2024-09-22 10:50:57,248 INFO [train.py:1266] (1/4) Training started
2024-09-22 10:50:57,249 INFO [train.py:1276] (1/4) Device: cuda:1
2024-09-22 10:50:57,251 INFO [train.py:1307] (1/4) Using dtype=torch.float16
2024-09-22 10:50:57,251 INFO [train.py:1308] (1/4) Use AMP=True
2024-09-22 10:50:57,251 INFO [train.py:1310] (1/4) {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'ignore_id': -1, 'label_smoothing': 0.1, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '44a9d5682af9fd3ef77074777e15278ec6d390eb', 'k2-git-date': 'Wed Sep 27 11:22:55 2023', 'lhotse-version': '1.17.0.dev+git.ccfc5b2c.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'cr-ctc', 'icefall-git-sha1': 'a6eead6c-clean', 'icefall-git-date': 'Mon Sep 9 10:10:08 2024', 'icefall-path': '/star-zw/workspace/zipformer/icefall_cr_ctc', 'k2-path': '/star-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/workspace/lhotse/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0905180047-6d6678bc6f-8cwvw', 'IP address': '10.30.5.48'}, 'world_size': 4, 'master_port': 12347, 'tensorboard': True, 'num_epochs': 50, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('zipformer/exp-cr-loss-scale-0.2-time-mask-ratio-2.5-scaled-masked-1-4'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'base_lr': 0.045, 'lr_batches': 7500, 'lr_epochs': 3.5, 'ref_duration': 600, 'context_size': 2, 'prune_range': 5, 'lm_scale': 0.25, 'am_scale': 0.0, 'simple_loss_scale': 0.5, 'ctc_loss_scale': 1.0, 'cr_loss_scale': 0.2, 'time_mask_ratio': 2.5, 'cr_loss_masked_scale': 1.0, 'attention_decoder_loss_scale': 0.8, 'seed': 42, 'print_diagnostics': False, 'inf_check': False, 'save_every_n': 4000, 'keep_last_k': 30, 'average_period': 200, 'use_fp16': True, 'use_bf16': False, 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'attention_decoder_dim': 512, 'attention_decoder_num_layers': 6, 'attention_decoder_attention_dim': 512, 'attention_decoder_num_heads': 8, 'attention_decoder_feedforward_dim': 2048, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': False, 'use_ctc': True, 'use_attention_decoder': False, 'use_cr_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 700, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': False, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'blank_id': 0, 'sos_id': 1, 'eos_id': 1, 'vocab_size': 500, 'dtype': torch.float16, 'use_autocast': True}
2024-09-22 10:50:57,251 INFO [train.py:1312] (1/4) About to create model
2024-09-22 10:50:57,900 INFO [train.py:1316] (1/4) Number of model parameters: 64250603
2024-09-22 10:50:57,901 INFO [train.py:752] (1/4) num_frame_masks: 25, max_frames_mask_fraction: 0.375
2024-09-22 10:51:02,812 INFO [train.py:1338] (1/4) Using DDP
2024-09-22 10:51:03,362 INFO [asr_datamodule.py:436] (1/4) About to get the shuffled train-clean-100,             train-clean-360 and train-other-500 cuts
2024-09-22 10:51:03,600 INFO [asr_datamodule.py:232] (1/4) Enable MUSAN
2024-09-22 10:51:03,600 INFO [asr_datamodule.py:233] (1/4) About to get Musan cuts
2024-09-22 10:51:05,244 INFO [asr_datamodule.py:279] (1/4) Disable SpecAugment
2024-09-22 10:51:05,244 INFO [asr_datamodule.py:281] (1/4) About to create train dataset
2024-09-22 10:51:05,244 INFO [asr_datamodule.py:308] (1/4) Using DynamicBucketingSampler.
2024-09-22 10:51:28,276 INFO [asr_datamodule.py:325] (1/4) About to create train dataloader
2024-09-22 10:51:28,277 INFO [asr_datamodule.py:453] (1/4) About to get dev-clean cuts
2024-09-22 10:51:28,278 INFO [asr_datamodule.py:460] (1/4) About to get dev-other cuts
2024-09-22 10:51:28,279 INFO [asr_datamodule.py:356] (1/4) About to create dev dataset
2024-09-22 10:51:28,478 INFO [asr_datamodule.py:373] (1/4) About to create dev dataloader
2024-09-22 10:51:28,478 INFO [train.py:1545] (1/4) Sanity check -- see if any of the batches in epoch 1 would cause OOM.
2024-09-22 10:55:04,876 INFO [train.py:1576] (1/4) Maximum memory allocated so far is 18729MB
2024-09-22 10:55:06,893 INFO [train.py:1576] (1/4) Maximum memory allocated so far is 18729MB
2024-09-22 10:55:09,107 INFO [train.py:1576] (1/4) Maximum memory allocated so far is 18997MB
2024-09-22 10:55:10,929 INFO [train.py:1576] (1/4) Maximum memory allocated so far is 18997MB
2024-09-22 10:55:13,009 INFO [train.py:1576] (1/4) Maximum memory allocated so far is 18997MB
2024-09-22 10:55:15,403 INFO [train.py:1576] (1/4) Maximum memory allocated so far is 18997MB
2024-09-22 10:56:01,485 INFO [train.py:1198] (1/4) Epoch 1, batch 0, loss[loss=5.004, ctc_loss=4.864, cr_loss=0.6964, over 17227.00 frames. ], tot_loss[loss=5.004, ctc_loss=4.864, cr_loss=0.6964, over 17227.00 frames. ], batch size: 50, lr: 2.25e-02, grad_scale: 2.0
2024-09-22 10:56:01,486 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 10:56:18,234 INFO [train.py:1230] (1/4) Epoch 1, validation: loss=4.756, ctc_loss=4.756, cr_loss=2.853e-15, over 944034.00 frames. 
2024-09-22 10:56:18,235 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 18997MB
2024-09-22 10:56:18,979 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.90 vs. limit=7.5
2024-09-22 10:56:20,873 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=6.41 vs. limit=7.5
2024-09-22 10:56:29,918 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=7.08 vs. limit=4.0
2024-09-22 10:56:36,876 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=9.68 vs. limit=7.5175
2024-09-22 10:56:38,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.min_positive, batch_count=46.666666666666664, ans=0.24953333333333333
2024-09-22 10:56:39,773 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.680e+03 4.888e+03 5.131e+03 6.578e+03 8.849e+03, threshold=2.053e+04, percent-clipped=0.0
2024-09-22 10:56:40,618 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=18.72 vs. limit=7.535
2024-09-22 10:56:55,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=93.33333333333333, ans=0.495625
2024-09-22 10:56:56,010 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=9.52 vs. limit=7.57
2024-09-22 10:56:59,062 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=12.18 vs. limit=7.535
2024-09-22 10:57:00,471 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.227e+03 2.818e+03 4.846e+03 6.578e+03 1.124e+04, threshold=1.938e+04, percent-clipped=0.0
2024-09-22 10:57:04,974 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=14.93 vs. limit=7.535
2024-09-22 10:57:32,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=20.84 vs. limit=7.64
2024-09-22 10:57:34,552 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=23.72 vs. limit=5.046666666666667
2024-09-22 10:57:36,833 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 8.493e+02 2.434e+03 3.497e+03 4.961e+03 1.124e+04, threshold=1.399e+04, percent-clipped=0.0
2024-09-22 10:57:38,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=186.66666666666666, ans=0.8934666666666667
2024-09-22 10:57:46,736 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=25.94 vs. limit=7.57
2024-09-22 10:57:48,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=186.66666666666666, ans=0.8934666666666667
2024-09-22 10:57:48,934 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=11.28 vs. limit=4.074666666666666
2024-09-22 10:57:51,513 INFO [train.py:1198] (1/4) Epoch 1, batch 50, loss[loss=1.448, ctc_loss=1.375, cr_loss=0.3659, over 17003.00 frames. ], tot_loss[loss=2.318, ctc_loss=2.255, cr_loss=0.3163, over 760262.29 frames. ], batch size: 51, lr: 2.48e-02, grad_scale: 0.5
2024-09-22 10:57:57,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=233.33333333333334, ans=0.09854166666666667
2024-09-22 10:58:00,000 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=15.82 vs. limit=5.058333333333334
2024-09-22 10:58:01,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=13.41 vs. limit=5.058333333333334
2024-09-22 10:58:01,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=22.70 vs. limit=7.675
2024-09-22 10:58:10,622 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=20.99 vs. limit=5.14
2024-09-22 10:58:14,703 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=21.64 vs. limit=7.605
2024-09-22 10:58:17,038 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=31.25 vs. limit=7.605
2024-09-22 10:58:43,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=326.6666666666667, ans=0.8885666666666667
2024-09-22 10:58:45,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=373.3333333333333, ans=0.0916
2024-09-22 10:58:55,154 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=23.42 vs. limit=7.64
2024-09-22 10:58:55,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.44 vs. limit=3.056
2024-09-22 10:59:02,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=17.94 vs. limit=7.64
2024-09-22 10:59:15,975 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=17.15 vs. limit=4.168
2024-09-22 10:59:16,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=58.41 vs. limit=7.6575
2024-09-22 10:59:24,437 INFO [train.py:1198] (1/4) Epoch 1, batch 100, loss[loss=1.275, ctc_loss=1.244, cr_loss=0.1527, over 14743.00 frames. ], tot_loss[loss=1.733, ctc_loss=1.683, cr_loss=0.2491, over 1326842.60 frames. ], batch size: 89, lr: 2.70e-02, grad_scale: 1.0
2024-09-22 10:59:28,076 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.417e+02 5.797e+02 1.227e+03 2.964e+03 1.124e+04, threshold=2.454e+03, percent-clipped=0.0
2024-09-22 10:59:45,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=28.72 vs. limit=7.885
2024-09-22 10:59:57,694 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=19.16 vs. limit=7.885
2024-09-22 10:59:59,262 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=25.04 vs. limit=7.6925
2024-09-22 11:00:04,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.max_abs, batch_count=560.0, ans=5.35
2024-09-22 11:00:10,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=51.41 vs. limit=7.71
2024-09-22 11:00:14,312 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=19.81 vs. limit=7.71
2024-09-22 11:00:14,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=23.64 vs. limit=7.92
2024-09-22 11:00:18,412 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=33.29 vs. limit=7.71
2024-09-22 11:00:25,666 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=23.06 vs. limit=7.7275
2024-09-22 11:00:31,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=16.05 vs. limit=5.303333333333334
2024-09-22 11:01:01,875 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=45.82 vs. limit=7.745
2024-09-22 11:01:04,606 INFO [train.py:1198] (1/4) Epoch 1, batch 150, loss[loss=1.255, ctc_loss=1.232, cr_loss=0.1129, over 17224.00 frames. ], tot_loss[loss=1.516, ctc_loss=1.477, cr_loss=0.1974, over 1775112.53 frames. ], batch size: 55, lr: 2.93e-02, grad_scale: 1.0
2024-09-22 11:01:07,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=11.84 vs. limit=7.7625
2024-09-22 11:01:09,254 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=22.93 vs. limit=5.35
2024-09-22 11:01:09,398 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=8.39 vs. limit=4.28
2024-09-22 11:01:27,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=746.6666666666666, ans=5.1866666666666665
2024-09-22 11:01:30,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=746.6666666666666, ans=0.46499999999999997
2024-09-22 11:01:42,342 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=10.09 vs. limit=5.198333333333333
2024-09-22 11:01:43,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=793.3333333333334, ans=0.4628125
2024-09-22 11:01:45,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=793.3333333333334, ans=0.4628125
2024-09-22 11:01:51,385 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=40.02 vs. limit=8.095
2024-09-22 11:01:59,066 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=30.01 vs. limit=7.815
2024-09-22 11:02:00,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=111.24 vs. limit=8.13
2024-09-22 11:02:04,811 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=21.64 vs. limit=7.815
2024-09-22 11:02:08,037 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=23.60 vs. limit=5.42
2024-09-22 11:02:10,595 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=19.72 vs. limit=7.815
2024-09-22 11:02:13,705 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.97 vs. limit=8.13
2024-09-22 11:02:28,907 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=11.22 vs. limit=7.8325
2024-09-22 11:02:30,877 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=13.19 vs. limit=5.221666666666667
2024-09-22 11:02:39,841 INFO [train.py:1198] (1/4) Epoch 1, batch 200, loss[loss=1.233, ctc_loss=1.206, cr_loss=0.1336, over 17113.00 frames. ], tot_loss[loss=1.404, ctc_loss=1.371, cr_loss=0.1688, over 2131102.37 frames. ], batch size: 49, lr: 3.15e-02, grad_scale: 2.0
2024-09-22 11:02:43,594 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 2.245e+02 3.033e+02 4.033e+02 1.104e+03, threshold=6.066e+02, percent-clipped=0.0
2024-09-22 11:02:44,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=120.84 vs. limit=7.85
2024-09-22 11:02:52,023 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.09 vs. limit=7.85
2024-09-22 11:03:00,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=980.0, ans=0.4540625
2024-09-22 11:03:04,555 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=118.49 vs. limit=7.8675
2024-09-22 11:03:22,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=1026.6666666666667, ans=0.28973333333333334
2024-09-22 11:03:23,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=1026.6666666666667, ans=5.256666666666667
2024-09-22 11:03:36,012 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.75 vs. limit=5.2683333333333335
2024-09-22 11:03:38,275 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=22.88 vs. limit=8.305
2024-09-22 11:03:40,239 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=51.07 vs. limit=7.9025
2024-09-22 11:03:43,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=1073.3333333333333, ans=0.15975
2024-09-22 11:03:44,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=1073.3333333333333, ans=0.8624333333333334
2024-09-22 11:03:58,353 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=24.98 vs. limit=7.92
2024-09-22 11:04:06,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=17.05 vs. limit=7.92
2024-09-22 11:04:07,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=1120.0, ans=0.4475
2024-09-22 11:04:12,401 INFO [train.py:1198] (1/4) Epoch 1, batch 250, loss[loss=1.183, ctc_loss=1.158, cr_loss=0.1286, over 16797.00 frames. ], tot_loss[loss=1.337, ctc_loss=1.306, cr_loss=0.1533, over 2397396.26 frames. ], batch size: 61, lr: 3.38e-02, grad_scale: 2.0
2024-09-22 11:04:14,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=1166.6666666666667, ans=0.35416666666666663
2024-09-22 11:04:27,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=1166.6666666666667, ans=0.04635416666666667
2024-09-22 11:04:29,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=16.84 vs. limit=7.955
2024-09-22 11:04:42,435 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=23.10 vs. limit=7.955
2024-09-22 11:04:44,735 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=28.69 vs. limit=7.955
2024-09-22 11:04:49,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=1260.0, ans=0.4409375
2024-09-22 11:04:51,831 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=156.92 vs. limit=7.9725
2024-09-22 11:05:00,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=1260.0, ans=0.2874
2024-09-22 11:05:36,892 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=8.866e+00
2024-09-22 11:05:38,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=1353.3333333333333, ans=0.4365625
2024-09-22 11:05:46,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=1400.0, ans=0.14750000000000002
2024-09-22 11:05:47,708 INFO [train.py:1198] (1/4) Epoch 1, batch 300, loss[loss=1.182, ctc_loss=1.151, cr_loss=0.1514, over 17216.00 frames. ], tot_loss[loss=1.297, ctc_loss=1.266, cr_loss=0.1516, over 2617372.79 frames. ], batch size: 47, lr: 3.60e-02, grad_scale: 4.0
2024-09-22 11:05:50,275 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=14.92 vs. limit=5.7
2024-09-22 11:05:50,545 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=11.92 vs. limit=8.55
2024-09-22 11:05:51,295 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.872e+02 2.621e+02 3.440e+02 6.626e+02, threshold=5.242e+02, percent-clipped=4.0
2024-09-22 11:05:57,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=1400.0, ans=0.286
2024-09-22 11:06:04,797 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=50.75 vs. limit=8.0425
2024-09-22 11:06:07,213 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=8.82 vs. limit=8.585
2024-09-22 11:06:08,770 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=63.98 vs. limit=8.0425
2024-09-22 11:06:14,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=1446.6666666666667, ans=0.4321875
2024-09-22 11:06:28,806 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.44 vs. limit=5.746666666666666
2024-09-22 11:06:34,500 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=11.82 vs. limit=8.620000000000001
2024-09-22 11:06:38,147 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=10.68 vs. limit=5.746666666666666
2024-09-22 11:06:41,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=1493.3333333333333, ans=0.31333333333333335
2024-09-22 11:06:44,157 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=13.70 vs. limit=8.06
2024-09-22 11:06:45,481 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=13.20 vs. limit=8.06
2024-09-22 11:06:52,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=1540.0, ans=0.3075
2024-09-22 11:07:00,346 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.21 vs. limit=8.655
2024-09-22 11:07:10,947 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=15.55 vs. limit=8.095
2024-09-22 11:07:20,871 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.28 vs. limit=3.238
2024-09-22 11:07:22,175 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.27 vs. limit=8.69
2024-09-22 11:07:25,016 INFO [train.py:1198] (1/4) Epoch 1, batch 350, loss[loss=1.06, ctc_loss=1.022, cr_loss=0.1902, over 17191.00 frames. ], tot_loss[loss=1.263, ctc_loss=1.23, cr_loss=0.1618, over 2786339.80 frames. ], batch size: 41, lr: 3.83e-02, grad_scale: 4.0
2024-09-22 11:07:50,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=1680.0, ans=0.42125
2024-09-22 11:07:51,201 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=91.64 vs. limit=8.13
2024-09-22 11:07:54,920 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=45.14 vs. limit=8.13
2024-09-22 11:07:54,982 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.13 vs. limit=8.76
2024-09-22 11:07:55,338 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=7.20 vs. limit=5.84
2024-09-22 11:08:09,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.66 vs. limit=8.795
2024-09-22 11:08:14,933 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=53.86 vs. limit=8.1475
2024-09-22 11:08:16,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=1726.6666666666667, ans=0.28273333333333334
2024-09-22 11:08:27,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=1773.3333333333333, ans=0.28226666666666667
2024-09-22 11:08:33,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=8.08 vs. limit=8.165
2024-09-22 11:08:49,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=1820.0, ans=0.8363
2024-09-22 11:08:55,232 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=49.91 vs. limit=8.1825
2024-09-22 11:09:00,231 INFO [train.py:1198] (1/4) Epoch 1, batch 400, loss[loss=1.092, ctc_loss=1.042, cr_loss=0.2475, over 17161.00 frames. ], tot_loss[loss=1.237, ctc_loss=1.201, cr_loss=0.1796, over 2913502.94 frames. ], batch size: 45, lr: 4.05e-02, grad_scale: 8.0
2024-09-22 11:09:03,043 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.84 vs. limit=8.2
2024-09-22 11:09:03,789 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.539e+02 3.204e+02 4.584e+02 1.114e+03, threshold=6.407e+02, percent-clipped=17.0
2024-09-22 11:09:35,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=1960.0, ans=0.2804
2024-09-22 11:09:42,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=1960.0, ans=0.08775000000000001
2024-09-22 11:09:48,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=1960.0, ans=0.408125
2024-09-22 11:09:57,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=5.15 vs. limit=4.802666666666667
2024-09-22 11:10:06,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=2006.6666666666667, ans=0.4059375
2024-09-22 11:10:07,062 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.53 vs. limit=9.004999999999999
2024-09-22 11:10:08,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=2006.6666666666667, ans=0.4059375
2024-09-22 11:10:10,539 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.48 vs. limit=9.004999999999999
2024-09-22 11:10:15,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=2053.3333333333335, ans=0.053799999999999994
2024-09-22 11:10:25,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=2053.3333333333335, ans=0.13449999999999998
2024-09-22 11:10:28,718 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=9.98 vs. limit=8.27
2024-09-22 11:10:31,077 INFO [train.py:1198] (1/4) Epoch 1, batch 450, loss[loss=1.138, ctc_loss=1.08, cr_loss=0.2896, over 17221.00 frames. ], tot_loss[loss=1.213, ctc_loss=1.173, cr_loss=0.2041, over 2997188.21 frames. ], batch size: 50, lr: 4.28e-02, grad_scale: 4.0
2024-09-22 11:10:45,187 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=38.33 vs. limit=8.2875
2024-09-22 11:10:48,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=2100.0, ans=0.771
2024-09-22 11:10:53,974 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.46 vs. limit=6.073333333333333
2024-09-22 11:11:03,512 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.21 vs. limit=8.305
2024-09-22 11:11:04,797 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.32 vs. limit=9.11
2024-09-22 11:11:24,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=2193.3333333333335, ans=0.3971875
2024-09-22 11:11:28,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=2240.0, ans=0.21999999999999997
2024-09-22 11:11:49,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.58 vs. limit=9.18
2024-09-22 11:11:51,045 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=19.30 vs. limit=8.3575
2024-09-22 11:11:55,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=2286.6666666666665, ans=0.2343
2024-09-22 11:12:03,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=2286.6666666666665, ans=0.11425
2024-09-22 11:12:05,552 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=5.72 vs. limit=4.914666666666666
2024-09-22 11:12:10,285 INFO [train.py:1198] (1/4) Epoch 1, batch 500, loss[loss=1.079, ctc_loss=1.016, cr_loss=0.3142, over 17088.00 frames. ], tot_loss[loss=1.18, ctc_loss=1.134, cr_loss=0.2319, over 3072730.47 frames. ], batch size: 49, lr: 4.49e-02, grad_scale: 8.0
2024-09-22 11:12:15,704 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.330e+02 2.678e+02 3.368e+02 4.496e+02 8.489e+02, threshold=6.737e+02, percent-clipped=3.0
2024-09-22 11:12:34,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=2380.0, ans=0.085125
2024-09-22 11:12:35,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=2380.0, ans=0.3884375
2024-09-22 11:12:50,967 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=8.23 vs. limit=8.41
2024-09-22 11:12:53,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=2426.6666666666665, ans=0.38625
2024-09-22 11:13:02,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=2473.3333333333335, ans=0.110875
2024-09-22 11:13:19,214 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.66 vs. limit=5.618333333333333
2024-09-22 11:13:42,938 INFO [train.py:1198] (1/4) Epoch 1, batch 550, loss[loss=0.922, ctc_loss=0.8542, cr_loss=0.339, over 17098.00 frames. ], tot_loss[loss=1.138, ctc_loss=1.086, cr_loss=0.2612, over 3122416.46 frames. ], batch size: 43, lr: 4.49e-02, grad_scale: 8.0
2024-09-22 11:13:43,770 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.42 vs. limit=8.4625
2024-09-22 11:13:59,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=2613.3333333333335, ans=0.3775
2024-09-22 11:14:06,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=2613.3333333333335, ans=0.3775
2024-09-22 11:14:11,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=2613.3333333333335, ans=0.8085333333333333
2024-09-22 11:14:50,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=2706.6666666666665, ans=0.37312500000000004
2024-09-22 11:14:51,558 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.49 vs. limit=8.515
2024-09-22 11:14:52,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=2753.3333333333335, ans=0.37093750000000003
2024-09-22 11:15:01,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=2753.3333333333335, ans=0.27246666666666663
2024-09-22 11:15:03,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=2753.3333333333335, ans=0.04139583333333333
2024-09-22 11:15:11,741 INFO [train.py:1198] (1/4) Epoch 1, batch 600, loss[loss=0.9007, ctc_loss=0.8139, cr_loss=0.434, over 17016.00 frames. ], tot_loss[loss=1.093, ctc_loss=1.034, cr_loss=0.2925, over 3157327.39 frames. ], batch size: 44, lr: 4.49e-02, grad_scale: 8.0
2024-09-22 11:15:12,712 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.18 vs. limit=8.55
2024-09-22 11:15:17,227 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.420e+02 3.324e+02 4.180e+02 8.567e+02, threshold=6.647e+02, percent-clipped=1.0
2024-09-22 11:15:23,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=2800.0, ans=0.037000000000000005
2024-09-22 11:15:28,581 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=2.594e-03
2024-09-22 11:15:37,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=2846.6666666666665, ans=0.22153333333333333
2024-09-22 11:16:04,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=2893.3333333333335, ans=0.2710666666666667
2024-09-22 11:16:12,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=2940.0, ans=0.7971
2024-09-22 11:16:19,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=2940.0, ans=0.08975
2024-09-22 11:16:30,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=2986.6666666666665, ans=0.36
2024-09-22 11:16:42,201 INFO [train.py:1198] (1/4) Epoch 1, batch 650, loss[loss=0.9239, ctc_loss=0.8427, cr_loss=0.4059, over 16563.00 frames. ], tot_loss[loss=1.045, ctc_loss=0.9799, cr_loss=0.3238, over 3208993.12 frames. ], batch size: 66, lr: 4.49e-02, grad_scale: 8.0
2024-09-22 11:17:26,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=3126.6666666666665, ans=0.08274999999999999
2024-09-22 11:17:31,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=3126.6666666666665, ans=0.029649999999999996
2024-09-22 11:17:36,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=3126.6666666666665, ans=0.08274999999999999
2024-09-22 11:17:49,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=3173.3333333333335, ans=0.26826666666666665
2024-09-22 11:17:49,879 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=6.06 vs. limit=5.793333333333333
2024-09-22 11:17:56,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=3220.0, ans=0.3490625
2024-09-22 11:18:05,464 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.38 vs. limit=5.805
2024-09-22 11:18:10,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=3220.0, ans=0.3490625
2024-09-22 11:18:14,946 INFO [train.py:1198] (1/4) Epoch 1, batch 700, loss[loss=0.73, ctc_loss=0.6458, cr_loss=0.4211, over 17039.00 frames. ], tot_loss[loss=0.9972, ctc_loss=0.9266, cr_loss=0.353, over 3248051.26 frames. ], batch size: 44, lr: 4.49e-02, grad_scale: 8.0
2024-09-22 11:18:20,264 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.357e+02 3.031e+02 4.358e+02 1.002e+03, threshold=6.062e+02, percent-clipped=7.0
2024-09-22 11:18:25,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=3266.6666666666665, ans=0.346875
2024-09-22 11:18:27,848 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.02 vs. limit=8.725
2024-09-22 11:18:37,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=3313.3333333333335, ans=0.07574999999999998
2024-09-22 11:18:53,552 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.08 vs. limit=10.02
2024-09-22 11:19:14,630 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.72 vs. limit=8.7775
2024-09-22 11:19:33,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=3453.3333333333335, ans=0.07049999999999998
2024-09-22 11:19:33,593 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=10.38 vs. limit=10.09
2024-09-22 11:19:44,683 INFO [train.py:1198] (1/4) Epoch 1, batch 750, loss[loss=0.8338, ctc_loss=0.7393, cr_loss=0.4727, over 17214.00 frames. ], tot_loss[loss=0.9529, ctc_loss=0.8776, cr_loss=0.3766, over 3271998.09 frames. ], batch size: 55, lr: 4.49e-02, grad_scale: 8.0
2024-09-22 11:19:59,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=3500.0, ans=0.3359375
2024-09-22 11:20:00,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=10.06 vs. limit=10.125
2024-09-22 11:20:00,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=3546.6666666666665, ans=0.7758666666666667
2024-09-22 11:20:17,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=3593.3333333333335, ans=0.3315625
2024-09-22 11:21:13,170 INFO [train.py:1198] (1/4) Epoch 1, batch 800, loss[loss=0.7622, ctc_loss=0.686, cr_loss=0.3809, over 17096.00 frames. ], tot_loss[loss=0.905, ctc_loss=0.8267, cr_loss=0.3914, over 3290904.93 frames. ], batch size: 49, lr: 4.49e-02, grad_scale: 16.0
2024-09-22 11:21:18,256 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.759e+02 2.614e+02 4.071e+02 6.304e+02 1.473e+03, threshold=8.142e+02, percent-clipped=26.0
2024-09-22 11:21:20,943 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.79 vs. limit=6.866666666666667
2024-09-22 11:21:25,923 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=14.84 vs. limit=10.3
2024-09-22 11:22:01,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=3826.6666666666665, ans=0.320625
2024-09-22 11:22:08,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=3873.3333333333335, ans=0.05474999999999999
2024-09-22 11:22:32,855 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.56 vs. limit=8.97
2024-09-22 11:22:39,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=3920.0, ans=0.05299999999999999
2024-09-22 11:22:44,172 INFO [train.py:1198] (1/4) Epoch 1, batch 850, loss[loss=0.7288, ctc_loss=0.649, cr_loss=0.3992, over 17096.00 frames. ], tot_loss[loss=0.8582, ctc_loss=0.7786, cr_loss=0.3979, over 3304010.58 frames. ], batch size: 49, lr: 4.49e-02, grad_scale: 16.0
2024-09-22 11:23:08,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=4013.3333333333335, ans=0.311875
2024-09-22 11:23:11,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=4013.3333333333335, ans=0.311875
2024-09-22 11:23:28,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=4060.0, ans=0.25939999999999996
2024-09-22 11:23:51,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=4106.666666666667, ans=0.2616
2024-09-22 11:24:02,157 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.36 vs. limit=9.0575
2024-09-22 11:24:02,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.11 vs. limit=10.615
2024-09-22 11:24:03,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=4153.333333333333, ans=0.009966666666666667
2024-09-22 11:24:11,781 INFO [train.py:1198] (1/4) Epoch 1, batch 900, loss[loss=0.62, ctc_loss=0.5333, cr_loss=0.4335, over 16926.00 frames. ], tot_loss[loss=0.8099, ctc_loss=0.7298, cr_loss=0.4007, over 3320883.65 frames. ], batch size: 42, lr: 4.48e-02, grad_scale: 16.0
2024-09-22 11:24:15,428 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 11:24:16,914 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.836e+02 3.761e+02 6.423e+02 1.326e+03, threshold=7.521e+02, percent-clipped=10.0
2024-09-22 11:24:34,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=4246.666666666667, ans=0.025
2024-09-22 11:25:13,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=4340.0, ans=0.2651
2024-09-22 11:25:17,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=4340.0, ans=0.2965625
2024-09-22 11:25:17,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=4340.0, ans=0.2965625
2024-09-22 11:25:32,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=4386.666666666667, ans=0.2561333333333333
2024-09-22 11:25:37,367 INFO [train.py:1198] (1/4) Epoch 1, batch 950, loss[loss=0.5838, ctc_loss=0.4986, cr_loss=0.4259, over 17278.00 frames. ], tot_loss[loss=0.7704, ctc_loss=0.6895, cr_loss=0.4046, over 3315984.79 frames. ], batch size: 46, lr: 4.48e-02, grad_scale: 16.0
2024-09-22 11:25:39,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=4433.333333333333, ans=0.009905797101449275
2024-09-22 11:25:51,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=4433.333333333333, ans=0.7943333333333333
2024-09-22 11:25:52,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=4480.0, ans=0.2552
2024-09-22 11:26:07,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=4480.0, ans=0.2552
2024-09-22 11:26:15,217 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.82 vs. limit=10.895
2024-09-22 11:26:21,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=4526.666666666667, ans=0.2547333333333333
2024-09-22 11:26:25,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=4526.666666666667, ans=0.07170833333333333
2024-09-22 11:26:43,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.78 vs. limit=9.215
2024-09-22 11:26:44,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=4573.333333333333, ans=0.7399333333333333
2024-09-22 11:26:53,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=4620.0, ans=0.009865217391304347
2024-09-22 11:27:05,077 INFO [train.py:1198] (1/4) Epoch 1, batch 1000, loss[loss=0.5618, ctc_loss=0.4764, cr_loss=0.4267, over 17144.00 frames. ], tot_loss[loss=0.7334, ctc_loss=0.6516, cr_loss=0.4089, over 3322622.20 frames. ], batch size: 45, lr: 4.48e-02, grad_scale: 16.0
2024-09-22 11:27:09,935 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.795e+02 3.933e+02 5.449e+02 1.373e+03, threshold=7.866e+02, percent-clipped=13.0
2024-09-22 11:27:24,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=4713.333333333333, ans=0.7350333333333334
2024-09-22 11:27:34,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=4713.333333333333, ans=0.2528666666666667
2024-09-22 11:27:36,044 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=5.66 vs. limit=5.8853333333333335
2024-09-22 11:27:37,159 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 11:27:56,291 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.99 vs. limit=11.07
2024-09-22 11:28:34,215 INFO [train.py:1198] (1/4) Epoch 1, batch 1050, loss[loss=0.6375, ctc_loss=0.543, cr_loss=0.4724, over 17008.00 frames. ], tot_loss[loss=0.6978, ctc_loss=0.6152, cr_loss=0.4126, over 3339926.32 frames. ], batch size: 51, lr: 4.48e-02, grad_scale: 16.0
2024-09-22 11:28:48,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=4900.0, ans=0.0
2024-09-22 11:28:58,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=4946.666666666667, ans=0.25053333333333333
2024-09-22 11:29:01,824 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.03 vs. limit=9.355
2024-09-22 11:29:44,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=5086.666666666667, ans=0.26156250000000003
2024-09-22 11:29:54,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=5086.666666666667, ans=0.035
2024-09-22 11:30:01,211 INFO [train.py:1198] (1/4) Epoch 1, batch 1100, loss[loss=0.5909, ctc_loss=0.5, cr_loss=0.4546, over 16559.00 frames. ], tot_loss[loss=0.6706, ctc_loss=0.5873, cr_loss=0.4165, over 3335404.04 frames. ], batch size: 66, lr: 4.48e-02, grad_scale: 16.0
2024-09-22 11:30:06,283 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.354e+02 3.241e+02 4.881e+02 1.077e+03, threshold=6.482e+02, percent-clipped=7.0
2024-09-22 11:30:21,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=5180.0, ans=0.2571875
2024-09-22 11:30:40,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=5226.666666666667, ans=0.24773333333333333
2024-09-22 11:30:41,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=5226.666666666667, ans=0.025
2024-09-22 11:31:19,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=5320.0, ans=0.250625
2024-09-22 11:31:25,796 INFO [train.py:1198] (1/4) Epoch 1, batch 1150, loss[loss=0.5079, ctc_loss=0.4237, cr_loss=0.4211, over 16967.00 frames. ], tot_loss[loss=0.6421, ctc_loss=0.5582, cr_loss=0.4193, over 3351730.10 frames. ], batch size: 42, lr: 4.47e-02, grad_scale: 16.0
2024-09-22 11:31:26,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=5366.666666666667, ans=0.03322916666666667
2024-09-22 11:31:29,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=5366.666666666667, ans=0.2805
2024-09-22 11:31:29,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.43 vs. limit=11.525
2024-09-22 11:31:32,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=5366.666666666667, ans=0.24843749999999998
2024-09-22 11:31:52,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=5413.333333333333, ans=0.24586666666666668
2024-09-22 11:32:14,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=5460.0, ans=0.009682608695652174
2024-09-22 11:32:26,434 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.59 vs. limit=9.565
2024-09-22 11:32:31,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=5506.666666666667, ans=0.2826
2024-09-22 11:32:58,122 INFO [train.py:1198] (1/4) Epoch 1, batch 1200, loss[loss=0.5803, ctc_loss=0.4921, cr_loss=0.4411, over 16963.00 frames. ], tot_loss[loss=0.6219, ctc_loss=0.5372, cr_loss=0.4235, over 3346938.92 frames. ], batch size: 58, lr: 4.47e-02, grad_scale: 32.0
2024-09-22 11:33:00,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=5600.0, ans=0.2375
2024-09-22 11:33:02,960 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.423e+02 3.283e+02 4.569e+02 8.108e+02, threshold=6.566e+02, percent-clipped=6.0
2024-09-22 11:33:04,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=5600.0, ans=0.2375
2024-09-22 11:33:05,607 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=13.03 vs. limit=11.7
2024-09-22 11:33:38,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=5693.333333333333, ans=0.24306666666666665
2024-09-22 11:33:51,146 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=1.016e-02
2024-09-22 11:34:03,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=5786.666666666667, ans=0.6974666666666667
2024-09-22 11:34:03,129 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=2.581e-03
2024-09-22 11:34:17,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=5786.666666666667, ans=0.2421333333333333
2024-09-22 11:34:18,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=5786.666666666667, ans=0.042555555555555555
2024-09-22 11:34:23,533 INFO [train.py:1198] (1/4) Epoch 1, batch 1250, loss[loss=0.4667, ctc_loss=0.3894, cr_loss=0.3866, over 17082.00 frames. ], tot_loss[loss=0.6021, ctc_loss=0.5169, cr_loss=0.4261, over 3348946.54 frames. ], batch size: 43, lr: 4.47e-02, grad_scale: 32.0
2024-09-22 11:34:39,624 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.11 vs. limit=11.91
2024-09-22 11:35:02,889 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.79 vs. limit=3.8890000000000002
2024-09-22 11:35:12,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=5973.333333333333, ans=0.0
2024-09-22 11:35:15,884 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.05 vs. limit=9.74
2024-09-22 11:35:47,097 INFO [train.py:1198] (1/4) Epoch 1, batch 1300, loss[loss=0.5398, ctc_loss=0.4498, cr_loss=0.45, over 16572.00 frames. ], tot_loss[loss=0.5838, ctc_loss=0.4982, cr_loss=0.4281, over 3345492.55 frames. ], batch size: 66, lr: 4.47e-02, grad_scale: 32.0
2024-09-22 11:35:52,195 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.150e+02 2.604e+02 3.500e+02 8.408e+02, threshold=5.208e+02, percent-clipped=5.0
2024-09-22 11:36:34,113 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-22 11:36:42,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=6206.666666666667, ans=0.23793333333333333
2024-09-22 11:37:09,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=6253.333333333333, ans=0.07
2024-09-22 11:37:12,778 INFO [train.py:1198] (1/4) Epoch 1, batch 1350, loss[loss=0.5787, ctc_loss=0.4872, cr_loss=0.4575, over 15017.00 frames. ], tot_loss[loss=0.5666, ctc_loss=0.4805, cr_loss=0.4306, over 3350353.36 frames. ], batch size: 89, lr: 4.46e-02, grad_scale: 32.0
2024-09-22 11:37:30,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=6346.666666666667, ans=12.26
2024-09-22 11:38:04,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=6393.333333333333, ans=0.2003125
2024-09-22 11:38:23,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=6486.666666666667, ans=0.2351333333333333
2024-09-22 11:38:32,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=6486.666666666667, ans=0.009459420289855072
2024-09-22 11:38:41,179 INFO [train.py:1198] (1/4) Epoch 1, batch 1400, loss[loss=0.4772, ctc_loss=0.3965, cr_loss=0.4035, over 17283.00 frames. ], tot_loss[loss=0.5517, ctc_loss=0.4654, cr_loss=0.4312, over 3342068.72 frames. ], batch size: 42, lr: 4.46e-02, grad_scale: 32.0
2024-09-22 11:38:41,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=6533.333333333333, ans=0.03944444444444445
2024-09-22 11:38:46,072 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.832e+02 2.453e+02 3.278e+02 5.014e+02 1.044e+03, threshold=6.556e+02, percent-clipped=21.0
2024-09-22 11:38:51,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=6533.333333333333, ans=0.19374999999999998
2024-09-22 11:39:01,990 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=10.88 vs. limit=12.434999999999999
2024-09-22 11:39:11,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=6580.0, ans=0.2342
2024-09-22 11:39:21,664 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.24 vs. limit=9.985
2024-09-22 11:39:36,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=6673.333333333333, ans=0.23326666666666668
2024-09-22 11:40:05,929 INFO [train.py:1198] (1/4) Epoch 1, batch 1450, loss[loss=0.515, ctc_loss=0.4175, cr_loss=0.4877, over 17039.00 frames. ], tot_loss[loss=0.5407, ctc_loss=0.4541, cr_loss=0.4328, over 3344797.69 frames. ], batch size: 52, lr: 4.46e-02, grad_scale: 32.0
2024-09-22 11:40:12,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=6766.666666666667, ans=0.1828125
2024-09-22 11:40:14,862 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.78 vs. limit=10.0375
2024-09-22 11:40:21,274 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.99 vs. limit=10.055
2024-09-22 11:40:26,237 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.16 vs. limit=12.61
2024-09-22 11:40:36,215 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.09 vs. limit=10.055
2024-09-22 11:40:48,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=6860.0, ans=0.0
2024-09-22 11:41:23,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=6953.333333333333, ans=0.03769444444444445
2024-09-22 11:41:27,762 INFO [train.py:1198] (1/4) Epoch 1, batch 1500, loss[loss=0.5734, ctc_loss=0.4651, cr_loss=0.5414, over 16016.00 frames. ], tot_loss[loss=0.5288, ctc_loss=0.4422, cr_loss=0.433, over 3340583.15 frames. ], batch size: 74, lr: 4.46e-02, grad_scale: 32.0
2024-09-22 11:41:30,349 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.99 vs. limit=10.125
2024-09-22 11:41:31,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=7000.0, ans=0.037500000000000006
2024-09-22 11:41:32,711 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.217e+02 3.113e+02 4.719e+02 9.117e+02, threshold=6.226e+02, percent-clipped=8.0
2024-09-22 11:41:36,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=7000.0, ans=0.171875
2024-09-22 11:41:51,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=7046.666666666667, ans=0.1696875
2024-09-22 11:42:00,704 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.62 vs. limit=12.785
2024-09-22 11:42:05,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=7093.333333333333, ans=0.009327536231884058
2024-09-22 11:42:12,494 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.72 vs. limit=4.064
2024-09-22 11:42:21,957 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.57 vs. limit=8.57
2024-09-22 11:42:34,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=7186.666666666667, ans=0.16312500000000002
2024-09-22 11:42:34,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=7186.666666666667, ans=0.036722222222222226
2024-09-22 11:42:39,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.06 vs. limit=10.195
2024-09-22 11:42:56,582 INFO [train.py:1198] (1/4) Epoch 1, batch 1550, loss[loss=0.5058, ctc_loss=0.4175, cr_loss=0.4414, over 17130.00 frames. ], tot_loss[loss=0.5165, ctc_loss=0.4299, cr_loss=0.4329, over 3351627.16 frames. ], batch size: 48, lr: 4.45e-02, grad_scale: 32.0
2024-09-22 11:43:18,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=7280.0, ans=9.55
2024-09-22 11:43:42,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=7326.666666666667, ans=0.22673333333333334
2024-09-22 11:43:45,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=7373.333333333333, ans=0.15437499999999998
2024-09-22 11:44:00,894 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.06 vs. limit=13.065000000000001
2024-09-22 11:44:03,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=7420.0, ans=0.009256521739130434
2024-09-22 11:44:03,978 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=7420.0, ans=0.025
2024-09-22 11:44:18,842 INFO [train.py:1198] (1/4) Epoch 1, batch 1600, loss[loss=0.3918, ctc_loss=0.3166, cr_loss=0.376, over 16217.00 frames. ], tot_loss[loss=0.507, ctc_loss=0.4202, cr_loss=0.4339, over 3360402.58 frames. ], batch size: 36, lr: 4.45e-02, grad_scale: 32.0
2024-09-22 11:44:22,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=7466.666666666667, ans=0.6386666666666667
2024-09-22 11:44:23,627 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.041e+02 2.519e+02 3.425e+02 6.490e+02, threshold=5.038e+02, percent-clipped=3.0
2024-09-22 11:44:59,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=7.17 vs. limit=10.335
2024-09-22 11:45:01,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten.whitening_limit, batch_count=7560.0, ans=13.17
2024-09-22 11:45:03,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=7560.0, ans=0.145625
2024-09-22 11:45:21,816 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.85 vs. limit=13.205
2024-09-22 11:45:27,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=7653.333333333333, ans=0.009205797101449276
2024-09-22 11:45:35,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=7653.333333333333, ans=0.14125
2024-09-22 11:45:42,058 INFO [train.py:1198] (1/4) Epoch 1, batch 1650, loss[loss=0.5283, ctc_loss=0.4335, cr_loss=0.4738, over 17208.00 frames. ], tot_loss[loss=0.4991, ctc_loss=0.4121, cr_loss=0.4349, over 3354084.98 frames. ], batch size: 55, lr: 4.45e-02, grad_scale: 32.0
2024-09-22 11:45:42,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=7700.0, ans=0.6305000000000001
2024-09-22 11:45:52,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=7700.0, ans=0.009195652173913044
2024-09-22 11:45:55,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=7700.0, ans=0.025
2024-09-22 11:45:58,997 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.08 vs. limit=10.405
2024-09-22 11:46:10,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=7746.666666666667, ans=0.034388888888888886
2024-09-22 11:46:24,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=7793.333333333333, ans=0.13468750000000002
2024-09-22 11:46:44,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=7840.0, ans=0.009165217391304348
2024-09-22 11:47:05,462 INFO [train.py:1198] (1/4) Epoch 1, batch 1700, loss[loss=0.4911, ctc_loss=0.3976, cr_loss=0.4674, over 17142.00 frames. ], tot_loss[loss=0.4947, ctc_loss=0.4073, cr_loss=0.4369, over 3347805.42 frames. ], batch size: 48, lr: 4.44e-02, grad_scale: 32.0
2024-09-22 11:47:10,186 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.014e+02 2.718e+02 3.727e+02 5.677e+02, threshold=5.436e+02, percent-clipped=4.0
2024-09-22 11:47:15,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=7933.333333333333, ans=0.128125
2024-09-22 11:47:19,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=7980.0, ans=0.12593749999999998
2024-09-22 11:47:28,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=7980.0, ans=0.03341666666666667
2024-09-22 11:47:33,291 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.41 vs. limit=10.4925
2024-09-22 11:48:01,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=8073.333333333333, ans=0.125
2024-09-22 11:48:14,921 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.94 vs. limit=10.545
2024-09-22 11:48:18,370 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.92 vs. limit=13.59
2024-09-22 11:48:29,758 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.84 vs. limit=13.625
2024-09-22 11:48:30,595 INFO [train.py:1198] (1/4) Epoch 1, batch 1750, loss[loss=0.4161, ctc_loss=0.3272, cr_loss=0.4446, over 17081.00 frames. ], tot_loss[loss=0.4889, ctc_loss=0.4013, cr_loss=0.438, over 3343595.56 frames. ], batch size: 43, lr: 4.44e-02, grad_scale: 32.0
2024-09-22 11:49:02,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=8260.0, ans=10.0
2024-09-22 11:49:13,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=8260.0, ans=0.125
2024-09-22 11:49:27,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=8306.666666666666, ans=0.6092666666666667
2024-09-22 11:49:31,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=8306.666666666666, ans=0.3246
2024-09-22 11:49:45,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=8353.333333333334, ans=0.6076333333333334
2024-09-22 11:49:53,985 INFO [train.py:1198] (1/4) Epoch 1, batch 1800, loss[loss=0.4855, ctc_loss=0.3911, cr_loss=0.4722, over 17204.00 frames. ], tot_loss[loss=0.4837, ctc_loss=0.3959, cr_loss=0.4387, over 3339210.15 frames. ], batch size: 55, lr: 4.44e-02, grad_scale: 32.0
2024-09-22 11:49:56,336 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.67 vs. limit=7.1
2024-09-22 11:49:58,878 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 1.919e+02 2.429e+02 3.208e+02 6.110e+02, threshold=4.858e+02, percent-clipped=5.0
2024-09-22 11:50:20,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=8446.666666666666, ans=0.03147222222222223
2024-09-22 11:50:29,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=8493.333333333334, ans=0.125
2024-09-22 11:50:37,950 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.16 vs. limit=10.685
2024-09-22 11:51:14,591 INFO [train.py:1198] (1/4) Epoch 1, batch 1850, loss[loss=0.4959, ctc_loss=0.4044, cr_loss=0.4578, over 17296.00 frames. ], tot_loss[loss=0.4796, ctc_loss=0.3916, cr_loss=0.4398, over 3335845.25 frames. ], batch size: 49, lr: 4.43e-02, grad_scale: 32.0
2024-09-22 11:51:23,931 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=11.03 vs. limit=13.975000000000001
2024-09-22 11:51:38,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.73 vs. limit=5.0
2024-09-22 11:51:51,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=8726.666666666666, ans=0.025
2024-09-22 11:52:29,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.23 vs. limit=4.323
2024-09-22 11:52:39,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=8866.666666666666, ans=0.125
2024-09-22 11:52:40,377 INFO [train.py:1198] (1/4) Epoch 1, batch 1900, loss[loss=0.4459, ctc_loss=0.3618, cr_loss=0.4205, over 16477.00 frames. ], tot_loss[loss=0.472, ctc_loss=0.3843, cr_loss=0.4382, over 3339339.60 frames. ], batch size: 66, lr: 4.43e-02, grad_scale: 32.0
2024-09-22 11:52:44,917 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 1.950e+02 2.731e+02 3.550e+02 1.054e+03, threshold=5.462e+02, percent-clipped=8.0
2024-09-22 11:53:08,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=14.69 vs. limit=14.184999999999999
2024-09-22 11:53:15,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=8960.0, ans=0.125
2024-09-22 11:53:25,364 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.30 vs. limit=10.86
2024-09-22 11:53:42,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=9006.666666666666, ans=0.029138888888888895
2024-09-22 11:54:02,942 INFO [train.py:1198] (1/4) Epoch 1, batch 1950, loss[loss=0.4875, ctc_loss=0.3922, cr_loss=0.4767, over 16993.00 frames. ], tot_loss[loss=0.4684, ctc_loss=0.3806, cr_loss=0.4389, over 3343475.89 frames. ], batch size: 56, lr: 4.43e-02, grad_scale: 32.0
2024-09-22 11:54:06,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=9100.0, ans=0.125
2024-09-22 11:54:13,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=9100.0, ans=0.125
2024-09-22 11:54:55,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=9240.0, ans=0.125
2024-09-22 11:55:01,921 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 11:55:08,763 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.36 vs. limit=4.393
2024-09-22 11:55:10,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.88 vs. limit=14.465
2024-09-22 11:55:16,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=9286.666666666666, ans=0.027972222222222228
2024-09-22 11:55:26,419 INFO [train.py:1198] (1/4) Epoch 1, batch 2000, loss[loss=0.3695, ctc_loss=0.2981, cr_loss=0.3568, over 17119.00 frames. ], tot_loss[loss=0.463, ctc_loss=0.3751, cr_loss=0.4393, over 3355079.17 frames. ], batch size: 40, lr: 4.42e-02, grad_scale: 32.0
2024-09-22 11:55:31,236 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 1.918e+02 2.442e+02 3.354e+02 7.763e+02, threshold=4.883e+02, percent-clipped=5.0
2024-09-22 11:55:44,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=9380.0, ans=0.2062
2024-09-22 11:56:03,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=9426.666666666666, ans=0.20573333333333332
2024-09-22 11:56:15,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=9473.333333333334, ans=0.125
2024-09-22 11:56:27,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=9473.333333333334, ans=0.125
2024-09-22 11:56:42,224 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.76 vs. limit=14.64
2024-09-22 11:56:49,511 INFO [train.py:1198] (1/4) Epoch 1, batch 2050, loss[loss=0.3831, ctc_loss=0.3068, cr_loss=0.3818, over 16277.00 frames. ], tot_loss[loss=0.4591, ctc_loss=0.3713, cr_loss=0.4391, over 3351216.34 frames. ], batch size: 36, lr: 4.42e-02, grad_scale: 32.0
2024-09-22 11:56:49,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=9566.666666666666, ans=0.125
2024-09-22 11:57:23,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=9660.0, ans=0.025
2024-09-22 11:57:27,580 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.96 vs. limit=7.415
2024-09-22 11:58:05,909 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.57 vs. limit=14.815000000000001
2024-09-22 11:58:08,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=9753.333333333334, ans=0.125
2024-09-22 11:58:14,743 INFO [train.py:1198] (1/4) Epoch 1, batch 2100, loss[loss=0.4405, ctc_loss=0.3546, cr_loss=0.4294, over 17232.00 frames. ], tot_loss[loss=0.4538, ctc_loss=0.3661, cr_loss=0.4386, over 3353580.85 frames. ], batch size: 50, lr: 4.42e-02, grad_scale: 32.0
2024-09-22 11:58:15,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=9800.0, ans=0.125
2024-09-22 11:58:19,529 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 1.883e+02 2.281e+02 3.077e+02 7.464e+02, threshold=4.562e+02, percent-clipped=6.0
2024-09-22 11:58:50,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=9893.333333333334, ans=0.15106666666666665
2024-09-22 11:59:14,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=9940.0, ans=0.5521
2024-09-22 11:59:29,204 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.08 vs. limit=11.245000000000001
2024-09-22 11:59:37,506 INFO [train.py:1198] (1/4) Epoch 1, batch 2150, loss[loss=0.4721, ctc_loss=0.3721, cr_loss=0.5, over 17225.00 frames. ], tot_loss[loss=0.4497, ctc_loss=0.3622, cr_loss=0.4376, over 3359909.67 frames. ], batch size: 55, lr: 4.41e-02, grad_scale: 32.0
2024-09-22 11:59:44,436 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.01 vs. limit=11.2625
2024-09-22 12:00:11,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=10126.666666666666, ans=0.125
2024-09-22 12:00:12,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=10126.666666666666, ans=0.5455666666666668
2024-09-22 12:00:22,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=10126.666666666666, ans=0.125
2024-09-22 12:00:23,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=10173.333333333334, ans=0.8517333333333333
2024-09-22 12:00:35,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=10173.333333333334, ans=0.025
2024-09-22 12:00:41,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=10220.0, ans=0.0
2024-09-22 12:00:44,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=10220.0, ans=0.1978
2024-09-22 12:00:55,359 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.06 vs. limit=11.3325
2024-09-22 12:00:58,121 INFO [train.py:1198] (1/4) Epoch 1, batch 2200, loss[loss=0.3982, ctc_loss=0.3162, cr_loss=0.41, over 16954.00 frames. ], tot_loss[loss=0.446, ctc_loss=0.3586, cr_loss=0.4369, over 3358847.15 frames. ], batch size: 42, lr: 4.41e-02, grad_scale: 32.0
2024-09-22 12:00:58,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=10266.666666666666, ans=0.125
2024-09-22 12:01:02,862 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.024e+02 2.529e+02 3.777e+02 5.736e+02, threshold=5.059e+02, percent-clipped=14.0
2024-09-22 12:01:08,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=10266.666666666666, ans=0.0
2024-09-22 12:01:11,219 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=10266.666666666666, ans=0.125
2024-09-22 12:01:24,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=10313.333333333334, ans=0.0
2024-09-22 12:01:25,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=10313.333333333334, ans=0.125
2024-09-22 12:02:17,121 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.27 vs. limit=4.568
2024-09-22 12:02:21,445 INFO [train.py:1198] (1/4) Epoch 1, batch 2250, loss[loss=0.3723, ctc_loss=0.2951, cr_loss=0.3863, over 15800.00 frames. ], tot_loss[loss=0.4432, ctc_loss=0.3558, cr_loss=0.4372, over 3353373.59 frames. ], batch size: 35, lr: 4.40e-02, grad_scale: 32.0
2024-09-22 12:02:32,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=10500.0, ans=0.195
2024-09-22 12:02:42,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=10546.666666666666, ans=0.022722222222222227
2024-09-22 12:02:42,897 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.07 vs. limit=11.455
2024-09-22 12:03:46,762 INFO [train.py:1198] (1/4) Epoch 1, batch 2300, loss[loss=0.3981, ctc_loss=0.3102, cr_loss=0.4393, over 17042.00 frames. ], tot_loss[loss=0.4388, ctc_loss=0.3516, cr_loss=0.4361, over 3359258.29 frames. ], batch size: 39, lr: 4.40e-02, grad_scale: 32.0
2024-09-22 12:03:51,576 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 1.850e+02 2.386e+02 2.971e+02 5.038e+02, threshold=4.772e+02, percent-clipped=0.0
2024-09-22 12:03:56,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=10733.333333333334, ans=0.125
2024-09-22 12:04:18,213 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.50 vs. limit=11.559999999999999
2024-09-22 12:04:43,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=10873.333333333334, ans=0.02136111111111111
2024-09-22 12:04:50,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=10873.333333333334, ans=0.5194333333333334
2024-09-22 12:04:57,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=10920.0, ans=0.09899494936611666
2024-09-22 12:05:09,823 INFO [train.py:1198] (1/4) Epoch 1, batch 2350, loss[loss=0.4247, ctc_loss=0.3347, cr_loss=0.4499, over 17356.00 frames. ], tot_loss[loss=0.4347, ctc_loss=0.3476, cr_loss=0.4352, over 3362146.49 frames. ], batch size: 48, lr: 4.40e-02, grad_scale: 32.0
2024-09-22 12:05:27,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=11013.333333333334, ans=0.008475362318840579
2024-09-22 12:05:32,523 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.20 vs. limit=11.629999999999999
2024-09-22 12:05:35,761 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.33 vs. limit=10.506666666666668
2024-09-22 12:05:46,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=11060.0, ans=0.125
2024-09-22 12:05:49,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=11060.0, ans=0.18939999999999999
2024-09-22 12:05:51,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.80 vs. limit=4.659
2024-09-22 12:06:20,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=11153.333333333334, ans=0.0
2024-09-22 12:06:27,226 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.77 vs. limit=15.865
2024-09-22 12:06:30,103 INFO [train.py:1198] (1/4) Epoch 1, batch 2400, loss[loss=0.4113, ctc_loss=0.3224, cr_loss=0.4445, over 17013.00 frames. ], tot_loss[loss=0.433, ctc_loss=0.3457, cr_loss=0.4363, over 3356107.20 frames. ], batch size: 44, lr: 4.39e-02, grad_scale: 32.0
2024-09-22 12:06:35,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=11200.0, ans=0.020000000000000004
2024-09-22 12:06:37,258 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 1.777e+02 2.025e+02 2.571e+02 5.493e+02, threshold=4.051e+02, percent-clipped=2.0
2024-09-22 12:06:37,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=11200.0, ans=0.125
2024-09-22 12:06:37,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=11200.0, ans=0.125
2024-09-22 12:06:59,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=11246.666666666666, ans=0.025
2024-09-22 12:07:03,372 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=16.55 vs. limit=15.97
2024-09-22 12:07:31,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=11340.0, ans=0.07
2024-09-22 12:07:57,471 INFO [train.py:1198] (1/4) Epoch 1, batch 2450, loss[loss=0.4624, ctc_loss=0.3719, cr_loss=0.4521, over 17002.00 frames. ], tot_loss[loss=0.4314, ctc_loss=0.3442, cr_loss=0.4362, over 3348874.58 frames. ], batch size: 53, lr: 4.39e-02, grad_scale: 64.0
2024-09-22 12:07:59,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=11433.333333333334, ans=0.019027777777777775
2024-09-22 12:08:08,213 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.85 vs. limit=11.7875
2024-09-22 12:08:18,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=11480.0, ans=0.125
2024-09-22 12:08:21,198 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=4.28 vs. limit=8.591999999999999
2024-09-22 12:08:37,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer_ff3.min_abs, batch_count=11526.666666666666, ans=0.2
2024-09-22 12:08:41,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=11526.666666666666, ans=0.125
2024-09-22 12:09:09,425 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.81 vs. limit=16.215
2024-09-22 12:09:10,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=11620.0, ans=0.125
2024-09-22 12:09:11,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=11620.0, ans=0.49330000000000007
2024-09-22 12:09:14,637 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.95 vs. limit=4.743
2024-09-22 12:09:18,266 INFO [train.py:1198] (1/4) Epoch 1, batch 2500, loss[loss=0.4593, ctc_loss=0.3608, cr_loss=0.4925, over 16922.00 frames. ], tot_loss[loss=0.4279, ctc_loss=0.341, cr_loss=0.4343, over 3337471.77 frames. ], batch size: 58, lr: 4.38e-02, grad_scale: 64.0
2024-09-22 12:09:22,320 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.93 vs. limit=11.875
2024-09-22 12:09:22,935 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.074e+02 2.928e+02 4.593e+02 9.871e+02, threshold=5.856e+02, percent-clipped=30.0
2024-09-22 12:09:37,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.81 vs. limit=8.685333333333332
2024-09-22 12:09:46,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=11713.333333333334, ans=0.017861111111111105
2024-09-22 12:09:54,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=11760.0, ans=0.125
2024-09-22 12:10:06,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=11760.0, ans=0.1824
2024-09-22 12:10:40,845 INFO [train.py:1198] (1/4) Epoch 1, batch 2550, loss[loss=0.474, ctc_loss=0.3865, cr_loss=0.4378, over 14954.00 frames. ], tot_loss[loss=0.4281, ctc_loss=0.341, cr_loss=0.4359, over 3346185.80 frames. ], batch size: 89, lr: 4.38e-02, grad_scale: 64.0
2024-09-22 12:10:44,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=11900.0, ans=0.181
2024-09-22 12:10:46,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=11900.0, ans=0.125
2024-09-22 12:10:49,461 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=9.83 vs. limit=11.9625
2024-09-22 12:11:12,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.59 vs. limit=11.997499999999999
2024-09-22 12:11:24,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=11993.333333333334, ans=0.016694444444444442
2024-09-22 12:11:29,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=12040.0, ans=0.17959999999999998
2024-09-22 12:11:52,163 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=16.00 vs. limit=16.564999999999998
2024-09-22 12:12:02,914 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.81 vs. limit=12.05
2024-09-22 12:12:04,477 INFO [train.py:1198] (1/4) Epoch 1, batch 2600, loss[loss=0.4303, ctc_loss=0.3468, cr_loss=0.4173, over 17355.00 frames. ], tot_loss[loss=0.4232, ctc_loss=0.3365, cr_loss=0.4337, over 3348047.28 frames. ], batch size: 48, lr: 4.37e-02, grad_scale: 64.0
2024-09-22 12:12:09,352 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 1.984e+02 2.668e+02 3.339e+02 5.918e+02, threshold=5.335e+02, percent-clipped=1.0
2024-09-22 12:12:29,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=12180.0, ans=0.025
2024-09-22 12:12:31,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=12180.0, ans=0.125
2024-09-22 12:12:43,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=12226.666666666666, ans=0.008211594202898551
2024-09-22 12:12:50,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=12226.666666666666, ans=0.125
2024-09-22 12:12:51,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=12226.666666666666, ans=0.125
2024-09-22 12:12:57,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=12273.333333333334, ans=0.015527777777777772
2024-09-22 12:13:29,613 INFO [train.py:1198] (1/4) Epoch 1, batch 2650, loss[loss=0.4329, ctc_loss=0.3364, cr_loss=0.4826, over 17140.00 frames. ], tot_loss[loss=0.42, ctc_loss=0.3334, cr_loss=0.4331, over 3357182.07 frames. ], batch size: 45, lr: 4.37e-02, grad_scale: 64.0
2024-09-22 12:13:39,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=12366.666666666666, ans=0.125
2024-09-22 12:13:43,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=12366.666666666666, ans=0.125
2024-09-22 12:13:52,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=12413.333333333334, ans=0.17586666666666667
2024-09-22 12:13:53,012 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.60 vs. limit=8.103333333333333
2024-09-22 12:13:55,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=12413.333333333334, ans=0.025
2024-09-22 12:14:13,990 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.99 vs. limit=16.845
2024-09-22 12:14:51,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=12600.0, ans=0.08345000000000001
2024-09-22 12:14:52,683 INFO [train.py:1198] (1/4) Epoch 1, batch 2700, loss[loss=0.419, ctc_loss=0.3332, cr_loss=0.429, over 17091.00 frames. ], tot_loss[loss=0.4194, ctc_loss=0.3326, cr_loss=0.4342, over 3355574.68 frames. ], batch size: 49, lr: 4.36e-02, grad_scale: 64.0
2024-09-22 12:14:54,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=12600.0, ans=0.125
2024-09-22 12:14:57,514 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 1.915e+02 2.536e+02 3.410e+02 5.700e+02, threshold=5.072e+02, percent-clipped=2.0
2024-09-22 12:15:15,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=12646.666666666666, ans=0.013972222222222226
2024-09-22 12:15:47,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=12740.0, ans=0.125
2024-09-22 12:16:12,417 INFO [train.py:1198] (1/4) Epoch 1, batch 2750, loss[loss=0.3781, ctc_loss=0.3013, cr_loss=0.3841, over 16707.00 frames. ], tot_loss[loss=0.4176, ctc_loss=0.331, cr_loss=0.433, over 3342132.84 frames. ], batch size: 37, lr: 4.36e-02, grad_scale: 64.0
2024-09-22 12:16:39,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=12880.0, ans=0.125
2024-09-22 12:16:44,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=12880.0, ans=0.125
2024-09-22 12:17:01,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=12973.333333333334, ans=0.125
2024-09-22 12:17:04,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=12973.333333333334, ans=0.0
2024-09-22 12:17:13,069 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=6.46 vs. limit=9.189333333333334
2024-09-22 12:17:18,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=12973.333333333334, ans=0.012611111111111108
2024-09-22 12:17:23,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=13020.0, ans=0.125
2024-09-22 12:17:34,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=13020.0, ans=0.125
2024-09-22 12:17:36,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=13020.0, ans=0.44430000000000003
2024-09-22 12:17:40,888 INFO [train.py:1198] (1/4) Epoch 1, batch 2800, loss[loss=0.3421, ctc_loss=0.2612, cr_loss=0.4046, over 16920.00 frames. ], tot_loss[loss=0.4147, ctc_loss=0.3283, cr_loss=0.4324, over 3347984.89 frames. ], batch size: 42, lr: 4.36e-02, grad_scale: 64.0
2024-09-22 12:17:45,584 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 1.867e+02 2.104e+02 2.772e+02 6.258e+02, threshold=4.209e+02, percent-clipped=2.0
2024-09-22 12:17:50,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=13066.666666666666, ans=0.008028985507246377
2024-09-22 12:18:08,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=13113.333333333334, ans=0.008018840579710146
2024-09-22 12:18:50,369 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.31 vs. limit=12.469999999999999
2024-09-22 12:18:59,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=13300.0, ans=0.007978260869565218
2024-09-22 12:19:00,770 INFO [train.py:1198] (1/4) Epoch 1, batch 2850, loss[loss=0.4111, ctc_loss=0.3224, cr_loss=0.4437, over 17050.00 frames. ], tot_loss[loss=0.4144, ctc_loss=0.3276, cr_loss=0.4337, over 3348206.95 frames. ], batch size: 56, lr: 4.35e-02, grad_scale: 32.0
2024-09-22 12:19:20,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=13346.666666666666, ans=0.125
2024-09-22 12:19:32,848 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.72 vs. limit=5.002
2024-09-22 12:19:43,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=13393.333333333334, ans=0.010861111111111106
2024-09-22 12:20:23,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=13533.333333333334, ans=0.125
2024-09-22 12:20:24,535 INFO [train.py:1198] (1/4) Epoch 1, batch 2900, loss[loss=0.3804, ctc_loss=0.2956, cr_loss=0.4243, over 17210.00 frames. ], tot_loss[loss=0.4114, ctc_loss=0.3248, cr_loss=0.4328, over 3353301.50 frames. ], batch size: 47, lr: 4.35e-02, grad_scale: 32.0
2024-09-22 12:20:31,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 1.863e+02 2.342e+02 3.249e+02 5.939e+02, threshold=4.685e+02, percent-clipped=7.0
2024-09-22 12:20:39,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=13580.0, ans=0.125
2024-09-22 12:21:02,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=13626.666666666666, ans=0.0
2024-09-22 12:21:15,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=13673.333333333334, ans=0.125
2024-09-22 12:21:36,523 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.79 vs. limit=17.79
2024-09-22 12:21:39,487 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.35 vs. limit=12.645
2024-09-22 12:21:47,308 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.44 vs. limit=17.825
2024-09-22 12:21:48,333 INFO [train.py:1198] (1/4) Epoch 1, batch 2950, loss[loss=0.3796, ctc_loss=0.2948, cr_loss=0.4241, over 17217.00 frames. ], tot_loss[loss=0.4088, ctc_loss=0.3225, cr_loss=0.4314, over 3356571.16 frames. ], batch size: 50, lr: 4.34e-02, grad_scale: 32.0
2024-09-22 12:22:53,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.58 vs. limit=8.476666666666667
2024-09-22 12:23:13,840 INFO [train.py:1198] (1/4) Epoch 1, batch 3000, loss[loss=0.4013, ctc_loss=0.3098, cr_loss=0.4579, over 17016.00 frames. ], tot_loss[loss=0.4079, ctc_loss=0.3216, cr_loss=0.4315, over 3349727.47 frames. ], batch size: 44, lr: 4.34e-02, grad_scale: 32.0
2024-09-22 12:23:13,841 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 12:23:29,187 INFO [train.py:1230] (1/4) Epoch 1, validation: loss=0.1235, ctc_loss=0.1235, cr_loss=7.044e-15, over 944034.00 frames. 
2024-09-22 12:23:29,188 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 12:23:35,639 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 1.886e+02 2.276e+02 2.833e+02 5.148e+02, threshold=4.553e+02, percent-clipped=2.0
2024-09-22 12:23:37,882 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.15 vs. limit=12.75
2024-09-22 12:24:02,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=14093.333333333334, ans=0.025
2024-09-22 12:24:14,176 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.87 vs. limit=5.114
2024-09-22 12:24:29,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=14140.0, ans=0.125
2024-09-22 12:24:47,936 INFO [train.py:1198] (1/4) Epoch 1, batch 3050, loss[loss=0.3639, ctc_loss=0.2798, cr_loss=0.4205, over 17091.00 frames. ], tot_loss[loss=0.4083, ctc_loss=0.322, cr_loss=0.4315, over 3339520.19 frames. ], batch size: 40, lr: 4.33e-02, grad_scale: 32.0
2024-09-22 12:24:48,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=14233.333333333334, ans=0.125
2024-09-22 12:25:10,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=14280.0, ans=0.125
2024-09-22 12:25:12,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=14280.0, ans=0.1572
2024-09-22 12:25:29,978 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.64 vs. limit=12.872499999999999
2024-09-22 12:26:06,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=14420.0, ans=0.125
2024-09-22 12:26:09,729 INFO [train.py:1198] (1/4) Epoch 1, batch 3100, loss[loss=0.404, ctc_loss=0.3151, cr_loss=0.4447, over 17297.00 frames. ], tot_loss[loss=0.4076, ctc_loss=0.3211, cr_loss=0.4324, over 3338231.84 frames. ], batch size: 46, lr: 4.33e-02, grad_scale: 32.0
2024-09-22 12:26:13,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=14466.666666666666, ans=0.3936666666666667
2024-09-22 12:26:15,852 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.870e+02 2.371e+02 3.057e+02 5.717e+02, threshold=4.743e+02, percent-clipped=5.0
2024-09-22 12:26:17,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=14466.666666666666, ans=0.007724637681159421
2024-09-22 12:26:30,239 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 12:26:31,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=14513.333333333334, ans=0.15486666666666665
2024-09-22 12:26:31,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=14513.333333333334, ans=0.125
2024-09-22 12:26:40,343 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=5.01 vs. limit=12.96
2024-09-22 12:26:47,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=14560.0, ans=0.125
2024-09-22 12:26:53,051 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.59 vs. limit=18.42
2024-09-22 12:26:56,311 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=5.30 vs. limit=9.842666666666666
2024-09-22 12:27:02,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=14606.666666666666, ans=0.125
2024-09-22 12:27:28,658 INFO [train.py:1198] (1/4) Epoch 1, batch 3150, loss[loss=0.3628, ctc_loss=0.2848, cr_loss=0.3898, over 17270.00 frames. ], tot_loss[loss=0.4051, ctc_loss=0.3187, cr_loss=0.4317, over 3338418.67 frames. ], batch size: 44, lr: 4.32e-02, grad_scale: 32.0
2024-09-22 12:27:45,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=14746.666666666666, ans=0.125
2024-09-22 12:28:14,762 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=18.70 vs. limit=18.630000000000003
2024-09-22 12:28:40,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=14886.666666666666, ans=0.15113333333333334
2024-09-22 12:28:47,524 INFO [train.py:1198] (1/4) Epoch 1, batch 3200, loss[loss=0.5293, ctc_loss=0.4316, cr_loss=0.4884, over 12477.00 frames. ], tot_loss[loss=0.4028, ctc_loss=0.3167, cr_loss=0.4308, over 3344652.65 frames. ], batch size: 123, lr: 4.32e-02, grad_scale: 32.0
2024-09-22 12:28:53,525 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 1.832e+02 2.280e+02 2.861e+02 6.877e+02, threshold=4.560e+02, percent-clipped=3.0
2024-09-22 12:29:06,901 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=14.48 vs. limit=13.1175
2024-09-22 12:29:11,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=14980.0, ans=0.125
2024-09-22 12:29:25,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=15026.666666666666, ans=0.0
2024-09-22 12:30:06,101 INFO [train.py:1198] (1/4) Epoch 1, batch 3250, loss[loss=0.4106, ctc_loss=0.3232, cr_loss=0.437, over 17349.00 frames. ], tot_loss[loss=0.3994, ctc_loss=0.3137, cr_loss=0.4285, over 3346612.95 frames. ], batch size: 48, lr: 4.31e-02, grad_scale: 32.0
2024-09-22 12:30:06,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer_na.min_abs, batch_count=15166.666666666666, ans=0.02
2024-09-22 12:30:09,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.62 vs. limit=10.066666666666666
2024-09-22 12:30:23,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=15213.333333333334, ans=0.0032777777777777753
2024-09-22 12:30:28,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=15213.333333333334, ans=0.125
2024-09-22 12:30:33,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=15213.333333333334, ans=0.00756231884057971
2024-09-22 12:30:35,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=15213.333333333334, ans=0.3675333333333334
2024-09-22 12:30:59,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=15306.666666666666, ans=0.125
2024-09-22 12:31:06,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=15306.666666666666, ans=0.0028888888888888853
2024-09-22 12:31:26,913 INFO [train.py:1198] (1/4) Epoch 1, batch 3300, loss[loss=0.3881, ctc_loss=0.3005, cr_loss=0.4381, over 16725.00 frames. ], tot_loss[loss=0.3979, ctc_loss=0.3123, cr_loss=0.4282, over 3346476.64 frames. ], batch size: 61, lr: 4.31e-02, grad_scale: 32.0
2024-09-22 12:31:32,209 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 12:31:33,368 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 1.835e+02 2.401e+02 3.313e+02 5.174e+02, threshold=4.802e+02, percent-clipped=5.0
2024-09-22 12:32:04,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=15493.333333333334, ans=0.007501449275362318
2024-09-22 12:32:10,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=15493.333333333334, ans=0.125
2024-09-22 12:32:16,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=15540.0, ans=0.125
2024-09-22 12:32:24,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=15540.0, ans=0.125
2024-09-22 12:32:26,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=15540.0, ans=0.125
2024-09-22 12:32:50,476 INFO [train.py:1198] (1/4) Epoch 1, batch 3350, loss[loss=0.3543, ctc_loss=0.2756, cr_loss=0.3936, over 17162.00 frames. ], tot_loss[loss=0.3962, ctc_loss=0.3107, cr_loss=0.4276, over 3349068.95 frames. ], batch size: 45, lr: 4.30e-02, grad_scale: 32.0
2024-09-22 12:33:11,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=15680.0, ans=0.125
2024-09-22 12:33:16,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=15680.0, ans=0.125
2024-09-22 12:33:20,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=15726.666666666666, ans=0.0011388888888888907
2024-09-22 12:33:55,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=15820.0, ans=0.0007500000000000007
2024-09-22 12:34:07,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=15866.666666666666, ans=0.04949747468305833
2024-09-22 12:34:09,480 INFO [train.py:1198] (1/4) Epoch 1, batch 3400, loss[loss=0.3579, ctc_loss=0.2782, cr_loss=0.3985, over 17040.00 frames. ], tot_loss[loss=0.3968, ctc_loss=0.311, cr_loss=0.429, over 3350694.69 frames. ], batch size: 39, lr: 4.29e-02, grad_scale: 32.0
2024-09-22 12:34:15,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=15866.666666666666, ans=13.45
2024-09-22 12:34:15,799 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 1.769e+02 2.096e+02 2.628e+02 4.837e+02, threshold=4.193e+02, percent-clipped=1.0
2024-09-22 12:34:24,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=15913.333333333334, ans=0.125
2024-09-22 12:34:57,865 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=7.56 vs. limit=9.001666666666665
2024-09-22 12:35:04,104 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=5.56 vs. limit=13.502500000000001
2024-09-22 12:35:06,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=16006.666666666666, ans=0.125
2024-09-22 12:35:23,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=16053.333333333334, ans=0.08946666666666664
2024-09-22 12:35:28,155 INFO [train.py:1198] (1/4) Epoch 1, batch 3450, loss[loss=0.3355, ctc_loss=0.2593, cr_loss=0.3811, over 16718.00 frames. ], tot_loss[loss=0.3945, ctc_loss=0.309, cr_loss=0.4275, over 3354052.52 frames. ], batch size: 37, lr: 4.29e-02, grad_scale: 32.0
2024-09-22 12:35:30,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=16100.0, ans=0.125
2024-09-22 12:35:45,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=16146.666666666666, ans=0.13853333333333334
2024-09-22 12:35:45,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=16146.666666666666, ans=0.125
2024-09-22 12:35:56,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=16146.666666666666, ans=0.125
2024-09-22 12:36:22,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=16240.0, ans=0.125
2024-09-22 12:36:29,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=16240.0, ans=0.125
2024-09-22 12:36:39,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=16286.666666666666, ans=0.0
2024-09-22 12:36:41,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=16286.666666666666, ans=0.13713333333333333
2024-09-22 12:36:48,696 INFO [train.py:1198] (1/4) Epoch 1, batch 3500, loss[loss=0.3297, ctc_loss=0.2483, cr_loss=0.4072, over 17052.00 frames. ], tot_loss[loss=0.3909, ctc_loss=0.3057, cr_loss=0.426, over 3358682.08 frames. ], batch size: 39, lr: 4.28e-02, grad_scale: 32.0
2024-09-22 12:36:54,800 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 1.781e+02 2.318e+02 3.100e+02 5.527e+02, threshold=4.636e+02, percent-clipped=9.0
2024-09-22 12:36:58,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=16333.333333333334, ans=0.0
2024-09-22 12:37:43,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=16473.333333333332, ans=0.125
2024-09-22 12:37:49,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=9.11 vs. limit=10.608
2024-09-22 12:37:51,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=16520.0, ans=0.0
2024-09-22 12:38:06,622 INFO [train.py:1198] (1/4) Epoch 1, batch 3550, loss[loss=0.4014, ctc_loss=0.3089, cr_loss=0.4627, over 17307.00 frames. ], tot_loss[loss=0.3932, ctc_loss=0.3076, cr_loss=0.4281, over 3358080.57 frames. ], batch size: 49, lr: 4.28e-02, grad_scale: 32.0
2024-09-22 12:38:06,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=16566.666666666668, ans=0.125
2024-09-22 12:38:17,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=16566.666666666668, ans=0.1343333333333333
2024-09-22 12:38:41,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=16660.0, ans=0.31690000000000007
2024-09-22 12:38:47,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=16660.0, ans=0.04949747468305833
2024-09-22 12:38:59,252 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=6.43 vs. limit=9.176666666666666
2024-09-22 12:39:06,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=16706.666666666668, ans=0.125
2024-09-22 12:39:09,554 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.55 vs. limit=10.701333333333332
2024-09-22 12:39:24,818 INFO [train.py:1198] (1/4) Epoch 1, batch 3600, loss[loss=0.396, ctc_loss=0.3103, cr_loss=0.4283, over 17013.00 frames. ], tot_loss[loss=0.3927, ctc_loss=0.3071, cr_loss=0.4281, over 3354635.94 frames. ], batch size: 56, lr: 4.27e-02, grad_scale: 32.0
2024-09-22 12:39:30,801 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 1.842e+02 2.057e+02 2.677e+02 5.057e+02, threshold=4.115e+02, percent-clipped=2.0
2024-09-22 12:39:57,699 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.82 vs. limit=9.223333333333333
2024-09-22 12:40:02,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=16893.333333333332, ans=0.3087333333333334
2024-09-22 12:40:11,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=16940.0, ans=0.0
2024-09-22 12:40:13,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=16940.0, ans=0.125
2024-09-22 12:40:22,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=16940.0, ans=0.0
2024-09-22 12:40:43,994 INFO [train.py:1198] (1/4) Epoch 1, batch 3650, loss[loss=0.3706, ctc_loss=0.2807, cr_loss=0.4492, over 17231.00 frames. ], tot_loss[loss=0.3918, ctc_loss=0.3061, cr_loss=0.4285, over 3353500.56 frames. ], batch size: 55, lr: 4.27e-02, grad_scale: 32.0
2024-09-22 12:41:12,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=17080.0, ans=0.125
2024-09-22 12:41:16,778 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.82 vs. limit=13.9225
2024-09-22 12:41:26,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=17126.666666666668, ans=0.0
2024-09-22 12:41:39,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=17173.333333333332, ans=0.0
2024-09-22 12:41:44,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=17173.333333333332, ans=0.1282666666666667
2024-09-22 12:41:49,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=21.05 vs. limit=20.415
2024-09-22 12:41:50,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=17220.0, ans=0.2973
2024-09-22 12:41:59,243 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=20.56 vs. limit=13.9575
2024-09-22 12:42:05,821 INFO [train.py:1198] (1/4) Epoch 1, batch 3700, loss[loss=0.3938, ctc_loss=0.3068, cr_loss=0.4347, over 16187.00 frames. ], tot_loss[loss=0.3901, ctc_loss=0.3045, cr_loss=0.4282, over 3355077.94 frames. ], batch size: 74, lr: 4.26e-02, grad_scale: 32.0
2024-09-22 12:42:12,057 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 1.892e+02 2.660e+02 3.633e+02 5.715e+02, threshold=5.320e+02, percent-clipped=15.0
2024-09-22 12:42:25,249 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=24.52 vs. limit=20.485
2024-09-22 12:43:06,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=17453.333333333332, ans=0.00707536231884058
2024-09-22 12:43:11,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.55 vs. limit=14.044999999999998
2024-09-22 12:43:18,890 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.44 vs. limit=14.044999999999998
2024-09-22 12:43:22,576 INFO [train.py:1198] (1/4) Epoch 1, batch 3750, loss[loss=0.4053, ctc_loss=0.3138, cr_loss=0.4573, over 17008.00 frames. ], tot_loss[loss=0.3903, ctc_loss=0.3047, cr_loss=0.428, over 3340911.70 frames. ], batch size: 53, lr: 4.26e-02, grad_scale: 32.0
2024-09-22 12:43:34,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=17500.0, ans=0.125
2024-09-22 12:43:42,667 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 12:43:47,940 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.45 vs. limit=11.018666666666668
2024-09-22 12:43:50,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=17546.666666666668, ans=0.125
2024-09-22 12:43:50,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=17546.666666666668, ans=0.007055072463768117
2024-09-22 12:43:54,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=17593.333333333332, ans=0.125
2024-09-22 12:44:21,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=17640.0, ans=0.2826000000000001
2024-09-22 12:44:40,336 INFO [train.py:1198] (1/4) Epoch 1, batch 3800, loss[loss=0.3887, ctc_loss=0.3041, cr_loss=0.423, over 17307.00 frames. ], tot_loss[loss=0.3915, ctc_loss=0.3057, cr_loss=0.4288, over 3318669.97 frames. ], batch size: 51, lr: 4.25e-02, grad_scale: 32.0
2024-09-22 12:44:46,416 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 1.780e+02 2.356e+02 3.200e+02 5.376e+02, threshold=4.713e+02, percent-clipped=1.0
2024-09-22 12:45:05,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=17780.0, ans=0.9278
2024-09-22 12:45:59,544 INFO [train.py:1198] (1/4) Epoch 1, batch 3850, loss[loss=0.4514, ctc_loss=0.3591, cr_loss=0.4616, over 15051.00 frames. ], tot_loss[loss=0.3939, ctc_loss=0.3082, cr_loss=0.4286, over 3281566.72 frames. ], batch size: 89, lr: 4.24e-02, grad_scale: 32.0
2024-09-22 12:46:06,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=17966.666666666668, ans=0.27116666666666667
2024-09-22 12:46:12,533 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=2.50 vs. limit=14.2375
2024-09-22 12:47:02,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=18153.333333333332, ans=0.0
2024-09-22 12:47:05,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=18153.333333333332, ans=0.125
2024-09-22 12:48:02,934 INFO [train.py:1198] (1/4) Epoch 2, batch 0, loss[loss=0.3712, ctc_loss=0.2871, cr_loss=0.4203, over 17075.00 frames. ], tot_loss[loss=0.3712, ctc_loss=0.2871, cr_loss=0.4203, over 17075.00 frames. ], batch size: 46, lr: 4.16e-02, grad_scale: 32.0
2024-09-22 12:48:02,934 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 12:48:18,079 INFO [train.py:1230] (1/4) Epoch 2, validation: loss=0.1169, ctc_loss=0.1169, cr_loss=1.034e-14, over 944034.00 frames. 
2024-09-22 12:48:18,079 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 12:48:30,952 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 1.944e+02 2.365e+02 3.007e+02 5.794e+02, threshold=4.731e+02, percent-clipped=1.0
2024-09-22 12:48:38,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=18228.0, ans=0.26202000000000014
2024-09-22 12:48:49,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=18228.0, ans=0.26202000000000014
2024-09-22 12:49:33,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=18368.0, ans=0.125
2024-09-22 12:49:36,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=18368.0, ans=0.11632
2024-09-22 12:49:39,680 INFO [train.py:1198] (1/4) Epoch 2, batch 50, loss[loss=0.3274, ctc_loss=0.2521, cr_loss=0.3764, over 17255.00 frames. ], tot_loss[loss=0.3845, ctc_loss=0.2994, cr_loss=0.4254, over 751446.96 frames. ], batch size: 42, lr: 4.15e-02, grad_scale: 32.0
2024-09-22 12:49:49,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=18414.666666666668, ans=0.09899494936611666
2024-09-22 12:50:05,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=18461.333333333332, ans=0.2538533333333334
2024-09-22 12:50:21,904 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=22.34 vs. limit=21.381
2024-09-22 12:50:43,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=18601.333333333332, ans=0.0
2024-09-22 12:50:43,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=18601.333333333332, ans=0.125
2024-09-22 12:50:59,411 INFO [train.py:1198] (1/4) Epoch 2, batch 100, loss[loss=0.3259, ctc_loss=0.2496, cr_loss=0.3817, over 16946.00 frames. ], tot_loss[loss=0.3821, ctc_loss=0.297, cr_loss=0.4252, over 1325188.78 frames. ], batch size: 42, lr: 4.15e-02, grad_scale: 32.0
2024-09-22 12:51:06,748 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.62 vs. limit=5.7972
2024-09-22 12:51:19,088 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 1.796e+02 2.128e+02 2.839e+02 5.119e+02, threshold=4.256e+02, percent-clipped=1.0
2024-09-22 12:51:25,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=18694.666666666668, ans=0.07
2024-09-22 12:51:51,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=18741.333333333332, ans=0.125
2024-09-22 12:52:25,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=18881.333333333332, ans=0.125
2024-09-22 12:52:26,281 INFO [train.py:1198] (1/4) Epoch 2, batch 150, loss[loss=0.3586, ctc_loss=0.2748, cr_loss=0.419, over 17209.00 frames. ], tot_loss[loss=0.3824, ctc_loss=0.2968, cr_loss=0.4279, over 1776246.17 frames. ], batch size: 47, lr: 4.14e-02, grad_scale: 32.0
2024-09-22 12:52:29,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=18881.333333333332, ans=0.025
2024-09-22 12:52:37,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=18881.333333333332, ans=0.125
2024-09-22 12:53:19,669 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=13.56 vs. limit=21.766
2024-09-22 12:53:36,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=19068.0, ans=0.10932
2024-09-22 12:53:51,161 INFO [train.py:1198] (1/4) Epoch 2, batch 200, loss[loss=0.3641, ctc_loss=0.282, cr_loss=0.4107, over 17169.00 frames. ], tot_loss[loss=0.381, ctc_loss=0.2955, cr_loss=0.4273, over 2130555.68 frames. ], batch size: 45, lr: 4.13e-02, grad_scale: 32.0
2024-09-22 12:54:03,746 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 1.786e+02 2.214e+02 2.969e+02 6.338e+02, threshold=4.427e+02, percent-clipped=7.0
2024-09-22 12:54:07,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=19161.333333333332, ans=0.125
2024-09-22 12:54:13,913 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.16 vs. limit=14.685500000000001
2024-09-22 12:54:37,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=19254.666666666668, ans=0.10745333333333332
2024-09-22 12:55:02,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=19301.333333333332, ans=0.2244533333333334
2024-09-22 12:55:07,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=19301.333333333332, ans=0.0
2024-09-22 12:55:10,171 INFO [train.py:1198] (1/4) Epoch 2, batch 250, loss[loss=0.396, ctc_loss=0.3091, cr_loss=0.435, over 17135.00 frames. ], tot_loss[loss=0.3794, ctc_loss=0.2941, cr_loss=0.4263, over 2399138.39 frames. ], batch size: 48, lr: 4.13e-02, grad_scale: 32.0
2024-09-22 12:55:15,763 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=7.55 vs. limit=9.837
2024-09-22 12:55:18,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=19348.0, ans=0.10652
2024-09-22 12:55:24,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=19394.666666666668, ans=0.10605333333333333
2024-09-22 12:55:26,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=19394.666666666668, ans=0.025
2024-09-22 12:55:56,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=19488.0, ans=0.0
2024-09-22 12:56:11,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=19488.0, ans=0.2179200000000001
2024-09-22 12:56:23,065 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.63 vs. limit=14.8255
2024-09-22 12:56:35,316 INFO [train.py:1198] (1/4) Epoch 2, batch 300, loss[loss=0.3609, ctc_loss=0.2755, cr_loss=0.4273, over 17281.00 frames. ], tot_loss[loss=0.3778, ctc_loss=0.2927, cr_loss=0.4255, over 2615212.86 frames. ], batch size: 46, lr: 4.12e-02, grad_scale: 32.0
2024-09-22 12:56:35,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=19581.333333333332, ans=0.018872666666666676
2024-09-22 12:56:35,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=19581.333333333332, ans=0.125
2024-09-22 12:56:48,373 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.759e+02 2.124e+02 2.853e+02 4.892e+02, threshold=4.248e+02, percent-clipped=3.0
2024-09-22 12:56:48,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=19581.333333333332, ans=0.21465333333333336
2024-09-22 12:56:58,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=19628.0, ans=0.2130200000000001
2024-09-22 12:57:16,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.13 vs. limit=5.9512
2024-09-22 12:57:46,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=19768.0, ans=0.125
2024-09-22 12:57:47,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=19768.0, ans=0.025
2024-09-22 12:57:54,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=19768.0, ans=0.10232000000000002
2024-09-22 12:57:57,758 INFO [train.py:1198] (1/4) Epoch 2, batch 350, loss[loss=0.3536, ctc_loss=0.2698, cr_loss=0.4191, over 17275.00 frames. ], tot_loss[loss=0.3766, ctc_loss=0.2916, cr_loss=0.4252, over 2775955.43 frames. ], batch size: 42, lr: 4.12e-02, grad_scale: 32.0
2024-09-22 12:58:12,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=19861.333333333332, ans=0.006551884057971015
2024-09-22 12:58:22,520 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.85 vs. limit=9.965333333333334
2024-09-22 12:58:24,432 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.36 vs. limit=7.972266666666666
2024-09-22 12:58:45,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=19908.0, ans=0.125
2024-09-22 12:59:01,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=19954.666666666668, ans=0.0
2024-09-22 12:59:12,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=20001.333333333332, ans=0.0
2024-09-22 12:59:20,247 INFO [train.py:1198] (1/4) Epoch 2, batch 400, loss[loss=0.3465, ctc_loss=0.2652, cr_loss=0.4064, over 17219.00 frames. ], tot_loss[loss=0.3741, ctc_loss=0.2894, cr_loss=0.4235, over 2907796.97 frames. ], batch size: 47, lr: 4.11e-02, grad_scale: 32.0
2024-09-22 12:59:32,897 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.825e+02 2.194e+02 2.995e+02 5.365e+02, threshold=4.388e+02, percent-clipped=4.0
2024-09-22 12:59:33,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=20048.0, ans=0.125
2024-09-22 12:59:42,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=20094.666666666668, ans=0.006501159420289854
2024-09-22 12:59:47,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=7.17 vs. limit=15.0
2024-09-22 13:00:00,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=20141.333333333332, ans=0.125
2024-09-22 13:00:13,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=20188.0, ans=0.1
2024-09-22 13:00:39,701 INFO [train.py:1198] (1/4) Epoch 2, batch 450, loss[loss=0.3657, ctc_loss=0.2808, cr_loss=0.4245, over 17363.00 frames. ], tot_loss[loss=0.3739, ctc_loss=0.2892, cr_loss=0.4234, over 2997641.60 frames. ], batch size: 48, lr: 4.10e-02, grad_scale: 32.0
2024-09-22 13:00:43,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=20281.333333333332, ans=0.125
2024-09-22 13:00:43,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=20281.333333333332, ans=0.1
2024-09-22 13:00:57,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=20328.0, ans=0.006450434782608696
2024-09-22 13:01:32,545 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.45 vs. limit=12.0
2024-09-22 13:01:35,765 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.65 vs. limit=12.0
2024-09-22 13:01:54,838 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=10.34 vs. limit=15.0
2024-09-22 13:02:00,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=20468.0, ans=0.125
2024-09-22 13:02:04,847 INFO [train.py:1198] (1/4) Epoch 2, batch 500, loss[loss=0.3269, ctc_loss=0.2484, cr_loss=0.3927, over 16684.00 frames. ], tot_loss[loss=0.3742, ctc_loss=0.2894, cr_loss=0.424, over 3074834.50 frames. ], batch size: 37, lr: 4.10e-02, grad_scale: 32.0
2024-09-22 13:02:13,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=20514.666666666668, ans=0.025
2024-09-22 13:02:13,748 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.48 vs. limit=12.0
2024-09-22 13:02:17,796 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 1.804e+02 2.205e+02 2.881e+02 5.655e+02, threshold=4.410e+02, percent-clipped=3.0
2024-09-22 13:02:19,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=20561.333333333332, ans=0.125
2024-09-22 13:02:23,262 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.65 vs. limit=10.0
2024-09-22 13:02:33,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=20561.333333333332, ans=0.09899494936611666
2024-09-22 13:02:52,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=20608.0, ans=0.1
2024-09-22 13:03:01,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.max_abs, batch_count=20654.666666666668, ans=10.0
2024-09-22 13:03:29,949 INFO [train.py:1198] (1/4) Epoch 2, batch 550, loss[loss=0.3732, ctc_loss=0.2895, cr_loss=0.4188, over 17012.00 frames. ], tot_loss[loss=0.3729, ctc_loss=0.2882, cr_loss=0.4235, over 3141945.57 frames. ], batch size: 53, lr: 4.09e-02, grad_scale: 32.0
2024-09-22 13:03:33,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=20748.0, ans=0.0063591304347826085
2024-09-22 13:04:22,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.87 vs. limit=22.5
2024-09-22 13:04:27,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=20888.0, ans=0.125
2024-09-22 13:04:27,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=20888.0, ans=0.0063286956521739135
2024-09-22 13:04:31,531 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.52 vs. limit=15.0
2024-09-22 13:04:37,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=20934.666666666668, ans=0.125
2024-09-22 13:04:40,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=20934.666666666668, ans=0.0
2024-09-22 13:04:49,741 INFO [train.py:1198] (1/4) Epoch 2, batch 600, loss[loss=0.3688, ctc_loss=0.2866, cr_loss=0.4111, over 17213.00 frames. ], tot_loss[loss=0.3728, ctc_loss=0.288, cr_loss=0.4238, over 3189576.06 frames. ], batch size: 47, lr: 4.09e-02, grad_scale: 32.0
2024-09-22 13:04:54,018 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.98 vs. limit=15.0
2024-09-22 13:04:56,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=20981.333333333332, ans=0.025
2024-09-22 13:05:01,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=20981.333333333332, ans=0.1
2024-09-22 13:05:02,738 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 1.714e+02 2.171e+02 2.480e+02 4.403e+02, threshold=4.342e+02, percent-clipped=0.0
2024-09-22 13:05:33,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=21074.666666666668, ans=0.1
2024-09-22 13:05:33,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.70 vs. limit=15.0
2024-09-22 13:06:15,351 INFO [train.py:1198] (1/4) Epoch 2, batch 650, loss[loss=0.3806, ctc_loss=0.2905, cr_loss=0.4502, over 16850.00 frames. ], tot_loss[loss=0.373, ctc_loss=0.2881, cr_loss=0.4245, over 3226941.29 frames. ], batch size: 58, lr: 4.08e-02, grad_scale: 32.0
2024-09-22 13:06:20,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=21214.666666666668, ans=0.0
2024-09-22 13:06:30,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=21261.333333333332, ans=0.125
2024-09-22 13:06:52,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=21308.0, ans=0.0
2024-09-22 13:07:03,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=21354.666666666668, ans=0.025
2024-09-22 13:07:06,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.min_positive, batch_count=21354.666666666668, ans=0.025
2024-09-22 13:07:14,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=21354.666666666668, ans=0.025
2024-09-22 13:07:21,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=21401.333333333332, ans=0.125
2024-09-22 13:07:25,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=21401.333333333332, ans=0.125
2024-09-22 13:07:37,761 INFO [train.py:1198] (1/4) Epoch 2, batch 700, loss[loss=0.3453, ctc_loss=0.2676, cr_loss=0.3881, over 17006.00 frames. ], tot_loss[loss=0.3694, ctc_loss=0.2849, cr_loss=0.4224, over 3263101.83 frames. ], batch size: 56, lr: 4.07e-02, grad_scale: 32.0
2024-09-22 13:07:50,782 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 1.762e+02 2.307e+02 2.806e+02 6.388e+02, threshold=4.614e+02, percent-clipped=12.0
2024-09-22 13:08:06,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=21494.666666666668, ans=0.2
2024-09-22 13:08:10,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=21541.333333333332, ans=0.125
2024-09-22 13:08:36,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=21588.0, ans=0.025
2024-09-22 13:08:42,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=21634.666666666668, ans=0.0
2024-09-22 13:08:59,998 INFO [train.py:1198] (1/4) Epoch 2, batch 750, loss[loss=0.3682, ctc_loss=0.284, cr_loss=0.4208, over 17084.00 frames. ], tot_loss[loss=0.369, ctc_loss=0.2844, cr_loss=0.4229, over 3290212.91 frames. ], batch size: 49, lr: 4.07e-02, grad_scale: 32.0
2024-09-22 13:09:43,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=21774.666666666668, ans=0.1
2024-09-22 13:10:05,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=21868.0, ans=0.125
2024-09-22 13:10:17,163 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.52 vs. limit=6.0
2024-09-22 13:10:17,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=21914.666666666668, ans=0.006105507246376812
2024-09-22 13:10:19,246 INFO [train.py:1198] (1/4) Epoch 2, batch 800, loss[loss=0.3539, ctc_loss=0.2725, cr_loss=0.4073, over 17048.00 frames. ], tot_loss[loss=0.3699, ctc_loss=0.2853, cr_loss=0.4233, over 3295741.11 frames. ], batch size: 46, lr: 4.06e-02, grad_scale: 32.0
2024-09-22 13:10:19,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=21914.666666666668, ans=0.125
2024-09-22 13:10:32,122 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.379e+02 1.875e+02 2.284e+02 2.792e+02 4.521e+02, threshold=4.569e+02, percent-clipped=0.0
2024-09-22 13:10:43,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=21961.333333333332, ans=0.0
2024-09-22 13:10:53,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=22008.0, ans=15.0
2024-09-22 13:11:38,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=22101.333333333332, ans=0.125
2024-09-22 13:11:39,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=22101.333333333332, ans=0.025
2024-09-22 13:11:40,105 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.10 vs. limit=15.0
2024-09-22 13:11:43,999 INFO [train.py:1198] (1/4) Epoch 2, batch 850, loss[loss=0.3454, ctc_loss=0.2651, cr_loss=0.4012, over 17111.00 frames. ], tot_loss[loss=0.3696, ctc_loss=0.2849, cr_loss=0.4236, over 3319931.84 frames. ], batch size: 40, lr: 4.06e-02, grad_scale: 32.0
2024-09-22 13:11:46,933 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.53 vs. limit=6.0
2024-09-22 13:11:53,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=22148.0, ans=0.006054782608695653
2024-09-22 13:12:09,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=22194.666666666668, ans=0.125
2024-09-22 13:12:14,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=22241.333333333332, ans=0.125
2024-09-22 13:12:14,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=22241.333333333332, ans=0.125
2024-09-22 13:12:23,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=22241.333333333332, ans=0.0
2024-09-22 13:13:02,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=22334.666666666668, ans=0.125
2024-09-22 13:13:05,748 INFO [train.py:1198] (1/4) Epoch 2, batch 900, loss[loss=0.3902, ctc_loss=0.3012, cr_loss=0.4446, over 17371.00 frames. ], tot_loss[loss=0.3696, ctc_loss=0.2848, cr_loss=0.4241, over 3336230.23 frames. ], batch size: 48, lr: 4.05e-02, grad_scale: 32.0
2024-09-22 13:13:15,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=22381.333333333332, ans=0.95
2024-09-22 13:13:21,505 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 1.677e+02 2.003e+02 2.906e+02 6.404e+02, threshold=4.006e+02, percent-clipped=3.0
2024-09-22 13:13:21,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=22381.333333333332, ans=0.1
2024-09-22 13:13:37,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=22428.0, ans=0.0
2024-09-22 13:14:03,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=22521.333333333332, ans=0.1
2024-09-22 13:14:28,989 INFO [train.py:1198] (1/4) Epoch 2, batch 950, loss[loss=0.3468, ctc_loss=0.2674, cr_loss=0.3971, over 17304.00 frames. ], tot_loss[loss=0.3691, ctc_loss=0.2842, cr_loss=0.4242, over 3349574.78 frames. ], batch size: 46, lr: 4.04e-02, grad_scale: 64.0
2024-09-22 13:14:31,652 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.11 vs. limit=22.5
2024-09-22 13:14:49,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=22661.333333333332, ans=0.0
2024-09-22 13:14:57,036 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.68 vs. limit=12.0
2024-09-22 13:15:16,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=22754.666666666668, ans=0.125
2024-09-22 13:15:45,891 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.58 vs. limit=15.0
2024-09-22 13:15:48,456 INFO [train.py:1198] (1/4) Epoch 2, batch 1000, loss[loss=0.39, ctc_loss=0.3005, cr_loss=0.4476, over 16917.00 frames. ], tot_loss[loss=0.369, ctc_loss=0.2841, cr_loss=0.4242, over 3350347.36 frames. ], batch size: 58, lr: 4.04e-02, grad_scale: 64.0
2024-09-22 13:16:07,762 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 1.793e+02 2.170e+02 2.654e+02 4.100e+02, threshold=4.339e+02, percent-clipped=2.0
2024-09-22 13:16:38,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=22941.333333333332, ans=0.125
2024-09-22 13:16:49,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=22988.0, ans=0.2
2024-09-22 13:16:56,236 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.57 vs. limit=15.0
2024-09-22 13:17:15,715 INFO [train.py:1198] (1/4) Epoch 2, batch 1050, loss[loss=0.3991, ctc_loss=0.3114, cr_loss=0.4386, over 16561.00 frames. ], tot_loss[loss=0.3667, ctc_loss=0.2824, cr_loss=0.4218, over 3352392.77 frames. ], batch size: 66, lr: 4.03e-02, grad_scale: 32.0
2024-09-22 13:17:17,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=23081.333333333332, ans=0.005851884057971015
2024-09-22 13:17:28,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=23081.333333333332, ans=0.2
2024-09-22 13:17:46,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=23174.666666666668, ans=0.125
2024-09-22 13:17:48,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=23174.666666666668, ans=0.005831594202898551
2024-09-22 13:17:49,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=23174.666666666668, ans=0.2
2024-09-22 13:18:07,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=23221.333333333332, ans=0.0
2024-09-22 13:18:18,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=23221.333333333332, ans=0.005821449275362319
2024-09-22 13:18:37,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=23314.666666666668, ans=0.125
2024-09-22 13:18:37,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=23314.666666666668, ans=0.5
2024-09-22 13:18:38,530 INFO [train.py:1198] (1/4) Epoch 2, batch 1100, loss[loss=0.3361, ctc_loss=0.2499, cr_loss=0.4312, over 17248.00 frames. ], tot_loss[loss=0.3665, ctc_loss=0.282, cr_loss=0.4223, over 3355746.01 frames. ], batch size: 42, lr: 4.03e-02, grad_scale: 32.0
2024-09-22 13:18:53,229 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.707e+02 2.038e+02 2.702e+02 5.370e+02, threshold=4.076e+02, percent-clipped=2.0
2024-09-22 13:19:04,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=23361.333333333332, ans=0.0
2024-09-22 13:19:09,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=23408.0, ans=0.1
2024-09-22 13:19:58,506 INFO [train.py:1198] (1/4) Epoch 2, batch 1150, loss[loss=0.3713, ctc_loss=0.2842, cr_loss=0.4356, over 17007.00 frames. ], tot_loss[loss=0.3653, ctc_loss=0.2811, cr_loss=0.4208, over 3358332.33 frames. ], batch size: 51, lr: 4.02e-02, grad_scale: 32.0
2024-09-22 13:20:42,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=23641.333333333332, ans=0.005730144927536232
2024-09-22 13:20:54,849 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=4.46 vs. limit=15.0
2024-09-22 13:21:23,360 INFO [train.py:1198] (1/4) Epoch 2, batch 1200, loss[loss=0.365, ctc_loss=0.2782, cr_loss=0.434, over 17011.00 frames. ], tot_loss[loss=0.366, ctc_loss=0.2818, cr_loss=0.4213, over 3341655.77 frames. ], batch size: 51, lr: 4.01e-02, grad_scale: 32.0
2024-09-22 13:21:37,785 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.890e+02 2.251e+02 2.751e+02 4.727e+02, threshold=4.502e+02, percent-clipped=4.0
2024-09-22 13:21:43,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=23828.0, ans=0.2
2024-09-22 13:21:45,236 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.89 vs. limit=15.0
2024-09-22 13:22:16,638 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=14.15 vs. limit=15.0
2024-09-22 13:22:19,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=23921.333333333332, ans=0.00566927536231884
2024-09-22 13:22:25,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=23921.333333333332, ans=0.125
2024-09-22 13:22:41,747 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=2.682e-03
2024-09-22 13:22:44,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=24014.666666666668, ans=0.125
2024-09-22 13:22:45,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=24014.666666666668, ans=0.1
2024-09-22 13:22:46,211 INFO [train.py:1198] (1/4) Epoch 2, batch 1250, loss[loss=0.4492, ctc_loss=0.3533, cr_loss=0.4791, over 15188.00 frames. ], tot_loss[loss=0.3659, ctc_loss=0.2817, cr_loss=0.4211, over 3344674.04 frames. ], batch size: 89, lr: 4.01e-02, grad_scale: 32.0
2024-09-22 13:22:51,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=24014.666666666668, ans=0.0
2024-09-22 13:22:51,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=24014.666666666668, ans=0.125
2024-09-22 13:22:54,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=24014.666666666668, ans=0.1
2024-09-22 13:22:58,543 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.98 vs. limit=15.0
2024-09-22 13:23:25,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=24108.0, ans=0.125
2024-09-22 13:23:27,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=24108.0, ans=0.0
2024-09-22 13:23:45,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=24154.666666666668, ans=0.005618550724637681
2024-09-22 13:23:48,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=24154.666666666668, ans=0.005618550724637681
2024-09-22 13:23:56,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer_ff3.min_abs, batch_count=24201.333333333332, ans=0.2
2024-09-22 13:24:02,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=24201.333333333332, ans=0.0
2024-09-22 13:24:08,732 INFO [train.py:1198] (1/4) Epoch 2, batch 1300, loss[loss=0.2919, ctc_loss=0.2173, cr_loss=0.3731, over 16968.00 frames. ], tot_loss[loss=0.3664, ctc_loss=0.282, cr_loss=0.4221, over 3343544.86 frames. ], batch size: 42, lr: 4.00e-02, grad_scale: 32.0
2024-09-22 13:24:10,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=24248.0, ans=0.025
2024-09-22 13:24:16,168 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.68 vs. limit=10.0
2024-09-22 13:24:23,423 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 1.885e+02 2.250e+02 3.074e+02 5.750e+02, threshold=4.500e+02, percent-clipped=7.0
2024-09-22 13:24:23,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=24294.666666666668, ans=0.1
2024-09-22 13:24:25,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=24294.666666666668, ans=0.04949747468305833
2024-09-22 13:24:28,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer_na.min_abs, batch_count=24294.666666666668, ans=0.02
2024-09-22 13:24:35,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=24294.666666666668, ans=0.0
2024-09-22 13:24:37,041 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.22 vs. limit=15.0
2024-09-22 13:24:52,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=24341.333333333332, ans=0.125
2024-09-22 13:25:14,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=24434.666666666668, ans=0.1
2024-09-22 13:25:16,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=24434.666666666668, ans=0.0
2024-09-22 13:25:28,592 INFO [train.py:1198] (1/4) Epoch 2, batch 1350, loss[loss=0.35, ctc_loss=0.2634, cr_loss=0.4334, over 17342.00 frames. ], tot_loss[loss=0.3663, ctc_loss=0.2818, cr_loss=0.4222, over 3340168.17 frames. ], batch size: 48, lr: 3.99e-02, grad_scale: 32.0
2024-09-22 13:25:38,663 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 13:25:56,721 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.62 vs. limit=15.0
2024-09-22 13:25:58,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=24528.0, ans=0.125
2024-09-22 13:26:10,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=24574.666666666668, ans=0.1
2024-09-22 13:26:25,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=24621.333333333332, ans=0.0055171014492753625
2024-09-22 13:26:25,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=24621.333333333332, ans=0.125
2024-09-22 13:26:41,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=24668.0, ans=0.0
2024-09-22 13:26:54,050 INFO [train.py:1198] (1/4) Epoch 2, batch 1400, loss[loss=0.4129, ctc_loss=0.318, cr_loss=0.4743, over 17098.00 frames. ], tot_loss[loss=0.3678, ctc_loss=0.2829, cr_loss=0.4244, over 3341217.60 frames. ], batch size: 49, lr: 3.99e-02, grad_scale: 32.0
2024-09-22 13:26:59,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=24714.666666666668, ans=0.1
2024-09-22 13:26:59,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=24714.666666666668, ans=0.09899494936611666
2024-09-22 13:27:11,157 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.006e+02 2.479e+02 3.166e+02 4.715e+02, threshold=4.958e+02, percent-clipped=3.0
2024-09-22 13:27:30,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=24808.0, ans=0.005476521739130435
2024-09-22 13:27:33,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=24808.0, ans=0.1
2024-09-22 13:28:01,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=24901.333333333332, ans=0.1
2024-09-22 13:28:16,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.89 vs. limit=15.0
2024-09-22 13:28:19,234 INFO [train.py:1198] (1/4) Epoch 2, batch 1450, loss[loss=0.4088, ctc_loss=0.3222, cr_loss=0.4331, over 15079.00 frames. ], tot_loss[loss=0.3669, ctc_loss=0.282, cr_loss=0.4248, over 3342502.46 frames. ], batch size: 89, lr: 3.98e-02, grad_scale: 32.0
2024-09-22 13:28:23,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.21 vs. limit=22.5
2024-09-22 13:28:25,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=24948.0, ans=0.0
2024-09-22 13:28:57,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=25041.333333333332, ans=0.125
2024-09-22 13:29:08,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=25088.0, ans=0.0
2024-09-22 13:29:11,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=25088.0, ans=0.005415652173913044
2024-09-22 13:29:21,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=25134.666666666668, ans=0.1
2024-09-22 13:29:23,267 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 13:29:32,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=25134.666666666668, ans=0.125
2024-09-22 13:29:32,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=25134.666666666668, ans=0.2
2024-09-22 13:29:38,894 INFO [train.py:1198] (1/4) Epoch 2, batch 1500, loss[loss=0.3553, ctc_loss=0.2718, cr_loss=0.4176, over 17079.00 frames. ], tot_loss[loss=0.3649, ctc_loss=0.2801, cr_loss=0.4237, over 3353895.31 frames. ], batch size: 46, lr: 3.98e-02, grad_scale: 32.0
2024-09-22 13:29:44,384 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.63 vs. limit=15.0
2024-09-22 13:29:47,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=25181.333333333332, ans=0.0
2024-09-22 13:29:52,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=25181.333333333332, ans=0.0
2024-09-22 13:29:53,535 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 1.701e+02 2.138e+02 2.969e+02 5.127e+02, threshold=4.276e+02, percent-clipped=1.0
2024-09-22 13:29:55,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=25228.0, ans=0.025
2024-09-22 13:30:09,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=25274.666666666668, ans=0.1
2024-09-22 13:30:39,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer_na.min_abs, batch_count=25321.333333333332, ans=0.02
2024-09-22 13:31:00,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=25368.0, ans=0.125
2024-09-22 13:31:03,514 INFO [train.py:1198] (1/4) Epoch 2, batch 1550, loss[loss=0.3963, ctc_loss=0.3107, cr_loss=0.4276, over 17050.00 frames. ], tot_loss[loss=0.3646, ctc_loss=0.2799, cr_loss=0.4237, over 3353312.40 frames. ], batch size: 56, lr: 3.97e-02, grad_scale: 32.0
2024-09-22 13:31:03,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=25414.666666666668, ans=0.005344637681159421
2024-09-22 13:31:08,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=25414.666666666668, ans=0.05
2024-09-22 13:31:14,272 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.35 vs. limit=6.0
2024-09-22 13:31:20,197 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.74 vs. limit=6.0
2024-09-22 13:31:34,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=25508.0, ans=0.1
2024-09-22 13:31:35,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=25508.0, ans=0.025
2024-09-22 13:31:39,266 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.25 vs. limit=15.0
2024-09-22 13:32:06,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=25554.666666666668, ans=0.005314202898550725
2024-09-22 13:32:25,136 INFO [train.py:1198] (1/4) Epoch 2, batch 1600, loss[loss=0.3287, ctc_loss=0.2515, cr_loss=0.386, over 16948.00 frames. ], tot_loss[loss=0.364, ctc_loss=0.2793, cr_loss=0.4235, over 3354617.52 frames. ], batch size: 42, lr: 3.96e-02, grad_scale: 32.0
2024-09-22 13:32:39,455 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 1.757e+02 1.985e+02 2.306e+02 3.282e+02, threshold=3.970e+02, percent-clipped=0.0
2024-09-22 13:32:54,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=25694.666666666668, ans=0.0
2024-09-22 13:32:56,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=25741.333333333332, ans=0.0
2024-09-22 13:33:40,269 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=17.54 vs. limit=22.5
2024-09-22 13:33:47,217 INFO [train.py:1198] (1/4) Epoch 2, batch 1650, loss[loss=0.3082, ctc_loss=0.2315, cr_loss=0.3834, over 17037.00 frames. ], tot_loss[loss=0.3636, ctc_loss=0.2791, cr_loss=0.4228, over 3351018.29 frames. ], batch size: 39, lr: 3.96e-02, grad_scale: 32.0
2024-09-22 13:34:24,789 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.66 vs. limit=15.0
2024-09-22 13:34:42,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=26021.333333333332, ans=0.0
2024-09-22 13:34:58,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=26068.0, ans=0.04949747468305833
2024-09-22 13:35:07,607 INFO [train.py:1198] (1/4) Epoch 2, batch 1700, loss[loss=0.4163, ctc_loss=0.3221, cr_loss=0.4707, over 16794.00 frames. ], tot_loss[loss=0.364, ctc_loss=0.2793, cr_loss=0.4233, over 3345740.14 frames. ], batch size: 61, lr: 3.95e-02, grad_scale: 32.0
2024-09-22 13:35:22,185 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 1.761e+02 2.144e+02 2.600e+02 4.141e+02, threshold=4.288e+02, percent-clipped=2.0
2024-09-22 13:35:27,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=26161.333333333332, ans=0.0
2024-09-22 13:35:53,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=26208.0, ans=0.05
2024-09-22 13:35:58,663 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.03 vs. limit=10.0
2024-09-22 13:36:33,102 INFO [train.py:1198] (1/4) Epoch 2, batch 1750, loss[loss=0.3123, ctc_loss=0.2361, cr_loss=0.3811, over 17194.00 frames. ], tot_loss[loss=0.3632, ctc_loss=0.2785, cr_loss=0.4231, over 3357243.65 frames. ], batch size: 41, lr: 3.94e-02, grad_scale: 32.0
2024-09-22 13:37:03,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten.whitening_limit, batch_count=26394.666666666668, ans=15.0
2024-09-22 13:37:08,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=14.24 vs. limit=15.0
2024-09-22 13:37:15,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=26441.333333333332, ans=0.2
2024-09-22 13:37:17,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=26441.333333333332, ans=0.125
2024-09-22 13:37:27,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.25 vs. limit=15.0
2024-09-22 13:37:42,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=26534.666666666668, ans=0.0
2024-09-22 13:37:57,654 INFO [train.py:1198] (1/4) Epoch 2, batch 1800, loss[loss=0.3422, ctc_loss=0.2657, cr_loss=0.3828, over 17225.00 frames. ], tot_loss[loss=0.3628, ctc_loss=0.2784, cr_loss=0.422, over 3345632.49 frames. ], batch size: 47, lr: 3.94e-02, grad_scale: 32.0
2024-09-22 13:38:11,941 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 1.770e+02 2.078e+02 2.651e+02 4.856e+02, threshold=4.156e+02, percent-clipped=2.0
2024-09-22 13:38:17,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=26628.0, ans=0.0
2024-09-22 13:38:25,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=26628.0, ans=0.0
2024-09-22 13:38:26,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=26628.0, ans=0.125
2024-09-22 13:38:34,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=26674.666666666668, ans=0.025
2024-09-22 13:38:53,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=26721.333333333332, ans=0.125
2024-09-22 13:39:08,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=26768.0, ans=0.2
2024-09-22 13:39:17,324 INFO [train.py:1198] (1/4) Epoch 2, batch 1850, loss[loss=0.4098, ctc_loss=0.3211, cr_loss=0.4439, over 15970.00 frames. ], tot_loss[loss=0.3638, ctc_loss=0.2792, cr_loss=0.423, over 3352189.57 frames. ], batch size: 74, lr: 3.93e-02, grad_scale: 32.0
2024-09-22 13:39:20,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=26814.666666666668, ans=0.125
2024-09-22 13:39:29,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=26814.666666666668, ans=0.125
2024-09-22 13:39:32,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=26861.333333333332, ans=0.125
2024-09-22 13:39:40,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=26861.333333333332, ans=0.0
2024-09-22 13:39:43,728 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.56 vs. limit=22.5
2024-09-22 13:39:50,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=26908.0, ans=0.125
2024-09-22 13:40:07,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=26954.666666666668, ans=0.005009855072463768
2024-09-22 13:40:10,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=26954.666666666668, ans=0.125
2024-09-22 13:40:32,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=27001.333333333332, ans=0.125
2024-09-22 13:40:39,945 INFO [train.py:1198] (1/4) Epoch 2, batch 1900, loss[loss=0.3808, ctc_loss=0.2915, cr_loss=0.4462, over 17082.00 frames. ], tot_loss[loss=0.3625, ctc_loss=0.2779, cr_loss=0.4231, over 3356868.14 frames. ], batch size: 46, lr: 3.92e-02, grad_scale: 32.0
2024-09-22 13:40:57,181 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.864e+02 2.204e+02 2.855e+02 4.990e+02, threshold=4.407e+02, percent-clipped=2.0
2024-09-22 13:41:15,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=27141.333333333332, ans=0.125
2024-09-22 13:41:31,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=27188.0, ans=0.125
2024-09-22 13:41:38,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=27188.0, ans=0.125
2024-09-22 13:41:52,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=27234.666666666668, ans=0.125
2024-09-22 13:41:59,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=27234.666666666668, ans=0.2
2024-09-22 13:42:05,868 INFO [train.py:1198] (1/4) Epoch 2, batch 1950, loss[loss=0.3662, ctc_loss=0.2775, cr_loss=0.4438, over 17302.00 frames. ], tot_loss[loss=0.3621, ctc_loss=0.2773, cr_loss=0.424, over 3361986.92 frames. ], batch size: 49, lr: 3.92e-02, grad_scale: 32.0
2024-09-22 13:42:48,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=27374.666666666668, ans=0.004918550724637681
2024-09-22 13:43:02,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=27421.333333333332, ans=0.125
2024-09-22 13:43:06,472 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=10.88 vs. limit=15.0
2024-09-22 13:43:16,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=27468.0, ans=0.125
2024-09-22 13:43:27,962 INFO [train.py:1198] (1/4) Epoch 2, batch 2000, loss[loss=0.4122, ctc_loss=0.3092, cr_loss=0.515, over 16072.00 frames. ], tot_loss[loss=0.3615, ctc_loss=0.2767, cr_loss=0.4239, over 3365814.26 frames. ], batch size: 74, lr: 3.91e-02, grad_scale: 32.0
2024-09-22 13:43:36,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=27514.666666666668, ans=0.035
2024-09-22 13:43:42,254 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 1.881e+02 2.197e+02 2.842e+02 5.136e+02, threshold=4.393e+02, percent-clipped=2.0
2024-09-22 13:44:30,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=27701.333333333332, ans=0.004847536231884058
2024-09-22 13:44:47,946 INFO [train.py:1198] (1/4) Epoch 2, batch 2050, loss[loss=0.3625, ctc_loss=0.2763, cr_loss=0.431, over 17295.00 frames. ], tot_loss[loss=0.3619, ctc_loss=0.2771, cr_loss=0.424, over 3370147.99 frames. ], batch size: 46, lr: 3.91e-02, grad_scale: 32.0
2024-09-22 13:44:49,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.82 vs. limit=12.0
2024-09-22 13:45:05,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=27794.666666666668, ans=0.2
2024-09-22 13:45:05,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=27794.666666666668, ans=0.125
2024-09-22 13:46:07,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=27934.666666666668, ans=0.1
2024-09-22 13:46:10,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=27934.666666666668, ans=0.2
2024-09-22 13:46:13,303 INFO [train.py:1198] (1/4) Epoch 2, batch 2100, loss[loss=0.3493, ctc_loss=0.2581, cr_loss=0.4556, over 16654.00 frames. ], tot_loss[loss=0.3621, ctc_loss=0.2774, cr_loss=0.4237, over 3356623.97 frames. ], batch size: 37, lr: 3.90e-02, grad_scale: 32.0
2024-09-22 13:46:27,943 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 1.746e+02 2.180e+02 2.623e+02 4.533e+02, threshold=4.360e+02, percent-clipped=2.0
2024-09-22 13:47:00,302 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=4.07 vs. limit=15.0
2024-09-22 13:47:22,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=28168.0, ans=0.2
2024-09-22 13:47:22,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=28168.0, ans=0.0
2024-09-22 13:47:28,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=28168.0, ans=0.0
2024-09-22 13:47:36,047 INFO [train.py:1198] (1/4) Epoch 2, batch 2150, loss[loss=0.3432, ctc_loss=0.2606, cr_loss=0.4129, over 17015.00 frames. ], tot_loss[loss=0.3613, ctc_loss=0.2766, cr_loss=0.4233, over 3367293.05 frames. ], batch size: 44, lr: 3.89e-02, grad_scale: 32.0
2024-09-22 13:47:37,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=28214.666666666668, ans=0.0
2024-09-22 13:47:40,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=28214.666666666668, ans=0.025
2024-09-22 13:47:49,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=28214.666666666668, ans=0.0
2024-09-22 13:48:15,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=28308.0, ans=0.125
2024-09-22 13:48:17,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=28308.0, ans=0.2
2024-09-22 13:48:23,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=28308.0, ans=0.025
2024-09-22 13:48:23,905 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.77 vs. limit=15.0
2024-09-22 13:48:58,600 INFO [train.py:1198] (1/4) Epoch 2, batch 2200, loss[loss=0.3236, ctc_loss=0.2466, cr_loss=0.3849, over 16692.00 frames. ], tot_loss[loss=0.3587, ctc_loss=0.2745, cr_loss=0.4212, over 3370008.45 frames. ], batch size: 37, lr: 3.89e-02, grad_scale: 32.0
2024-09-22 13:49:00,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=28448.0, ans=0.2
2024-09-22 13:49:03,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=28448.0, ans=0.125
2024-09-22 13:49:05,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=28448.0, ans=0.09899494936611666
2024-09-22 13:49:12,809 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.723e+02 2.085e+02 2.498e+02 4.255e+02, threshold=4.169e+02, percent-clipped=0.0
2024-09-22 13:49:13,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=28494.666666666668, ans=0.025
2024-09-22 13:49:21,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=28494.666666666668, ans=0.2
2024-09-22 13:49:32,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=28541.333333333332, ans=0.0
2024-09-22 13:49:43,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=28541.333333333332, ans=0.125
2024-09-22 13:50:18,353 INFO [train.py:1198] (1/4) Epoch 2, batch 2250, loss[loss=0.3725, ctc_loss=0.2793, cr_loss=0.466, over 17090.00 frames. ], tot_loss[loss=0.3589, ctc_loss=0.2747, cr_loss=0.421, over 3368022.21 frames. ], batch size: 49, lr: 3.88e-02, grad_scale: 32.0
2024-09-22 13:50:23,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=28681.333333333332, ans=0.2
2024-09-22 13:50:23,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=28681.333333333332, ans=0.1
2024-09-22 13:50:49,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=28728.0, ans=0.1
2024-09-22 13:50:59,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=28774.666666666668, ans=0.125
2024-09-22 13:51:07,467 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=16.05 vs. limit=15.0
2024-09-22 13:51:43,311 INFO [train.py:1198] (1/4) Epoch 2, batch 2300, loss[loss=0.3661, ctc_loss=0.2817, cr_loss=0.4218, over 17166.00 frames. ], tot_loss[loss=0.3596, ctc_loss=0.2751, cr_loss=0.4224, over 3374962.68 frames. ], batch size: 45, lr: 3.87e-02, grad_scale: 32.0
2024-09-22 13:51:43,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=28914.666666666668, ans=0.125
2024-09-22 13:52:00,470 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 1.812e+02 2.268e+02 2.899e+02 4.767e+02, threshold=4.537e+02, percent-clipped=4.0
2024-09-22 13:52:09,150 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=9.90 vs. limit=15.0
2024-09-22 13:52:20,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=29008.0, ans=0.1
2024-09-22 13:52:37,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=29054.666666666668, ans=0.004553333333333333
2024-09-22 13:52:57,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=29101.333333333332, ans=0.125
2024-09-22 13:53:07,831 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.56 vs. limit=15.0
2024-09-22 13:53:08,675 INFO [train.py:1198] (1/4) Epoch 2, batch 2350, loss[loss=0.3126, ctc_loss=0.2377, cr_loss=0.3743, over 17110.00 frames. ], tot_loss[loss=0.361, ctc_loss=0.2761, cr_loss=0.4245, over 3366108.05 frames. ], batch size: 40, lr: 3.87e-02, grad_scale: 32.0
2024-09-22 13:53:08,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=29148.0, ans=0.5
2024-09-22 13:53:48,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=29241.333333333332, ans=0.125
2024-09-22 13:53:53,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=29241.333333333332, ans=0.0
2024-09-22 13:54:18,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=29334.666666666668, ans=0.125
2024-09-22 13:54:27,694 INFO [train.py:1198] (1/4) Epoch 2, batch 2400, loss[loss=0.4664, ctc_loss=0.3789, cr_loss=0.4376, over 11575.00 frames. ], tot_loss[loss=0.3619, ctc_loss=0.2769, cr_loss=0.4254, over 3358742.69 frames. ], batch size: 123, lr: 3.86e-02, grad_scale: 32.0
2024-09-22 13:54:41,835 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.409e+02 1.714e+02 1.995e+02 2.694e+02 4.976e+02, threshold=3.990e+02, percent-clipped=1.0
2024-09-22 13:54:57,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.34 vs. limit=12.0
2024-09-22 13:55:23,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=29521.333333333332, ans=0.2
2024-09-22 13:55:52,213 INFO [train.py:1198] (1/4) Epoch 2, batch 2450, loss[loss=0.3665, ctc_loss=0.2795, cr_loss=0.4351, over 17011.00 frames. ], tot_loss[loss=0.3613, ctc_loss=0.2763, cr_loss=0.4252, over 3364017.56 frames. ], batch size: 52, lr: 3.86e-02, grad_scale: 32.0
2024-09-22 13:56:03,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=29614.666666666668, ans=0.1
2024-09-22 13:56:43,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=29754.666666666668, ans=0.125
2024-09-22 13:57:04,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=29801.333333333332, ans=0.125
2024-09-22 13:57:14,734 INFO [train.py:1198] (1/4) Epoch 2, batch 2500, loss[loss=0.3134, ctc_loss=0.2338, cr_loss=0.3976, over 17088.00 frames. ], tot_loss[loss=0.3613, ctc_loss=0.2763, cr_loss=0.4251, over 3359153.55 frames. ], batch size: 43, lr: 3.85e-02, grad_scale: 32.0
2024-09-22 13:57:16,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=29848.0, ans=0.125
2024-09-22 13:57:29,030 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 1.845e+02 2.309e+02 3.069e+02 4.385e+02, threshold=4.618e+02, percent-clipped=5.0
2024-09-22 13:57:29,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=29894.666666666668, ans=0.125
2024-09-22 13:57:43,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=29894.666666666668, ans=0.125
2024-09-22 13:57:49,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=29941.333333333332, ans=0.025
2024-09-22 13:58:06,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=29988.0, ans=0.1
2024-09-22 13:58:17,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=29988.0, ans=0.125
2024-09-22 13:58:33,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=30034.666666666668, ans=0.0
2024-09-22 13:58:36,482 INFO [train.py:1198] (1/4) Epoch 2, batch 2550, loss[loss=0.3472, ctc_loss=0.2605, cr_loss=0.4338, over 17021.00 frames. ], tot_loss[loss=0.3605, ctc_loss=0.2756, cr_loss=0.4244, over 3349033.07 frames. ], batch size: 51, lr: 3.84e-02, grad_scale: 32.0
2024-09-22 13:58:48,058 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 13:58:58,270 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.44 vs. limit=15.0
2024-09-22 13:58:59,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=30128.0, ans=0.125
2024-09-22 13:59:05,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=30128.0, ans=0.1
2024-09-22 13:59:05,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=30128.0, ans=0.125
2024-09-22 13:59:24,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=30221.333333333332, ans=0.0
2024-09-22 13:59:41,535 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=21.70 vs. limit=22.5
2024-09-22 13:59:52,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.max_abs, batch_count=30268.0, ans=10.0
2024-09-22 13:59:56,515 INFO [train.py:1198] (1/4) Epoch 2, batch 2600, loss[loss=0.3618, ctc_loss=0.2756, cr_loss=0.4307, over 17095.00 frames. ], tot_loss[loss=0.3611, ctc_loss=0.276, cr_loss=0.4253, over 3347540.92 frames. ], batch size: 49, lr: 3.84e-02, grad_scale: 32.0
2024-09-22 14:00:11,073 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 1.903e+02 2.207e+02 2.848e+02 4.508e+02, threshold=4.414e+02, percent-clipped=0.0
2024-09-22 14:00:23,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=30361.333333333332, ans=0.0
2024-09-22 14:00:26,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=30361.333333333332, ans=0.2
2024-09-22 14:00:29,486 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.15 vs. limit=8.0
2024-09-22 14:00:54,192 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.42 vs. limit=12.0
2024-09-22 14:01:20,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=30548.0, ans=0.1
2024-09-22 14:01:21,837 INFO [train.py:1198] (1/4) Epoch 2, batch 2650, loss[loss=0.3829, ctc_loss=0.2952, cr_loss=0.4386, over 16655.00 frames. ], tot_loss[loss=0.3575, ctc_loss=0.2729, cr_loss=0.423, over 3358329.45 frames. ], batch size: 66, lr: 3.83e-02, grad_scale: 32.0
2024-09-22 14:01:28,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=30548.0, ans=0.0
2024-09-22 14:01:39,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=30594.666666666668, ans=0.0
2024-09-22 14:01:41,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=30594.666666666668, ans=0.004218550724637681
2024-09-22 14:01:59,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=30641.333333333332, ans=0.2
2024-09-22 14:02:09,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=30641.333333333332, ans=0.00420840579710145
2024-09-22 14:02:19,402 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.15 vs. limit=10.0
2024-09-22 14:02:46,375 INFO [train.py:1198] (1/4) Epoch 2, batch 2700, loss[loss=0.3248, ctc_loss=0.2445, cr_loss=0.4018, over 17154.00 frames. ], tot_loss[loss=0.3564, ctc_loss=0.2719, cr_loss=0.4226, over 3368283.95 frames. ], batch size: 45, lr: 3.82e-02, grad_scale: 32.0
2024-09-22 14:03:00,819 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 1.814e+02 2.097e+02 2.446e+02 4.164e+02, threshold=4.194e+02, percent-clipped=0.0
2024-09-22 14:03:01,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=30828.0, ans=0.004167826086956521
2024-09-22 14:03:04,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=30828.0, ans=0.125
2024-09-22 14:03:05,138 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=15.01 vs. limit=22.5
2024-09-22 14:03:07,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=30828.0, ans=0.125
2024-09-22 14:03:26,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=30874.666666666668, ans=0.125
2024-09-22 14:03:39,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=30921.333333333332, ans=0.0
2024-09-22 14:03:55,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=30968.0, ans=0.0
2024-09-22 14:04:05,949 INFO [train.py:1198] (1/4) Epoch 2, batch 2750, loss[loss=0.3678, ctc_loss=0.2843, cr_loss=0.4171, over 16129.00 frames. ], tot_loss[loss=0.3565, ctc_loss=0.2719, cr_loss=0.423, over 3359744.52 frames. ], batch size: 74, lr: 3.82e-02, grad_scale: 32.0
2024-09-22 14:04:35,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=31061.333333333332, ans=0.0
2024-09-22 14:04:55,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=31154.666666666668, ans=0.0
2024-09-22 14:04:57,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=31154.666666666668, ans=0.125
2024-09-22 14:05:05,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=31154.666666666668, ans=0.125
2024-09-22 14:05:07,472 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=14.98 vs. limit=15.0
2024-09-22 14:05:10,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=31201.333333333332, ans=0.5
2024-09-22 14:05:31,359 INFO [train.py:1198] (1/4) Epoch 2, batch 2800, loss[loss=0.3472, ctc_loss=0.267, cr_loss=0.4011, over 16485.00 frames. ], tot_loss[loss=0.3573, ctc_loss=0.2725, cr_loss=0.4244, over 3363477.03 frames. ], batch size: 66, lr: 3.81e-02, grad_scale: 32.0
2024-09-22 14:05:45,623 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 1.818e+02 2.147e+02 2.657e+02 4.230e+02, threshold=4.294e+02, percent-clipped=1.0
2024-09-22 14:05:49,447 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.55 vs. limit=10.0
2024-09-22 14:05:52,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=31294.666666666668, ans=0.1
2024-09-22 14:06:17,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=31388.0, ans=0.0
2024-09-22 14:06:27,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=31388.0, ans=0.125
2024-09-22 14:06:49,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=31434.666666666668, ans=0.1
2024-09-22 14:06:53,503 INFO [train.py:1198] (1/4) Epoch 2, batch 2850, loss[loss=0.3494, ctc_loss=0.2636, cr_loss=0.4288, over 17323.00 frames. ], tot_loss[loss=0.3569, ctc_loss=0.272, cr_loss=0.4242, over 3370683.08 frames. ], batch size: 46, lr: 3.80e-02, grad_scale: 32.0
2024-09-22 14:07:31,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=31574.666666666668, ans=0.025
2024-09-22 14:08:15,441 INFO [train.py:1198] (1/4) Epoch 2, batch 2900, loss[loss=0.3274, ctc_loss=0.2489, cr_loss=0.3924, over 17253.00 frames. ], tot_loss[loss=0.3559, ctc_loss=0.2713, cr_loss=0.4227, over 3369749.80 frames. ], batch size: 44, lr: 3.80e-02, grad_scale: 32.0
2024-09-22 14:08:24,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=31714.666666666668, ans=0.125
2024-09-22 14:08:29,968 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.781e+02 2.101e+02 2.670e+02 4.501e+02, threshold=4.202e+02, percent-clipped=1.0
2024-09-22 14:08:56,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.32 vs. limit=15.0
2024-09-22 14:09:02,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=31854.666666666668, ans=0.125
2024-09-22 14:09:10,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=31854.666666666668, ans=0.2
2024-09-22 14:09:14,327 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=5.37 vs. limit=15.0
2024-09-22 14:09:35,328 INFO [train.py:1198] (1/4) Epoch 2, batch 2950, loss[loss=0.3427, ctc_loss=0.2562, cr_loss=0.4324, over 17147.00 frames. ], tot_loss[loss=0.3553, ctc_loss=0.2707, cr_loss=0.4231, over 3379262.72 frames. ], batch size: 48, lr: 3.79e-02, grad_scale: 32.0
2024-09-22 14:10:09,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.91 vs. limit=6.0
2024-09-22 14:10:18,099 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 14:10:25,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=32041.333333333332, ans=0.0
2024-09-22 14:10:50,744 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 14:10:51,002 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.97 vs. limit=12.0
2024-09-22 14:10:53,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=32134.666666666668, ans=0.1
2024-09-22 14:10:59,597 INFO [train.py:1198] (1/4) Epoch 2, batch 3000, loss[loss=0.3016, ctc_loss=0.2204, cr_loss=0.4063, over 16951.00 frames. ], tot_loss[loss=0.3558, ctc_loss=0.2711, cr_loss=0.4232, over 3362211.24 frames. ], batch size: 42, lr: 3.79e-02, grad_scale: 32.0
2024-09-22 14:10:59,597 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 14:11:15,222 INFO [train.py:1230] (1/4) Epoch 2, validation: loss=0.0967, ctc_loss=0.0967, cr_loss=8.169e-15, over 944034.00 frames. 
2024-09-22 14:11:15,223 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 14:11:29,193 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 1.814e+02 2.374e+02 2.965e+02 6.190e+02, threshold=4.748e+02, percent-clipped=4.0
2024-09-22 14:11:38,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=32228.0, ans=0.0038634782608695647
2024-09-22 14:11:51,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=32274.666666666668, ans=0.0038533333333333336
2024-09-22 14:11:52,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.min_positive, batch_count=32274.666666666668, ans=0.05
2024-09-22 14:12:33,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=32414.666666666668, ans=0.0
2024-09-22 14:12:35,205 INFO [train.py:1198] (1/4) Epoch 2, batch 3050, loss[loss=0.4492, ctc_loss=0.3656, cr_loss=0.418, over 12396.00 frames. ], tot_loss[loss=0.3551, ctc_loss=0.2706, cr_loss=0.4223, over 3363766.86 frames. ], batch size: 123, lr: 3.78e-02, grad_scale: 32.0
2024-09-22 14:12:48,943 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.97 vs. limit=12.0
2024-09-22 14:12:54,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=32461.333333333332, ans=0.125
2024-09-22 14:13:12,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=32508.0, ans=0.0038026086956521746
2024-09-22 14:13:48,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=32601.333333333332, ans=0.0
2024-09-22 14:13:52,646 INFO [train.py:1198] (1/4) Epoch 2, batch 3100, loss[loss=0.3463, ctc_loss=0.2602, cr_loss=0.4308, over 17296.00 frames. ], tot_loss[loss=0.3582, ctc_loss=0.2736, cr_loss=0.4232, over 3330353.75 frames. ], batch size: 49, lr: 3.77e-02, grad_scale: 32.0
2024-09-22 14:14:10,888 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 1.798e+02 2.252e+02 2.859e+02 4.646e+02, threshold=4.504e+02, percent-clipped=0.0
2024-09-22 14:14:21,125 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.65 vs. limit=10.0
2024-09-22 14:14:33,205 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.22 vs. limit=15.0
2024-09-22 14:14:40,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=32788.0, ans=0.0037417391304347828
2024-09-22 14:14:45,980 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.75 vs. limit=10.0
2024-09-22 14:14:57,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=32834.666666666664, ans=0.125
2024-09-22 14:15:08,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=32834.666666666664, ans=0.125
2024-09-22 14:15:12,721 INFO [train.py:1198] (1/4) Epoch 2, batch 3150, loss[loss=0.3881, ctc_loss=0.3016, cr_loss=0.4326, over 16597.00 frames. ], tot_loss[loss=0.3581, ctc_loss=0.2735, cr_loss=0.4227, over 3335298.07 frames. ], batch size: 66, lr: 3.77e-02, grad_scale: 32.0
2024-09-22 14:15:22,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=32881.333333333336, ans=0.125
2024-09-22 14:15:52,673 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=25.66 vs. limit=22.5
2024-09-22 14:16:10,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.max_abs, batch_count=33021.333333333336, ans=10.0
2024-09-22 14:16:30,722 INFO [train.py:1198] (1/4) Epoch 2, batch 3200, loss[loss=0.3027, ctc_loss=0.226, cr_loss=0.3838, over 17097.00 frames. ], tot_loss[loss=0.3567, ctc_loss=0.2723, cr_loss=0.4219, over 3332468.01 frames. ], batch size: 43, lr: 3.76e-02, grad_scale: 32.0
2024-09-22 14:16:38,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=33114.666666666664, ans=0.1
2024-09-22 14:16:45,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=33161.333333333336, ans=0.125
2024-09-22 14:16:46,457 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 1.721e+02 1.992e+02 2.305e+02 3.966e+02, threshold=3.983e+02, percent-clipped=0.0
2024-09-22 14:17:17,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=33254.666666666664, ans=0.0036402898550724648
2024-09-22 14:17:22,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=33254.666666666664, ans=0.125
2024-09-22 14:17:37,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=33301.333333333336, ans=0.125
2024-09-22 14:17:48,522 INFO [train.py:1198] (1/4) Epoch 2, batch 3250, loss[loss=0.3236, ctc_loss=0.2349, cr_loss=0.4434, over 17077.00 frames. ], tot_loss[loss=0.3554, ctc_loss=0.271, cr_loss=0.4219, over 3349874.89 frames. ], batch size: 43, lr: 3.75e-02, grad_scale: 32.0
2024-09-22 14:17:51,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=33348.0, ans=0.0
2024-09-22 14:18:02,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=33394.666666666664, ans=0.025
2024-09-22 14:18:08,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=33394.666666666664, ans=0.0
2024-09-22 14:18:24,907 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=24.86 vs. limit=22.5
2024-09-22 14:18:29,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=33441.333333333336, ans=0.025
2024-09-22 14:19:06,343 INFO [train.py:1198] (1/4) Epoch 2, batch 3300, loss[loss=0.3283, ctc_loss=0.2469, cr_loss=0.4067, over 17047.00 frames. ], tot_loss[loss=0.3555, ctc_loss=0.2711, cr_loss=0.4222, over 3350381.86 frames. ], batch size: 39, lr: 3.75e-02, grad_scale: 32.0
2024-09-22 14:19:13,304 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.54 vs. limit=15.0
2024-09-22 14:19:22,039 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 1.852e+02 2.253e+02 3.068e+02 5.078e+02, threshold=4.507e+02, percent-clipped=5.0
2024-09-22 14:19:34,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=33628.0, ans=0.125
2024-09-22 14:20:28,449 INFO [train.py:1198] (1/4) Epoch 2, batch 3350, loss[loss=0.3816, ctc_loss=0.293, cr_loss=0.4433, over 16433.00 frames. ], tot_loss[loss=0.3572, ctc_loss=0.2724, cr_loss=0.4239, over 3346245.42 frames. ], batch size: 66, lr: 3.74e-02, grad_scale: 32.0
2024-09-22 14:20:40,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=33814.666666666664, ans=0.125
2024-09-22 14:20:43,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.24 vs. limit=22.5
2024-09-22 14:20:50,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.39 vs. limit=12.0
2024-09-22 14:21:03,533 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.56 vs. limit=15.0
2024-09-22 14:21:07,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=33908.0, ans=0.125
2024-09-22 14:21:38,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=34001.333333333336, ans=0.125
2024-09-22 14:21:46,546 INFO [train.py:1198] (1/4) Epoch 2, batch 3400, loss[loss=0.2866, ctc_loss=0.2132, cr_loss=0.3668, over 17256.00 frames. ], tot_loss[loss=0.3554, ctc_loss=0.2708, cr_loss=0.4231, over 3351688.63 frames. ], batch size: 44, lr: 3.74e-02, grad_scale: 32.0
2024-09-22 14:21:51,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=34048.0, ans=0.0
2024-09-22 14:21:54,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=34048.0, ans=0.125
2024-09-22 14:21:57,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=34048.0, ans=0.0
2024-09-22 14:22:02,234 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 1.761e+02 2.090e+02 2.503e+02 3.941e+02, threshold=4.179e+02, percent-clipped=0.0
2024-09-22 14:22:36,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=34188.0, ans=0.025
2024-09-22 14:22:42,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=34188.0, ans=0.125
2024-09-22 14:22:44,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=34188.0, ans=0.1
2024-09-22 14:22:48,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=34188.0, ans=0.5
2024-09-22 14:23:04,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=34234.666666666664, ans=0.1
2024-09-22 14:23:07,087 INFO [train.py:1198] (1/4) Epoch 2, batch 3450, loss[loss=0.3596, ctc_loss=0.2746, cr_loss=0.4249, over 16445.00 frames. ], tot_loss[loss=0.3556, ctc_loss=0.2709, cr_loss=0.4237, over 3348087.74 frames. ], batch size: 66, lr: 3.73e-02, grad_scale: 16.0
2024-09-22 14:23:08,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=34281.333333333336, ans=0.0
2024-09-22 14:23:43,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=34374.666666666664, ans=10.0
2024-09-22 14:24:07,235 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=22.97 vs. limit=22.5
2024-09-22 14:24:08,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=34421.333333333336, ans=0.025
2024-09-22 14:24:26,671 INFO [train.py:1198] (1/4) Epoch 2, batch 3500, loss[loss=0.3157, ctc_loss=0.2379, cr_loss=0.3891, over 17037.00 frames. ], tot_loss[loss=0.3537, ctc_loss=0.2691, cr_loss=0.4227, over 3355320.14 frames. ], batch size: 39, lr: 3.72e-02, grad_scale: 16.0
2024-09-22 14:24:39,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=34514.666666666664, ans=0.0
2024-09-22 14:24:44,120 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 1.790e+02 2.190e+02 2.610e+02 4.602e+02, threshold=4.381e+02, percent-clipped=3.0
2024-09-22 14:24:57,157 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.07 vs. limit=15.0
2024-09-22 14:24:58,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=34608.0, ans=0.025
2024-09-22 14:25:11,503 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.65 vs. limit=10.0
2024-09-22 14:25:14,491 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.94 vs. limit=15.0
2024-09-22 14:25:19,163 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.74 vs. limit=22.5
2024-09-22 14:25:33,182 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.19 vs. limit=10.0
2024-09-22 14:25:35,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=34701.333333333336, ans=0.1
2024-09-22 14:25:35,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=34701.333333333336, ans=0.0033257971014492745
2024-09-22 14:25:42,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=34701.333333333336, ans=0.125
2024-09-22 14:25:44,830 INFO [train.py:1198] (1/4) Epoch 2, batch 3550, loss[loss=0.3042, ctc_loss=0.2282, cr_loss=0.3802, over 17172.00 frames. ], tot_loss[loss=0.3544, ctc_loss=0.2698, cr_loss=0.4231, over 3352638.59 frames. ], batch size: 45, lr: 3.72e-02, grad_scale: 16.0
2024-09-22 14:25:53,136 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.33 vs. limit=15.0
2024-09-22 14:26:01,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=34794.666666666664, ans=0.0
2024-09-22 14:26:12,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=34794.666666666664, ans=0.125
2024-09-22 14:26:25,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=34841.333333333336, ans=0.04949747468305833
2024-09-22 14:26:27,353 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.65 vs. limit=15.0
2024-09-22 14:26:39,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=34888.0, ans=0.025
2024-09-22 14:26:53,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=34934.666666666664, ans=0.05
2024-09-22 14:27:02,440 INFO [train.py:1198] (1/4) Epoch 2, batch 3600, loss[loss=0.3536, ctc_loss=0.27, cr_loss=0.4181, over 17356.00 frames. ], tot_loss[loss=0.354, ctc_loss=0.2695, cr_loss=0.4227, over 3360702.52 frames. ], batch size: 48, lr: 3.71e-02, grad_scale: 32.0
2024-09-22 14:27:19,633 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 1.714e+02 2.123e+02 2.581e+02 4.355e+02, threshold=4.245e+02, percent-clipped=0.0
2024-09-22 14:28:02,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=35121.333333333336, ans=0.125
2024-09-22 14:28:18,047 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=5.08 vs. limit=15.0
2024-09-22 14:28:20,465 INFO [train.py:1198] (1/4) Epoch 2, batch 3650, loss[loss=0.4025, ctc_loss=0.3157, cr_loss=0.4339, over 16696.00 frames. ], tot_loss[loss=0.3554, ctc_loss=0.2707, cr_loss=0.4238, over 3340610.23 frames. ], batch size: 61, lr: 3.70e-02, grad_scale: 32.0
2024-09-22 14:28:29,160 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=5.85 vs. limit=12.0
2024-09-22 14:28:33,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=35214.666666666664, ans=0.025
2024-09-22 14:29:12,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=35354.666666666664, ans=0.1
2024-09-22 14:29:20,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=35354.666666666664, ans=0.125
2024-09-22 14:29:28,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=35401.333333333336, ans=0.125
2024-09-22 14:29:28,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=35401.333333333336, ans=0.125
2024-09-22 14:29:28,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=35401.333333333336, ans=0.07
2024-09-22 14:29:40,841 INFO [train.py:1198] (1/4) Epoch 2, batch 3700, loss[loss=0.3307, ctc_loss=0.2482, cr_loss=0.4128, over 17133.00 frames. ], tot_loss[loss=0.3544, ctc_loss=0.2698, cr_loss=0.4231, over 3348947.10 frames. ], batch size: 48, lr: 3.70e-02, grad_scale: 16.0
2024-09-22 14:29:56,005 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=20.61 vs. limit=22.5
2024-09-22 14:29:59,940 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 1.763e+02 2.146e+02 2.787e+02 4.998e+02, threshold=4.291e+02, percent-clipped=2.0
2024-09-22 14:30:16,249 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.48 vs. limit=10.0
2024-09-22 14:30:37,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=35588.0, ans=0.1
2024-09-22 14:30:40,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=35588.0, ans=0.2
2024-09-22 14:30:46,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=7.35 vs. limit=15.0
2024-09-22 14:30:59,291 INFO [train.py:1198] (1/4) Epoch 2, batch 3750, loss[loss=0.4591, ctc_loss=0.3641, cr_loss=0.4749, over 11970.00 frames. ], tot_loss[loss=0.3544, ctc_loss=0.27, cr_loss=0.422, over 3343348.38 frames. ], batch size: 123, lr: 3.69e-02, grad_scale: 16.0
2024-09-22 14:31:53,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=35821.333333333336, ans=0.125
2024-09-22 14:31:55,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=35821.333333333336, ans=0.2
2024-09-22 14:32:07,881 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 14:32:12,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=35868.0, ans=0.0
2024-09-22 14:32:15,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=35868.0, ans=0.2
2024-09-22 14:32:17,562 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.82 vs. limit=15.0
2024-09-22 14:32:18,707 INFO [train.py:1198] (1/4) Epoch 2, batch 3800, loss[loss=0.3208, ctc_loss=0.2474, cr_loss=0.367, over 17095.00 frames. ], tot_loss[loss=0.355, ctc_loss=0.2706, cr_loss=0.422, over 3322969.21 frames. ], batch size: 39, lr: 3.69e-02, grad_scale: 16.0
2024-09-22 14:32:31,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=35914.666666666664, ans=0.125
2024-09-22 14:32:34,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=35961.333333333336, ans=0.0030518840579710146
2024-09-22 14:32:37,259 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 1.782e+02 2.185e+02 2.503e+02 5.708e+02, threshold=4.370e+02, percent-clipped=5.0
2024-09-22 14:33:07,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=36054.666666666664, ans=0.2
2024-09-22 14:33:19,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=36101.333333333336, ans=0.2
2024-09-22 14:33:35,545 INFO [train.py:1198] (1/4) Epoch 2, batch 3850, loss[loss=0.4656, ctc_loss=0.379, cr_loss=0.433, over 12042.00 frames. ], tot_loss[loss=0.3596, ctc_loss=0.2749, cr_loss=0.4235, over 3277450.06 frames. ], batch size: 123, lr: 3.68e-02, grad_scale: 16.0
2024-09-22 14:33:46,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=36148.0, ans=0.1
2024-09-22 14:33:50,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=36194.666666666664, ans=0.025
2024-09-22 14:34:16,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=36241.333333333336, ans=0.125
2024-09-22 14:35:36,876 INFO [train.py:1198] (1/4) Epoch 3, batch 0, loss[loss=0.3515, ctc_loss=0.2734, cr_loss=0.3904, over 17111.00 frames. ], tot_loss[loss=0.3515, ctc_loss=0.2734, cr_loss=0.3904, over 17111.00 frames. ], batch size: 40, lr: 3.49e-02, grad_scale: 32.0
2024-09-22 14:35:36,876 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 14:35:52,236 INFO [train.py:1230] (1/4) Epoch 3, validation: loss=0.1002, ctc_loss=0.1002, cr_loss=7.948e-15, over 944034.00 frames. 
2024-09-22 14:35:52,237 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 14:35:55,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=36362.666666666664, ans=0.2
2024-09-22 14:35:58,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=36362.666666666664, ans=0.0
2024-09-22 14:36:13,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=36409.333333333336, ans=0.09899494936611666
2024-09-22 14:36:20,867 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 1.866e+02 2.182e+02 2.689e+02 4.735e+02, threshold=4.364e+02, percent-clipped=1.0
2024-09-22 14:36:24,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=36409.333333333336, ans=0.125
2024-09-22 14:36:32,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=36456.0, ans=0.2
2024-09-22 14:36:40,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=36456.0, ans=0.1
2024-09-22 14:36:51,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=36502.666666666664, ans=0.125
2024-09-22 14:37:17,946 INFO [train.py:1198] (1/4) Epoch 3, batch 50, loss[loss=0.4422, ctc_loss=0.3562, cr_loss=0.4303, over 11619.00 frames. ], tot_loss[loss=0.349, ctc_loss=0.2656, cr_loss=0.4171, over 741870.71 frames. ], batch size: 123, lr: 3.49e-02, grad_scale: 32.0
2024-09-22 14:37:28,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=36596.0, ans=0.125
2024-09-22 14:37:29,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=36596.0, ans=0.2
2024-09-22 14:37:31,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=36596.0, ans=0.2
2024-09-22 14:37:47,827 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=22.63 vs. limit=22.5
2024-09-22 14:37:59,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=36689.333333333336, ans=0.5
2024-09-22 14:38:28,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=36782.666666666664, ans=0.5
2024-09-22 14:38:31,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=36782.666666666664, ans=0.125
2024-09-22 14:38:41,849 INFO [train.py:1198] (1/4) Epoch 3, batch 100, loss[loss=0.3443, ctc_loss=0.2581, cr_loss=0.431, over 17037.00 frames. ], tot_loss[loss=0.3491, ctc_loss=0.2653, cr_loss=0.4192, over 1314947.13 frames. ], batch size: 39, lr: 3.48e-02, grad_scale: 32.0
2024-09-22 14:38:47,271 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.33 vs. limit=15.0
2024-09-22 14:39:07,121 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 1.767e+02 2.113e+02 2.664e+02 5.595e+02, threshold=4.227e+02, percent-clipped=3.0
2024-09-22 14:39:12,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=36922.666666666664, ans=0.0
2024-09-22 14:40:01,180 INFO [train.py:1198] (1/4) Epoch 3, batch 150, loss[loss=0.3198, ctc_loss=0.2371, cr_loss=0.4134, over 17188.00 frames. ], tot_loss[loss=0.3437, ctc_loss=0.2602, cr_loss=0.4175, over 1772258.00 frames. ], batch size: 41, lr: 3.47e-02, grad_scale: 32.0
2024-09-22 14:40:12,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=37062.666666666664, ans=0.125
2024-09-22 14:41:17,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=37249.333333333336, ans=0.0027718840579710148
2024-09-22 14:41:23,508 INFO [train.py:1198] (1/4) Epoch 3, batch 200, loss[loss=0.3822, ctc_loss=0.2935, cr_loss=0.444, over 17021.00 frames. ], tot_loss[loss=0.3462, ctc_loss=0.2624, cr_loss=0.419, over 2121500.44 frames. ], batch size: 53, lr: 3.47e-02, grad_scale: 32.0
2024-09-22 14:41:24,331 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.42 vs. limit=12.0
2024-09-22 14:41:27,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=37296.0, ans=0.125
2024-09-22 14:41:48,947 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.08 vs. limit=15.0
2024-09-22 14:41:50,321 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.47 vs. limit=15.0
2024-09-22 14:41:51,077 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 1.679e+02 1.968e+02 2.454e+02 4.058e+02, threshold=3.935e+02, percent-clipped=0.0
2024-09-22 14:42:02,773 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=6.80 vs. limit=15.0
2024-09-22 14:42:43,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=37482.666666666664, ans=0.09899494936611666
2024-09-22 14:42:50,682 INFO [train.py:1198] (1/4) Epoch 3, batch 250, loss[loss=0.3467, ctc_loss=0.2634, cr_loss=0.4165, over 17139.00 frames. ], tot_loss[loss=0.3464, ctc_loss=0.2623, cr_loss=0.4205, over 2404154.54 frames. ], batch size: 48, lr: 3.46e-02, grad_scale: 32.0
2024-09-22 14:43:01,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=37529.333333333336, ans=0.2
2024-09-22 14:43:14,011 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.79 vs. limit=22.5
2024-09-22 14:43:14,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn1.whiten.whitening_limit, batch_count=37576.0, ans=22.5
2024-09-22 14:43:14,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=37576.0, ans=0.2
2024-09-22 14:43:59,994 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-22 14:44:12,443 INFO [train.py:1198] (1/4) Epoch 3, batch 300, loss[loss=0.3615, ctc_loss=0.2785, cr_loss=0.4151, over 17147.00 frames. ], tot_loss[loss=0.3465, ctc_loss=0.2624, cr_loss=0.4207, over 2614846.96 frames. ], batch size: 48, lr: 3.46e-02, grad_scale: 32.0
2024-09-22 14:44:26,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=37809.333333333336, ans=0.125
2024-09-22 14:44:34,114 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=16.98 vs. limit=22.5
2024-09-22 14:44:37,527 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.687e+02 1.987e+02 2.495e+02 5.356e+02, threshold=3.975e+02, percent-clipped=4.0
2024-09-22 14:45:06,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=37902.666666666664, ans=0.125
2024-09-22 14:45:11,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=37902.666666666664, ans=0.125
2024-09-22 14:45:13,573 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.38 vs. limit=15.0
2024-09-22 14:45:14,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=37949.333333333336, ans=0.0
2024-09-22 14:45:22,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=37949.333333333336, ans=0.125
2024-09-22 14:45:31,398 INFO [train.py:1198] (1/4) Epoch 3, batch 350, loss[loss=0.3658, ctc_loss=0.2791, cr_loss=0.4336, over 17310.00 frames. ], tot_loss[loss=0.3464, ctc_loss=0.2621, cr_loss=0.4214, over 2784505.83 frames. ], batch size: 51, lr: 3.45e-02, grad_scale: 32.0
2024-09-22 14:45:52,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=38042.666666666664, ans=0.125
2024-09-22 14:46:03,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=38089.333333333336, ans=0.00258927536231884
2024-09-22 14:46:30,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=38136.0, ans=0.1
2024-09-22 14:46:33,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=38136.0, ans=0.125
2024-09-22 14:46:57,024 INFO [train.py:1198] (1/4) Epoch 3, batch 400, loss[loss=0.3661, ctc_loss=0.2703, cr_loss=0.479, over 17354.00 frames. ], tot_loss[loss=0.3467, ctc_loss=0.2625, cr_loss=0.4208, over 2904543.52 frames. ], batch size: 48, lr: 3.45e-02, grad_scale: 32.0
2024-09-22 14:47:05,894 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.76 vs. limit=10.0
2024-09-22 14:47:06,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=38229.333333333336, ans=0.125
2024-09-22 14:47:25,984 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.690e+02 1.971e+02 2.820e+02 5.296e+02, threshold=3.942e+02, percent-clipped=6.0
2024-09-22 14:48:19,966 INFO [train.py:1198] (1/4) Epoch 3, batch 450, loss[loss=0.3382, ctc_loss=0.2517, cr_loss=0.4327, over 17142.00 frames. ], tot_loss[loss=0.3448, ctc_loss=0.2609, cr_loss=0.4197, over 3008899.30 frames. ], batch size: 48, lr: 3.44e-02, grad_scale: 32.0
2024-09-22 14:49:27,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=38649.333333333336, ans=0.0
2024-09-22 14:49:41,802 INFO [train.py:1198] (1/4) Epoch 3, batch 500, loss[loss=0.3852, ctc_loss=0.2976, cr_loss=0.4381, over 17016.00 frames. ], tot_loss[loss=0.3439, ctc_loss=0.2601, cr_loss=0.4193, over 3091687.67 frames. ], batch size: 44, lr: 3.43e-02, grad_scale: 32.0
2024-09-22 14:50:07,129 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.795e+02 2.310e+02 2.743e+02 4.759e+02, threshold=4.620e+02, percent-clipped=4.0
2024-09-22 14:50:07,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=38742.666666666664, ans=0.2
2024-09-22 14:51:01,266 INFO [train.py:1198] (1/4) Epoch 3, batch 550, loss[loss=0.4033, ctc_loss=0.3047, cr_loss=0.4928, over 17066.00 frames. ], tot_loss[loss=0.345, ctc_loss=0.2608, cr_loss=0.421, over 3154149.64 frames. ], batch size: 56, lr: 3.43e-02, grad_scale: 32.0
2024-09-22 14:51:21,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=38976.0, ans=0.0023965217391304352
2024-09-22 14:51:36,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=39022.666666666664, ans=0.1
2024-09-22 14:51:37,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=39022.666666666664, ans=0.125
2024-09-22 14:51:51,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=39022.666666666664, ans=0.0
2024-09-22 14:52:10,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=39116.0, ans=0.125
2024-09-22 14:52:28,850 INFO [train.py:1198] (1/4) Epoch 3, batch 600, loss[loss=0.3308, ctc_loss=0.2491, cr_loss=0.4086, over 17315.00 frames. ], tot_loss[loss=0.3459, ctc_loss=0.2617, cr_loss=0.4212, over 3187960.85 frames. ], batch size: 46, lr: 3.42e-02, grad_scale: 32.0
2024-09-22 14:52:29,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=39162.666666666664, ans=0.2
2024-09-22 14:52:37,514 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.84 vs. limit=15.0
2024-09-22 14:52:54,491 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.738e+02 2.025e+02 2.578e+02 4.577e+02, threshold=4.049e+02, percent-clipped=0.0
2024-09-22 14:53:18,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=39302.666666666664, ans=0.0023255072463768123
2024-09-22 14:53:34,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=39349.333333333336, ans=0.2
2024-09-22 14:53:34,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=39349.333333333336, ans=0.2
2024-09-22 14:53:38,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=39349.333333333336, ans=0.0
2024-09-22 14:53:47,385 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.40 vs. limit=12.0
2024-09-22 14:53:51,256 INFO [train.py:1198] (1/4) Epoch 3, batch 650, loss[loss=0.3356, ctc_loss=0.2527, cr_loss=0.4144, over 17038.00 frames. ], tot_loss[loss=0.3471, ctc_loss=0.2627, cr_loss=0.4221, over 3211005.86 frames. ], batch size: 44, lr: 3.42e-02, grad_scale: 32.0
2024-09-22 14:54:14,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.97 vs. limit=10.0
2024-09-22 14:54:33,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=39489.333333333336, ans=0.2
2024-09-22 14:54:56,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=39582.666666666664, ans=0.125
2024-09-22 14:55:10,516 INFO [train.py:1198] (1/4) Epoch 3, batch 700, loss[loss=0.3856, ctc_loss=0.2949, cr_loss=0.4538, over 17013.00 frames. ], tot_loss[loss=0.3463, ctc_loss=0.2618, cr_loss=0.4225, over 3246434.26 frames. ], batch size: 53, lr: 3.41e-02, grad_scale: 32.0
2024-09-22 14:55:20,926 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.11 vs. limit=15.0
2024-09-22 14:55:30,011 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.72 vs. limit=15.0
2024-09-22 14:55:35,725 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.431e+02 1.707e+02 1.931e+02 2.238e+02 4.906e+02, threshold=3.863e+02, percent-clipped=1.0
2024-09-22 14:55:40,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=39722.666666666664, ans=0.125
2024-09-22 14:56:32,766 INFO [train.py:1198] (1/4) Epoch 3, batch 750, loss[loss=0.3428, ctc_loss=0.2567, cr_loss=0.4307, over 17008.00 frames. ], tot_loss[loss=0.3461, ctc_loss=0.2616, cr_loss=0.4225, over 3261610.89 frames. ], batch size: 56, lr: 3.41e-02, grad_scale: 32.0
2024-09-22 14:56:56,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=39909.333333333336, ans=0.125
2024-09-22 14:56:58,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=39909.333333333336, ans=0.125
2024-09-22 14:57:01,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=39909.333333333336, ans=0.125
2024-09-22 14:57:30,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=40002.666666666664, ans=0.125
2024-09-22 14:57:35,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.66 vs. limit=15.0
2024-09-22 14:57:56,182 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.00 vs. limit=22.5
2024-09-22 14:57:57,149 INFO [train.py:1198] (1/4) Epoch 3, batch 800, loss[loss=0.3081, ctc_loss=0.2357, cr_loss=0.3624, over 16707.00 frames. ], tot_loss[loss=0.3462, ctc_loss=0.2616, cr_loss=0.4227, over 3280923.67 frames. ], batch size: 37, lr: 3.40e-02, grad_scale: 32.0
2024-09-22 14:58:20,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=40142.666666666664, ans=0.1
2024-09-22 14:58:25,181 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.263e+02 1.611e+02 1.885e+02 2.376e+02 4.057e+02, threshold=3.771e+02, percent-clipped=1.0
2024-09-22 14:58:31,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=40189.333333333336, ans=0.125
2024-09-22 14:59:18,681 INFO [train.py:1198] (1/4) Epoch 3, batch 850, loss[loss=0.3861, ctc_loss=0.3008, cr_loss=0.4263, over 15894.00 frames. ], tot_loss[loss=0.3449, ctc_loss=0.2605, cr_loss=0.4218, over 3301451.00 frames. ], batch size: 74, lr: 3.39e-02, grad_scale: 32.0
2024-09-22 15:00:38,408 INFO [train.py:1198] (1/4) Epoch 3, batch 900, loss[loss=0.432, ctc_loss=0.344, cr_loss=0.44, over 12444.00 frames. ], tot_loss[loss=0.344, ctc_loss=0.2598, cr_loss=0.4208, over 3303988.17 frames. ], batch size: 123, lr: 3.39e-02, grad_scale: 32.0
2024-09-22 15:01:06,321 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.646e+02 1.943e+02 2.321e+02 3.880e+02, threshold=3.887e+02, percent-clipped=1.0
2024-09-22 15:01:35,002 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.32 vs. limit=5.0
2024-09-22 15:01:35,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=40702.666666666664, ans=0.1
2024-09-22 15:01:41,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=40702.666666666664, ans=0.0
2024-09-22 15:02:03,456 INFO [train.py:1198] (1/4) Epoch 3, batch 950, loss[loss=0.2972, ctc_loss=0.2246, cr_loss=0.3631, over 17196.00 frames. ], tot_loss[loss=0.3415, ctc_loss=0.2576, cr_loss=0.4192, over 3318673.45 frames. ], batch size: 41, lr: 3.38e-02, grad_scale: 32.0
2024-09-22 15:02:19,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=40796.0, ans=0.0020008695652173926
2024-09-22 15:02:32,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=40842.666666666664, ans=0.125
2024-09-22 15:02:35,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.15 vs. limit=15.0
2024-09-22 15:02:41,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=40889.333333333336, ans=0.125
2024-09-22 15:03:21,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=40982.666666666664, ans=0.2
2024-09-22 15:03:28,684 INFO [train.py:1198] (1/4) Epoch 3, batch 1000, loss[loss=0.4529, ctc_loss=0.3663, cr_loss=0.4328, over 11739.00 frames. ], tot_loss[loss=0.3419, ctc_loss=0.2581, cr_loss=0.4191, over 3325930.95 frames. ], batch size: 123, lr: 3.38e-02, grad_scale: 32.0
2024-09-22 15:03:30,997 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.29 vs. limit=15.0
2024-09-22 15:03:38,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=41029.333333333336, ans=0.001950144927536231
2024-09-22 15:03:45,401 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.81 vs. limit=12.0
2024-09-22 15:03:48,705 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.20 vs. limit=22.5
2024-09-22 15:03:54,234 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 1.784e+02 2.139e+02 2.624e+02 4.654e+02, threshold=4.278e+02, percent-clipped=1.0
2024-09-22 15:03:54,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=41076.0, ans=0.125
2024-09-22 15:04:07,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=22.83 vs. limit=22.5
2024-09-22 15:04:35,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=41216.0, ans=0.05
2024-09-22 15:04:38,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=41216.0, ans=0.125
2024-09-22 15:04:39,196 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.56 vs. limit=15.0
2024-09-22 15:04:48,223 INFO [train.py:1198] (1/4) Epoch 3, batch 1050, loss[loss=0.3209, ctc_loss=0.2375, cr_loss=0.4167, over 17215.00 frames. ], tot_loss[loss=0.3424, ctc_loss=0.2585, cr_loss=0.4197, over 3325342.60 frames. ], batch size: 50, lr: 3.37e-02, grad_scale: 32.0
2024-09-22 15:05:10,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=41309.333333333336, ans=0.2
2024-09-22 15:05:58,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=41449.333333333336, ans=0.0
2024-09-22 15:06:10,764 INFO [train.py:1198] (1/4) Epoch 3, batch 1100, loss[loss=0.2736, ctc_loss=0.1979, cr_loss=0.3784, over 17091.00 frames. ], tot_loss[loss=0.3417, ctc_loss=0.2576, cr_loss=0.4204, over 3339955.32 frames. ], batch size: 40, lr: 3.37e-02, grad_scale: 32.0
2024-09-22 15:06:17,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=41496.0, ans=0.025
2024-09-22 15:06:20,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=41496.0, ans=0.125
2024-09-22 15:06:35,964 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 1.612e+02 1.917e+02 2.437e+02 4.278e+02, threshold=3.834e+02, percent-clipped=2.0
2024-09-22 15:06:42,562 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=2.92 vs. limit=15.0
2024-09-22 15:06:46,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=41589.333333333336, ans=0.025
2024-09-22 15:06:57,975 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.76 vs. limit=15.0
2024-09-22 15:07:06,189 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.56 vs. limit=22.5
2024-09-22 15:07:35,513 INFO [train.py:1198] (1/4) Epoch 3, batch 1150, loss[loss=0.3781, ctc_loss=0.2923, cr_loss=0.4292, over 17034.00 frames. ], tot_loss[loss=0.3405, ctc_loss=0.2567, cr_loss=0.4188, over 3328787.87 frames. ], batch size: 52, lr: 3.36e-02, grad_scale: 32.0
2024-09-22 15:07:42,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=41729.333333333336, ans=0.0
2024-09-22 15:07:43,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=41729.333333333336, ans=0.125
2024-09-22 15:07:55,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=41776.0, ans=0.0
2024-09-22 15:07:57,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=41776.0, ans=0.125
2024-09-22 15:08:01,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=41776.0, ans=0.125
2024-09-22 15:08:03,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=41776.0, ans=0.2
2024-09-22 15:08:10,212 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.87 vs. limit=15.0
2024-09-22 15:08:17,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=41822.666666666664, ans=0.125
2024-09-22 15:08:18,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=41822.666666666664, ans=0.125
2024-09-22 15:08:39,738 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:08:56,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=14.08 vs. limit=15.0
2024-09-22 15:08:57,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=41962.666666666664, ans=0.2
2024-09-22 15:08:58,267 INFO [train.py:1198] (1/4) Epoch 3, batch 1200, loss[loss=0.3705, ctc_loss=0.2803, cr_loss=0.4506, over 16709.00 frames. ], tot_loss[loss=0.3384, ctc_loss=0.2547, cr_loss=0.4182, over 3347860.84 frames. ], batch size: 61, lr: 3.36e-02, grad_scale: 32.0
2024-09-22 15:09:09,953 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=15.39 vs. limit=15.0
2024-09-22 15:09:23,540 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.644e+02 1.881e+02 2.304e+02 4.141e+02, threshold=3.762e+02, percent-clipped=3.0
2024-09-22 15:09:44,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=42102.666666666664, ans=0.125
2024-09-22 15:10:03,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=42149.333333333336, ans=0.0
2024-09-22 15:10:09,117 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=7.72 vs. limit=15.0
2024-09-22 15:10:12,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=42149.333333333336, ans=0.125
2024-09-22 15:10:17,333 INFO [train.py:1198] (1/4) Epoch 3, batch 1250, loss[loss=0.3426, ctc_loss=0.2601, cr_loss=0.4127, over 17273.00 frames. ], tot_loss[loss=0.341, ctc_loss=0.2572, cr_loss=0.4191, over 3337251.74 frames. ], batch size: 46, lr: 3.35e-02, grad_scale: 32.0
2024-09-22 15:10:26,381 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.98 vs. limit=22.5
2024-09-22 15:10:55,393 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.56 vs. limit=15.0
2024-09-22 15:11:07,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=42336.0, ans=0.125
2024-09-22 15:11:19,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=42336.0, ans=0.1
2024-09-22 15:11:31,001 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.81 vs. limit=12.0
2024-09-22 15:11:41,930 INFO [train.py:1198] (1/4) Epoch 3, batch 1300, loss[loss=0.3076, ctc_loss=0.2318, cr_loss=0.3788, over 17086.00 frames. ], tot_loss[loss=0.3394, ctc_loss=0.2555, cr_loss=0.4194, over 3348068.51 frames. ], batch size: 40, lr: 3.34e-02, grad_scale: 32.0
2024-09-22 15:11:45,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=42429.333333333336, ans=0.1
2024-09-22 15:12:02,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=42476.0, ans=0.125
2024-09-22 15:12:09,821 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 1.777e+02 1.981e+02 2.466e+02 4.544e+02, threshold=3.962e+02, percent-clipped=3.0
2024-09-22 15:12:10,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=42476.0, ans=0.125
2024-09-22 15:12:13,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=42476.0, ans=0.025
2024-09-22 15:12:18,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=42522.666666666664, ans=0.0
2024-09-22 15:12:18,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=42522.666666666664, ans=0.07
2024-09-22 15:12:19,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=42522.666666666664, ans=0.125
2024-09-22 15:12:34,750 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=11.93 vs. limit=15.0
2024-09-22 15:12:51,704 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.34 vs. limit=15.0
2024-09-22 15:13:06,078 INFO [train.py:1198] (1/4) Epoch 3, batch 1350, loss[loss=0.4505, ctc_loss=0.3622, cr_loss=0.4414, over 11558.00 frames. ], tot_loss[loss=0.3435, ctc_loss=0.2591, cr_loss=0.4222, over 3322616.58 frames. ], batch size: 123, lr: 3.34e-02, grad_scale: 32.0
2024-09-22 15:13:14,413 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:13:31,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=42709.333333333336, ans=0.125
2024-09-22 15:13:59,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=42802.666666666664, ans=0.0
2024-09-22 15:13:59,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=42802.666666666664, ans=0.0
2024-09-22 15:14:03,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=42802.666666666664, ans=0.0015646376811594212
2024-09-22 15:14:15,408 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.20 vs. limit=15.0
2024-09-22 15:14:25,893 INFO [train.py:1198] (1/4) Epoch 3, batch 1400, loss[loss=0.3614, ctc_loss=0.276, cr_loss=0.4273, over 17140.00 frames. ], tot_loss[loss=0.3402, ctc_loss=0.2564, cr_loss=0.4189, over 3335488.86 frames. ], batch size: 48, lr: 3.33e-02, grad_scale: 32.0
2024-09-22 15:14:26,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=42896.0, ans=0.125
2024-09-22 15:14:44,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=42942.666666666664, ans=0.125
2024-09-22 15:14:51,781 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 1.674e+02 1.940e+02 2.218e+02 3.917e+02, threshold=3.881e+02, percent-clipped=0.0
2024-09-22 15:15:00,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=42989.333333333336, ans=0.0015240579710144916
2024-09-22 15:15:06,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=42989.333333333336, ans=0.0015240579710144916
2024-09-22 15:15:09,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=42989.333333333336, ans=0.125
2024-09-22 15:15:10,123 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.61 vs. limit=6.0
2024-09-22 15:15:20,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=43036.0, ans=0.125
2024-09-22 15:15:26,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=43036.0, ans=0.125
2024-09-22 15:15:42,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=43082.666666666664, ans=0.125
2024-09-22 15:15:49,145 INFO [train.py:1198] (1/4) Epoch 3, batch 1450, loss[loss=0.3362, ctc_loss=0.2484, cr_loss=0.4392, over 17304.00 frames. ], tot_loss[loss=0.3403, ctc_loss=0.2563, cr_loss=0.42, over 3348836.33 frames. ], batch size: 49, lr: 3.33e-02, grad_scale: 32.0
2024-09-22 15:15:55,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=43129.333333333336, ans=0.125
2024-09-22 15:15:59,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=43129.333333333336, ans=0.125
2024-09-22 15:16:27,987 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:16:39,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=43269.333333333336, ans=0.1
2024-09-22 15:17:13,618 INFO [train.py:1198] (1/4) Epoch 3, batch 1500, loss[loss=0.3441, ctc_loss=0.2551, cr_loss=0.4453, over 17002.00 frames. ], tot_loss[loss=0.3388, ctc_loss=0.2552, cr_loss=0.4183, over 3352369.77 frames. ], batch size: 51, lr: 3.32e-02, grad_scale: 32.0
2024-09-22 15:17:16,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.65 vs. limit=10.0
2024-09-22 15:17:38,772 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 1.638e+02 2.002e+02 2.354e+02 3.823e+02, threshold=4.005e+02, percent-clipped=0.0
2024-09-22 15:18:02,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=43502.666666666664, ans=0.125
2024-09-22 15:18:02,709 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.18 vs. limit=15.0
2024-09-22 15:18:06,005 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=9.24 vs. limit=15.0
2024-09-22 15:18:22,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=43549.333333333336, ans=0.0
2024-09-22 15:18:35,371 INFO [train.py:1198] (1/4) Epoch 3, batch 1550, loss[loss=0.3607, ctc_loss=0.2672, cr_loss=0.4674, over 17220.00 frames. ], tot_loss[loss=0.3387, ctc_loss=0.255, cr_loss=0.4183, over 3357969.67 frames. ], batch size: 50, lr: 3.32e-02, grad_scale: 32.0
2024-09-22 15:18:43,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=43596.0, ans=0.025
2024-09-22 15:19:28,739 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=14.41 vs. limit=15.0
2024-09-22 15:19:36,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.39 vs. limit=15.0
2024-09-22 15:19:46,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=43782.666666666664, ans=0.07
2024-09-22 15:19:55,280 INFO [train.py:1198] (1/4) Epoch 3, batch 1600, loss[loss=0.3169, ctc_loss=0.2382, cr_loss=0.3932, over 17350.00 frames. ], tot_loss[loss=0.3397, ctc_loss=0.2556, cr_loss=0.4203, over 3355050.93 frames. ], batch size: 48, lr: 3.31e-02, grad_scale: 32.0
2024-09-22 15:20:15,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.80 vs. limit=15.0
2024-09-22 15:20:20,844 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 1.634e+02 1.960e+02 2.382e+02 4.201e+02, threshold=3.920e+02, percent-clipped=2.0
2024-09-22 15:20:27,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=43922.666666666664, ans=0.125
2024-09-22 15:20:48,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=43969.333333333336, ans=0.0
2024-09-22 15:21:13,623 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.48 vs. limit=6.0
2024-09-22 15:21:15,917 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:21:17,086 INFO [train.py:1198] (1/4) Epoch 3, batch 1650, loss[loss=0.3243, ctc_loss=0.2401, cr_loss=0.4209, over 17363.00 frames. ], tot_loss[loss=0.3405, ctc_loss=0.2564, cr_loss=0.4207, over 3346392.14 frames. ], batch size: 48, lr: 3.31e-02, grad_scale: 32.0
2024-09-22 15:21:22,289 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.79 vs. limit=15.0
2024-09-22 15:21:37,661 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.20 vs. limit=6.0
2024-09-22 15:21:38,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=44109.333333333336, ans=0.0
2024-09-22 15:21:50,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=44156.0, ans=0.125
2024-09-22 15:22:09,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=44202.666666666664, ans=0.1
2024-09-22 15:22:14,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=44202.666666666664, ans=0.0
2024-09-22 15:22:19,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=44202.666666666664, ans=0.125
2024-09-22 15:22:41,997 INFO [train.py:1198] (1/4) Epoch 3, batch 1700, loss[loss=0.3636, ctc_loss=0.2786, cr_loss=0.4252, over 16997.00 frames. ], tot_loss[loss=0.3397, ctc_loss=0.2556, cr_loss=0.4207, over 3355082.50 frames. ], batch size: 56, lr: 3.30e-02, grad_scale: 32.0
2024-09-22 15:22:47,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=44296.0, ans=0.0
2024-09-22 15:23:09,724 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.715e+02 2.359e+02 2.941e+02 4.631e+02, threshold=4.717e+02, percent-clipped=4.0
2024-09-22 15:23:49,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=44482.666666666664, ans=0.125
2024-09-22 15:23:57,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=44482.666666666664, ans=0.1
2024-09-22 15:24:03,723 INFO [train.py:1198] (1/4) Epoch 3, batch 1750, loss[loss=0.4298, ctc_loss=0.3298, cr_loss=0.4999, over 15271.00 frames. ], tot_loss[loss=0.3392, ctc_loss=0.2551, cr_loss=0.4202, over 3353497.42 frames. ], batch size: 89, lr: 3.30e-02, grad_scale: 32.0
2024-09-22 15:24:12,357 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.92 vs. limit=22.5
2024-09-22 15:24:29,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=44576.0, ans=0.1
2024-09-22 15:24:34,365 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=10.38 vs. limit=12.0
2024-09-22 15:24:56,928 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=8.00 vs. limit=15.0
2024-09-22 15:25:04,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=44669.333333333336, ans=0.125
2024-09-22 15:25:09,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.95 vs. limit=15.0
2024-09-22 15:25:25,623 INFO [train.py:1198] (1/4) Epoch 3, batch 1800, loss[loss=0.3368, ctc_loss=0.2555, cr_loss=0.4064, over 17206.00 frames. ], tot_loss[loss=0.3383, ctc_loss=0.2543, cr_loss=0.4199, over 3356182.75 frames. ], batch size: 41, lr: 3.29e-02, grad_scale: 64.0
2024-09-22 15:25:37,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=44762.666666666664, ans=0.0
2024-09-22 15:25:48,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=44809.333333333336, ans=0.1
2024-09-22 15:25:48,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=44809.333333333336, ans=0.125
2024-09-22 15:25:50,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=44809.333333333336, ans=0.2
2024-09-22 15:25:53,005 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 1.793e+02 2.251e+02 2.697e+02 4.483e+02, threshold=4.502e+02, percent-clipped=0.0
2024-09-22 15:26:02,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=44856.0, ans=0.125
2024-09-22 15:26:15,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=44902.666666666664, ans=0.04949747468305833
2024-09-22 15:26:30,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=44949.333333333336, ans=0.125
2024-09-22 15:26:47,445 INFO [train.py:1198] (1/4) Epoch 3, batch 1850, loss[loss=0.3158, ctc_loss=0.2428, cr_loss=0.3651, over 16952.00 frames. ], tot_loss[loss=0.3373, ctc_loss=0.2534, cr_loss=0.4196, over 3365494.04 frames. ], batch size: 42, lr: 3.29e-02, grad_scale: 32.0
2024-09-22 15:27:36,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=45136.0, ans=0.001057391304347826
2024-09-22 15:27:36,808 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=15.02 vs. limit=15.0
2024-09-22 15:27:42,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=45136.0, ans=0.125
2024-09-22 15:28:12,189 INFO [train.py:1198] (1/4) Epoch 3, batch 1900, loss[loss=0.3489, ctc_loss=0.2631, cr_loss=0.4292, over 17092.00 frames. ], tot_loss[loss=0.3374, ctc_loss=0.2535, cr_loss=0.4192, over 3357510.87 frames. ], batch size: 49, lr: 3.28e-02, grad_scale: 32.0
2024-09-22 15:28:23,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=45229.333333333336, ans=0.125
2024-09-22 15:28:25,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=45229.333333333336, ans=0.125
2024-09-22 15:28:25,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=45229.333333333336, ans=0.125
2024-09-22 15:28:38,955 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.740e+02 2.164e+02 2.812e+02 4.193e+02, threshold=4.328e+02, percent-clipped=0.0
2024-09-22 15:28:47,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=45322.666666666664, ans=0.125
2024-09-22 15:29:09,978 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=45369.333333333336, ans=0.0
2024-09-22 15:29:32,029 INFO [train.py:1198] (1/4) Epoch 3, batch 1950, loss[loss=0.3907, ctc_loss=0.2996, cr_loss=0.4555, over 15106.00 frames. ], tot_loss[loss=0.3385, ctc_loss=0.2545, cr_loss=0.4202, over 3350035.07 frames. ], batch size: 89, lr: 3.27e-02, grad_scale: 32.0
2024-09-22 15:29:53,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=45509.333333333336, ans=0.125
2024-09-22 15:30:02,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=45556.0, ans=0.00096608695652174
2024-09-22 15:30:05,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=45556.0, ans=0.125
2024-09-22 15:30:15,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=45556.0, ans=0.125
2024-09-22 15:30:51,629 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.09 vs. limit=15.0
2024-09-22 15:30:54,281 INFO [train.py:1198] (1/4) Epoch 3, batch 2000, loss[loss=0.3215, ctc_loss=0.2366, cr_loss=0.4247, over 17091.00 frames. ], tot_loss[loss=0.3372, ctc_loss=0.2534, cr_loss=0.4192, over 3357380.60 frames. ], batch size: 40, lr: 3.27e-02, grad_scale: 32.0
2024-09-22 15:31:08,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=45742.666666666664, ans=0.05
2024-09-22 15:31:23,956 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 1.768e+02 1.993e+02 2.472e+02 5.161e+02, threshold=3.986e+02, percent-clipped=2.0
2024-09-22 15:31:27,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=45789.333333333336, ans=0.125
2024-09-22 15:31:27,656 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:31:34,138 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:31:57,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=45836.0, ans=0.0009052173913043481
2024-09-22 15:32:19,442 INFO [train.py:1198] (1/4) Epoch 3, batch 2050, loss[loss=0.3066, ctc_loss=0.2292, cr_loss=0.3869, over 17063.00 frames. ], tot_loss[loss=0.3375, ctc_loss=0.2537, cr_loss=0.4193, over 3355575.34 frames. ], batch size: 46, lr: 3.26e-02, grad_scale: 32.0
2024-09-22 15:32:31,397 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.98 vs. limit=15.0
2024-09-22 15:33:08,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=46069.333333333336, ans=0.125
2024-09-22 15:33:09,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=46069.333333333336, ans=0.125
2024-09-22 15:33:41,356 INFO [train.py:1198] (1/4) Epoch 3, batch 2100, loss[loss=0.3898, ctc_loss=0.2944, cr_loss=0.4766, over 17203.00 frames. ], tot_loss[loss=0.3372, ctc_loss=0.2533, cr_loss=0.4195, over 3358654.97 frames. ], batch size: 55, lr: 3.26e-02, grad_scale: 32.0
2024-09-22 15:34:02,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=46209.333333333336, ans=0.125
2024-09-22 15:34:08,496 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.827e+02 2.117e+02 2.620e+02 4.403e+02, threshold=4.235e+02, percent-clipped=1.0
2024-09-22 15:34:27,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=46302.666666666664, ans=0.2
2024-09-22 15:34:54,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=46349.333333333336, ans=0.0
2024-09-22 15:35:01,055 INFO [train.py:1198] (1/4) Epoch 3, batch 2150, loss[loss=0.3968, ctc_loss=0.3053, cr_loss=0.4574, over 15152.00 frames. ], tot_loss[loss=0.3363, ctc_loss=0.2527, cr_loss=0.4181, over 3351252.89 frames. ], batch size: 89, lr: 3.25e-02, grad_scale: 32.0
2024-09-22 15:35:11,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=46396.0, ans=0.125
2024-09-22 15:35:11,556 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.53 vs. limit=15.0
2024-09-22 15:35:12,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=46396.0, ans=0.025
2024-09-22 15:35:12,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=46396.0, ans=0.1
2024-09-22 15:35:20,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=46442.666666666664, ans=0.125
2024-09-22 15:35:20,489 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.63 vs. limit=15.0
2024-09-22 15:35:26,160 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:35:39,724 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.56 vs. limit=15.0
2024-09-22 15:36:14,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=46582.666666666664, ans=0.125
2024-09-22 15:36:16,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.60 vs. limit=15.0
2024-09-22 15:36:22,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=46582.666666666664, ans=0.125
2024-09-22 15:36:25,581 INFO [train.py:1198] (1/4) Epoch 3, batch 2200, loss[loss=0.3552, ctc_loss=0.2698, cr_loss=0.4272, over 16586.00 frames. ], tot_loss[loss=0.3366, ctc_loss=0.2529, cr_loss=0.4183, over 3348118.09 frames. ], batch size: 66, lr: 3.25e-02, grad_scale: 32.0
2024-09-22 15:36:31,230 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.44 vs. limit=15.0
2024-09-22 15:36:55,287 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.636e+02 2.012e+02 2.486e+02 4.697e+02, threshold=4.025e+02, percent-clipped=4.0
2024-09-22 15:37:11,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=46722.666666666664, ans=0.125
2024-09-22 15:37:50,689 INFO [train.py:1198] (1/4) Epoch 3, batch 2250, loss[loss=0.3941, ctc_loss=0.3015, cr_loss=0.463, over 17009.00 frames. ], tot_loss[loss=0.3367, ctc_loss=0.253, cr_loss=0.4189, over 3354612.45 frames. ], batch size: 56, lr: 3.24e-02, grad_scale: 32.0
2024-09-22 15:38:09,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=46909.333333333336, ans=0.125
2024-09-22 15:38:27,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=46956.0, ans=0.2
2024-09-22 15:39:01,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=47049.333333333336, ans=0.0
2024-09-22 15:39:10,365 INFO [train.py:1198] (1/4) Epoch 3, batch 2300, loss[loss=0.2831, ctc_loss=0.2092, cr_loss=0.3695, over 17090.00 frames. ], tot_loss[loss=0.3352, ctc_loss=0.2515, cr_loss=0.4183, over 3360574.02 frames. ], batch size: 40, lr: 3.24e-02, grad_scale: 32.0
2024-09-22 15:39:17,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=47096.0, ans=0.2
2024-09-22 15:39:26,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=47142.666666666664, ans=0.0
2024-09-22 15:39:35,193 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.69 vs. limit=15.0
2024-09-22 15:39:37,770 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.712e+02 2.236e+02 2.750e+02 4.925e+02, threshold=4.473e+02, percent-clipped=7.0
2024-09-22 15:40:23,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=47282.666666666664, ans=0.025
2024-09-22 15:40:25,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=47282.666666666664, ans=0.0
2024-09-22 15:40:27,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=47282.666666666664, ans=0.04949747468305833
2024-09-22 15:40:33,161 INFO [train.py:1198] (1/4) Epoch 3, batch 2350, loss[loss=0.3691, ctc_loss=0.2853, cr_loss=0.4192, over 17305.00 frames. ], tot_loss[loss=0.3358, ctc_loss=0.2522, cr_loss=0.4182, over 3360702.74 frames. ], batch size: 51, lr: 3.23e-02, grad_scale: 32.0
2024-09-22 15:41:00,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=47376.0, ans=0.0005704347826086965
2024-09-22 15:41:05,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=47422.666666666664, ans=0.0005602898550724645
2024-09-22 15:41:33,899 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 15:41:58,589 INFO [train.py:1198] (1/4) Epoch 3, batch 2400, loss[loss=0.3236, ctc_loss=0.2392, cr_loss=0.422, over 17264.00 frames. ], tot_loss[loss=0.3364, ctc_loss=0.2527, cr_loss=0.4185, over 3351387.44 frames. ], batch size: 42, lr: 3.23e-02, grad_scale: 32.0
2024-09-22 15:42:25,878 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 1.787e+02 2.027e+02 2.376e+02 4.296e+02, threshold=4.054e+02, percent-clipped=0.0
2024-09-22 15:42:30,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=47609.333333333336, ans=0.2
2024-09-22 15:42:40,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=47656.0, ans=0.125
2024-09-22 15:42:41,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=47656.0, ans=0.125
2024-09-22 15:42:54,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=47702.666666666664, ans=0.125
2024-09-22 15:43:05,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=47749.333333333336, ans=0.5
2024-09-22 15:43:21,634 INFO [train.py:1198] (1/4) Epoch 3, batch 2450, loss[loss=0.2984, ctc_loss=0.2218, cr_loss=0.3833, over 17168.00 frames. ], tot_loss[loss=0.337, ctc_loss=0.2532, cr_loss=0.4192, over 3349830.13 frames. ], batch size: 41, lr: 3.22e-02, grad_scale: 32.0
2024-09-22 15:43:32,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=47796.0, ans=0.0
2024-09-22 15:43:32,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=47796.0, ans=0.1
2024-09-22 15:43:51,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=47889.333333333336, ans=0.0
2024-09-22 15:43:55,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=47889.333333333336, ans=0.00045884057971014476
2024-09-22 15:44:14,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=47936.0, ans=0.1
2024-09-22 15:44:27,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=47982.666666666664, ans=0.125
2024-09-22 15:44:38,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=47982.666666666664, ans=0.125
2024-09-22 15:44:41,202 INFO [train.py:1198] (1/4) Epoch 3, batch 2500, loss[loss=0.3047, ctc_loss=0.2289, cr_loss=0.379, over 17038.00 frames. ], tot_loss[loss=0.336, ctc_loss=0.2522, cr_loss=0.4191, over 3355511.64 frames. ], batch size: 39, lr: 3.22e-02, grad_scale: 32.0
2024-09-22 15:44:44,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=48029.333333333336, ans=0.0
2024-09-22 15:44:59,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.65 vs. limit=15.0
2024-09-22 15:45:08,320 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 1.703e+02 1.872e+02 2.246e+02 3.567e+02, threshold=3.744e+02, percent-clipped=0.0
2024-09-22 15:45:30,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=48169.333333333336, ans=0.125
2024-09-22 15:45:32,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.59 vs. limit=22.5
2024-09-22 15:45:46,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=48216.0, ans=0.00038782608695652095
2024-09-22 15:45:49,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=48216.0, ans=0.00038782608695652095
2024-09-22 15:46:03,938 INFO [train.py:1198] (1/4) Epoch 3, batch 2550, loss[loss=0.4506, ctc_loss=0.3604, cr_loss=0.4507, over 11868.00 frames. ], tot_loss[loss=0.3355, ctc_loss=0.2516, cr_loss=0.4191, over 3355078.16 frames. ], batch size: 123, lr: 3.21e-02, grad_scale: 32.0
2024-09-22 15:46:16,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=48262.666666666664, ans=0.125
2024-09-22 15:46:18,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=48262.666666666664, ans=0.1
2024-09-22 15:46:47,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=48356.0, ans=0.0
2024-09-22 15:47:00,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=48402.666666666664, ans=0.1
2024-09-22 15:47:05,982 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=7.54 vs. limit=15.0
2024-09-22 15:47:28,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.13 vs. limit=22.5
2024-09-22 15:47:31,517 INFO [train.py:1198] (1/4) Epoch 3, batch 2600, loss[loss=0.2996, ctc_loss=0.2279, cr_loss=0.3583, over 16993.00 frames. ], tot_loss[loss=0.3344, ctc_loss=0.2508, cr_loss=0.4182, over 3352866.36 frames. ], batch size: 44, lr: 3.21e-02, grad_scale: 32.0
2024-09-22 15:47:38,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=48496.0, ans=0.5
2024-09-22 15:47:44,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=48496.0, ans=0.125
2024-09-22 15:47:58,749 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 1.709e+02 1.993e+02 2.394e+02 4.094e+02, threshold=3.987e+02, percent-clipped=1.0
2024-09-22 15:48:01,003 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=14.11 vs. limit=15.0
2024-09-22 15:48:13,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=48589.333333333336, ans=0.0
2024-09-22 15:48:42,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=48682.666666666664, ans=0.125
2024-09-22 15:48:51,832 INFO [train.py:1198] (1/4) Epoch 3, batch 2650, loss[loss=0.3482, ctc_loss=0.2684, cr_loss=0.3989, over 16066.00 frames. ], tot_loss[loss=0.3351, ctc_loss=0.2514, cr_loss=0.4185, over 3349124.79 frames. ], batch size: 74, lr: 3.20e-02, grad_scale: 32.0
2024-09-22 15:49:01,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=48729.333333333336, ans=0.0
2024-09-22 15:49:17,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=48776.0, ans=0.1
2024-09-22 15:49:21,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=48776.0, ans=0.125
2024-09-22 15:49:24,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=48822.666666666664, ans=0.125
2024-09-22 15:49:58,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=48916.0, ans=0.05
2024-09-22 15:50:14,727 INFO [train.py:1198] (1/4) Epoch 3, batch 2700, loss[loss=0.4349, ctc_loss=0.3478, cr_loss=0.4355, over 11552.00 frames. ], tot_loss[loss=0.3338, ctc_loss=0.2501, cr_loss=0.4183, over 3357220.08 frames. ], batch size: 123, lr: 3.20e-02, grad_scale: 32.0
2024-09-22 15:50:27,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=48962.666666666664, ans=0.2
2024-09-22 15:50:29,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=49009.333333333336, ans=0.00021536231884057913
2024-09-22 15:50:37,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=49009.333333333336, ans=0.125
2024-09-22 15:50:40,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=49009.333333333336, ans=0.125
2024-09-22 15:50:41,576 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.782e+02 2.096e+02 2.443e+02 4.661e+02, threshold=4.192e+02, percent-clipped=1.0
2024-09-22 15:51:19,166 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.64 vs. limit=15.0
2024-09-22 15:51:24,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=49149.333333333336, ans=0.0
2024-09-22 15:51:39,723 INFO [train.py:1198] (1/4) Epoch 3, batch 2750, loss[loss=0.2869, ctc_loss=0.2082, cr_loss=0.3937, over 17067.00 frames. ], tot_loss[loss=0.3317, ctc_loss=0.2484, cr_loss=0.4164, over 3360171.26 frames. ], batch size: 46, lr: 3.19e-02, grad_scale: 32.0
2024-09-22 15:51:40,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=49196.0, ans=0.125
2024-09-22 15:51:43,671 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.74 vs. limit=15.0
2024-09-22 15:52:04,240 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=12.60 vs. limit=15.0
2024-09-22 15:52:05,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=49242.666666666664, ans=0.125
2024-09-22 15:52:30,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=49336.0, ans=0.0
2024-09-22 15:52:32,822 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=7.93 vs. limit=15.0
2024-09-22 15:52:32,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2.whitening_limit, batch_count=49336.0, ans=15.0
2024-09-22 15:52:40,910 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.33 vs. limit=22.5
2024-09-22 15:53:01,836 INFO [train.py:1198] (1/4) Epoch 3, batch 2800, loss[loss=0.368, ctc_loss=0.2809, cr_loss=0.4353, over 16594.00 frames. ], tot_loss[loss=0.3331, ctc_loss=0.2497, cr_loss=0.4166, over 3356361.40 frames. ], batch size: 66, lr: 3.19e-02, grad_scale: 32.0
2024-09-22 15:53:07,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=28.23 vs. limit=22.5
2024-09-22 15:53:13,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=49429.333333333336, ans=0.125
2024-09-22 15:53:16,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=49476.0, ans=0.125
2024-09-22 15:53:29,476 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 1.759e+02 2.001e+02 2.340e+02 4.757e+02, threshold=4.003e+02, percent-clipped=1.0
2024-09-22 15:53:36,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=49522.666666666664, ans=0.025
2024-09-22 15:53:36,426 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=15.53 vs. limit=15.0
2024-09-22 15:53:38,060 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.36 vs. limit=10.0
2024-09-22 15:53:47,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=49522.666666666664, ans=0.125
2024-09-22 15:53:55,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=49569.333333333336, ans=0.125
2024-09-22 15:54:08,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=49616.0, ans=0.125
2024-09-22 15:54:12,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=49616.0, ans=0.125
2024-09-22 15:54:22,390 INFO [train.py:1198] (1/4) Epoch 3, batch 2850, loss[loss=0.3282, ctc_loss=0.2442, cr_loss=0.4202, over 16697.00 frames. ], tot_loss[loss=0.3335, ctc_loss=0.25, cr_loss=0.4171, over 3349220.58 frames. ], batch size: 61, lr: 3.18e-02, grad_scale: 32.0
2024-09-22 15:54:22,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=49662.666666666664, ans=0.0
2024-09-22 15:54:38,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=49709.333333333336, ans=0.1
2024-09-22 15:54:40,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=49709.333333333336, ans=0.125
2024-09-22 15:54:50,607 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.71 vs. limit=15.0
2024-09-22 15:54:51,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=49709.333333333336, ans=0.125
2024-09-22 15:55:26,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten.whitening_limit, batch_count=49802.666666666664, ans=22.5
2024-09-22 15:55:41,143 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=9.01 vs. limit=10.0
2024-09-22 15:55:45,085 INFO [train.py:1198] (1/4) Epoch 3, batch 2900, loss[loss=0.3653, ctc_loss=0.2789, cr_loss=0.4319, over 16731.00 frames. ], tot_loss[loss=0.3325, ctc_loss=0.2493, cr_loss=0.416, over 3353013.49 frames. ], batch size: 61, lr: 3.18e-02, grad_scale: 32.0
2024-09-22 15:55:51,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=49896.0, ans=0.0
2024-09-22 15:56:14,387 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 1.745e+02 2.067e+02 2.607e+02 4.355e+02, threshold=4.133e+02, percent-clipped=1.0
2024-09-22 15:56:16,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=49942.666666666664, ans=0.125
2024-09-22 15:56:41,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten.whitening_limit, batch_count=50036.0, ans=22.5
2024-09-22 15:56:50,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=50036.0, ans=0.125
2024-09-22 15:57:01,012 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=22.17 vs. limit=22.5
2024-09-22 15:57:09,736 INFO [train.py:1198] (1/4) Epoch 3, batch 2950, loss[loss=0.36, ctc_loss=0.2767, cr_loss=0.4163, over 14836.00 frames. ], tot_loss[loss=0.3324, ctc_loss=0.2493, cr_loss=0.4158, over 3347222.94 frames. ], batch size: 89, lr: 3.17e-02, grad_scale: 32.0
2024-09-22 15:57:16,529 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.74 vs. limit=15.0
2024-09-22 15:57:29,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=50176.0, ans=0.025
2024-09-22 15:57:48,143 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.47 vs. limit=15.0
2024-09-22 15:58:25,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=50316.0, ans=0.125
2024-09-22 15:58:31,265 INFO [train.py:1198] (1/4) Epoch 3, batch 3000, loss[loss=0.3776, ctc_loss=0.294, cr_loss=0.418, over 15109.00 frames. ], tot_loss[loss=0.3333, ctc_loss=0.2498, cr_loss=0.4172, over 3352696.01 frames. ], batch size: 89, lr: 3.17e-02, grad_scale: 32.0
2024-09-22 15:58:31,265 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 15:58:46,491 INFO [train.py:1230] (1/4) Epoch 3, validation: loss=0.08436, ctc_loss=0.08436, cr_loss=7.957e-15, over 944034.00 frames. 
2024-09-22 15:58:46,492 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 15:58:58,108 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.31 vs. limit=22.5
2024-09-22 15:59:12,854 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 1.636e+02 1.924e+02 2.171e+02 3.615e+02, threshold=3.848e+02, percent-clipped=0.0
2024-09-22 15:59:16,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=50456.0, ans=0.2
2024-09-22 15:59:30,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=50456.0, ans=0.125
2024-09-22 15:59:47,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=50549.333333333336, ans=0.0
2024-09-22 16:00:03,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=50596.0, ans=0.015
2024-09-22 16:00:04,589 INFO [train.py:1198] (1/4) Epoch 3, batch 3050, loss[loss=0.3211, ctc_loss=0.2429, cr_loss=0.3911, over 16348.00 frames. ], tot_loss[loss=0.3332, ctc_loss=0.2496, cr_loss=0.4179, over 3353476.73 frames. ], batch size: 36, lr: 3.16e-02, grad_scale: 32.0
2024-09-22 16:00:06,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=50596.0, ans=0.125
2024-09-22 16:00:08,626 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=23.94 vs. limit=22.5
2024-09-22 16:00:15,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=50596.0, ans=0.1
2024-09-22 16:00:17,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=50596.0, ans=0.0
2024-09-22 16:00:23,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=50642.666666666664, ans=0.125
2024-09-22 16:00:53,227 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 16:01:05,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=50782.666666666664, ans=0.0
2024-09-22 16:01:15,289 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.26 vs. limit=15.0
2024-09-22 16:01:22,485 INFO [train.py:1198] (1/4) Epoch 3, batch 3100, loss[loss=0.363, ctc_loss=0.275, cr_loss=0.4398, over 17019.00 frames. ], tot_loss[loss=0.3334, ctc_loss=0.2496, cr_loss=0.4187, over 3357989.99 frames. ], batch size: 51, lr: 3.16e-02, grad_scale: 32.0
2024-09-22 16:01:22,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=50829.333333333336, ans=0.0
2024-09-22 16:01:25,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=50829.333333333336, ans=0.125
2024-09-22 16:01:32,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=50829.333333333336, ans=0.125
2024-09-22 16:01:38,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=50876.0, ans=0.125
2024-09-22 16:01:44,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=50876.0, ans=0.125
2024-09-22 16:01:49,066 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.630e+02 1.881e+02 2.229e+02 3.253e+02, threshold=3.762e+02, percent-clipped=0.0
2024-09-22 16:01:57,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=50922.666666666664, ans=0.0
2024-09-22 16:02:04,799 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.55 vs. limit=15.0
2024-09-22 16:02:11,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=50969.333333333336, ans=0.125
2024-09-22 16:02:43,156 INFO [train.py:1198] (1/4) Epoch 3, batch 3150, loss[loss=0.3636, ctc_loss=0.2769, cr_loss=0.4331, over 16613.00 frames. ], tot_loss[loss=0.3338, ctc_loss=0.2499, cr_loss=0.4197, over 3361054.30 frames. ], batch size: 66, lr: 3.15e-02, grad_scale: 32.0
2024-09-22 16:03:06,359 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.25 vs. limit=6.0
2024-09-22 16:03:17,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=51156.0, ans=0.125
2024-09-22 16:03:34,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=51202.666666666664, ans=0.0
2024-09-22 16:03:40,028 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=7.36 vs. limit=15.0
2024-09-22 16:04:00,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=51249.333333333336, ans=0.0
2024-09-22 16:04:03,495 INFO [train.py:1198] (1/4) Epoch 3, batch 3200, loss[loss=0.3648, ctc_loss=0.2724, cr_loss=0.4619, over 17025.00 frames. ], tot_loss[loss=0.3327, ctc_loss=0.249, cr_loss=0.419, over 3361705.76 frames. ], batch size: 51, lr: 3.15e-02, grad_scale: 32.0
2024-09-22 16:04:25,040 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.85 vs. limit=15.0
2024-09-22 16:04:27,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=51342.666666666664, ans=0.04949747468305833
2024-09-22 16:04:30,052 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.638e+02 1.865e+02 2.186e+02 5.181e+02, threshold=3.729e+02, percent-clipped=1.0
2024-09-22 16:04:36,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=51389.333333333336, ans=0.09899494936611666
2024-09-22 16:04:53,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=51436.0, ans=0.0
2024-09-22 16:04:56,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=51436.0, ans=0.0
2024-09-22 16:04:59,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=51436.0, ans=0.025
2024-09-22 16:05:02,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=51436.0, ans=0.125
2024-09-22 16:05:23,508 INFO [train.py:1198] (1/4) Epoch 3, batch 3250, loss[loss=0.3592, ctc_loss=0.2657, cr_loss=0.4677, over 17294.00 frames. ], tot_loss[loss=0.3324, ctc_loss=0.2485, cr_loss=0.4195, over 3366015.89 frames. ], batch size: 51, lr: 3.14e-02, grad_scale: 32.0
2024-09-22 16:05:27,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=51529.333333333336, ans=0.2
2024-09-22 16:05:38,249 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=22.45 vs. limit=22.5
2024-09-22 16:05:51,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=51576.0, ans=0.0
2024-09-22 16:06:04,234 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 16:06:06,104 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.26 vs. limit=15.0
2024-09-22 16:06:43,309 INFO [train.py:1198] (1/4) Epoch 3, batch 3300, loss[loss=0.3674, ctc_loss=0.2779, cr_loss=0.4477, over 17235.00 frames. ], tot_loss[loss=0.3329, ctc_loss=0.2489, cr_loss=0.4203, over 3365567.23 frames. ], batch size: 50, lr: 3.14e-02, grad_scale: 32.0
2024-09-22 16:06:45,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=51762.666666666664, ans=0.04949747468305833
2024-09-22 16:06:49,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=51762.666666666664, ans=0.2
2024-09-22 16:06:52,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=51762.666666666664, ans=0.125
2024-09-22 16:07:02,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=51809.333333333336, ans=0.2
2024-09-22 16:07:09,973 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.647e+02 1.953e+02 2.348e+02 4.155e+02, threshold=3.905e+02, percent-clipped=1.0
2024-09-22 16:08:01,372 INFO [train.py:1198] (1/4) Epoch 3, batch 3350, loss[loss=0.3556, ctc_loss=0.2669, cr_loss=0.4438, over 17050.00 frames. ], tot_loss[loss=0.334, ctc_loss=0.25, cr_loss=0.4199, over 3350435.17 frames. ], batch size: 52, lr: 3.13e-02, grad_scale: 32.0
2024-09-22 16:08:04,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=51996.0, ans=0.0
2024-09-22 16:08:12,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=51996.0, ans=0.0
2024-09-22 16:08:18,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=52042.666666666664, ans=0.125
2024-09-22 16:08:24,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=52042.666666666664, ans=0.0
2024-09-22 16:08:27,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=52042.666666666664, ans=0.1
2024-09-22 16:09:06,541 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.94 vs. limit=10.0
2024-09-22 16:09:10,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=52182.666666666664, ans=0.0
2024-09-22 16:09:19,277 INFO [train.py:1198] (1/4) Epoch 3, batch 3400, loss[loss=0.2746, ctc_loss=0.202, cr_loss=0.3631, over 16949.00 frames. ], tot_loss[loss=0.3324, ctc_loss=0.2486, cr_loss=0.4194, over 3366697.04 frames. ], batch size: 42, lr: 3.13e-02, grad_scale: 32.0
2024-09-22 16:09:22,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=52229.333333333336, ans=0.0
2024-09-22 16:09:24,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=52229.333333333336, ans=0.125
2024-09-22 16:09:33,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=52276.0, ans=0.07
2024-09-22 16:09:45,958 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.677e+02 1.939e+02 2.464e+02 4.534e+02, threshold=3.878e+02, percent-clipped=3.0
2024-09-22 16:09:50,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=52322.666666666664, ans=0.09899494936611666
2024-09-22 16:10:22,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=52416.0, ans=0.025
2024-09-22 16:10:37,705 INFO [train.py:1198] (1/4) Epoch 3, batch 3450, loss[loss=0.3627, ctc_loss=0.2733, cr_loss=0.447, over 17295.00 frames. ], tot_loss[loss=0.3322, ctc_loss=0.2484, cr_loss=0.4186, over 3351945.42 frames. ], batch size: 51, lr: 3.12e-02, grad_scale: 32.0
2024-09-22 16:11:01,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=52509.333333333336, ans=0.0
2024-09-22 16:11:29,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=52602.666666666664, ans=0.125
2024-09-22 16:11:32,118 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.53 vs. limit=8.0
2024-09-22 16:11:38,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=52649.333333333336, ans=0.125
2024-09-22 16:11:57,412 INFO [train.py:1198] (1/4) Epoch 3, batch 3500, loss[loss=0.3171, ctc_loss=0.2386, cr_loss=0.3927, over 17273.00 frames. ], tot_loss[loss=0.3326, ctc_loss=0.2488, cr_loss=0.4189, over 3351788.00 frames. ], batch size: 44, lr: 3.12e-02, grad_scale: 32.0
2024-09-22 16:11:57,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=52696.0, ans=0.125
2024-09-22 16:12:08,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=52696.0, ans=0.025
2024-09-22 16:12:24,122 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 1.808e+02 2.110e+02 2.675e+02 4.151e+02, threshold=4.220e+02, percent-clipped=2.0
2024-09-22 16:12:36,939 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 16:13:15,557 INFO [train.py:1198] (1/4) Epoch 3, batch 3550, loss[loss=0.3023, ctc_loss=0.2181, cr_loss=0.421, over 17245.00 frames. ], tot_loss[loss=0.3325, ctc_loss=0.2486, cr_loss=0.4191, over 3353825.15 frames. ], batch size: 44, lr: 3.11e-02, grad_scale: 32.0
2024-09-22 16:13:36,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=52976.0, ans=0.125
2024-09-22 16:13:42,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=52976.0, ans=0.125
2024-09-22 16:13:51,120 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.93 vs. limit=15.0
2024-09-22 16:13:58,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=53022.666666666664, ans=0.125
2024-09-22 16:14:05,051 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.31 vs. limit=15.0
2024-09-22 16:14:06,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.39 vs. limit=15.0
2024-09-22 16:14:23,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=53116.0, ans=0.0
2024-09-22 16:14:31,208 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 16:14:35,690 INFO [train.py:1198] (1/4) Epoch 3, batch 3600, loss[loss=0.3006, ctc_loss=0.2205, cr_loss=0.4003, over 17259.00 frames. ], tot_loss[loss=0.3303, ctc_loss=0.2469, cr_loss=0.417, over 3361820.20 frames. ], batch size: 44, lr: 3.11e-02, grad_scale: 32.0
2024-09-22 16:15:01,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=53209.333333333336, ans=0.0
2024-09-22 16:15:04,297 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.804e+02 2.185e+02 2.634e+02 3.942e+02, threshold=4.371e+02, percent-clipped=0.0
2024-09-22 16:15:09,629 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.76 vs. limit=15.0
2024-09-22 16:15:30,318 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.49 vs. limit=22.5
2024-09-22 16:15:35,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=53302.666666666664, ans=15.0
2024-09-22 16:15:57,743 INFO [train.py:1198] (1/4) Epoch 3, batch 3650, loss[loss=0.375, ctc_loss=0.2837, cr_loss=0.4564, over 15061.00 frames. ], tot_loss[loss=0.3298, ctc_loss=0.2467, cr_loss=0.4158, over 3348633.19 frames. ], batch size: 89, lr: 3.10e-02, grad_scale: 32.0
2024-09-22 16:16:31,312 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.52 vs. limit=22.5
2024-09-22 16:16:48,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=53536.0, ans=0.05
2024-09-22 16:16:55,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=53536.0, ans=0.1
2024-09-22 16:16:57,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=53536.0, ans=0.1
2024-09-22 16:17:12,086 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=21.55 vs. limit=22.5
2024-09-22 16:17:16,109 INFO [train.py:1198] (1/4) Epoch 3, batch 3700, loss[loss=0.3433, ctc_loss=0.2581, cr_loss=0.4263, over 17359.00 frames. ], tot_loss[loss=0.33, ctc_loss=0.2469, cr_loss=0.4158, over 3347764.98 frames. ], batch size: 48, lr: 3.10e-02, grad_scale: 32.0
2024-09-22 16:17:16,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=53629.333333333336, ans=0.1
2024-09-22 16:17:31,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=53676.0, ans=0.0
2024-09-22 16:17:42,218 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.642e+02 1.878e+02 2.249e+02 4.018e+02, threshold=3.757e+02, percent-clipped=0.0
2024-09-22 16:18:02,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=53769.333333333336, ans=0.95
2024-09-22 16:18:15,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=53769.333333333336, ans=0.025
2024-09-22 16:18:26,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.78 vs. limit=10.0
2024-09-22 16:18:34,047 INFO [train.py:1198] (1/4) Epoch 3, batch 3750, loss[loss=0.3196, ctc_loss=0.2376, cr_loss=0.4097, over 16952.00 frames. ], tot_loss[loss=0.3313, ctc_loss=0.2479, cr_loss=0.4166, over 3330147.28 frames. ], batch size: 42, lr: 3.10e-02, grad_scale: 32.0
2024-09-22 16:18:54,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=53909.333333333336, ans=0.125
2024-09-22 16:19:06,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=53956.0, ans=0.0
2024-09-22 16:19:14,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=53956.0, ans=0.025
2024-09-22 16:19:45,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=54049.333333333336, ans=0.125
2024-09-22 16:19:51,719 INFO [train.py:1198] (1/4) Epoch 3, batch 3800, loss[loss=0.3151, ctc_loss=0.2351, cr_loss=0.3997, over 16209.00 frames. ], tot_loss[loss=0.3327, ctc_loss=0.2491, cr_loss=0.4177, over 3318201.57 frames. ], batch size: 36, lr: 3.09e-02, grad_scale: 32.0
2024-09-22 16:20:08,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=54142.666666666664, ans=0.125
2024-09-22 16:20:17,980 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 1.642e+02 1.883e+02 2.367e+02 4.025e+02, threshold=3.766e+02, percent-clipped=5.0
2024-09-22 16:20:36,531 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.61 vs. limit=22.5
2024-09-22 16:21:03,695 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=10.23 vs. limit=15.0
2024-09-22 16:21:10,249 INFO [train.py:1198] (1/4) Epoch 3, batch 3850, loss[loss=0.3502, ctc_loss=0.2608, cr_loss=0.4467, over 17006.00 frames. ], tot_loss[loss=0.3346, ctc_loss=0.2507, cr_loss=0.4192, over 3300268.76 frames. ], batch size: 51, lr: 3.09e-02, grad_scale: 64.0
2024-09-22 16:21:16,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=54329.333333333336, ans=0.125
2024-09-22 16:21:19,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=54329.333333333336, ans=0.07
2024-09-22 16:21:21,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=54329.333333333336, ans=0.1
2024-09-22 16:21:21,596 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.10 vs. limit=10.0
2024-09-22 16:21:35,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=54376.0, ans=0.125
2024-09-22 16:23:12,167 INFO [train.py:1198] (1/4) Epoch 4, batch 0, loss[loss=0.3338, ctc_loss=0.2522, cr_loss=0.408, over 17061.00 frames. ], tot_loss[loss=0.3338, ctc_loss=0.2522, cr_loss=0.408, over 17061.00 frames. ], batch size: 46, lr: 2.88e-02, grad_scale: 32.0
2024-09-22 16:23:12,167 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 16:23:27,758 INFO [train.py:1230] (1/4) Epoch 4, validation: loss=0.08466, ctc_loss=0.08466, cr_loss=9.003e-15, over 944034.00 frames. 
2024-09-22 16:23:27,759 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 16:24:04,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=54637.333333333336, ans=0.125
2024-09-22 16:24:06,065 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 1.944e+02 2.324e+02 2.751e+02 6.786e+02, threshold=4.649e+02, percent-clipped=3.0
2024-09-22 16:24:21,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=54684.0, ans=0.025
2024-09-22 16:24:50,274 INFO [train.py:1198] (1/4) Epoch 4, batch 50, loss[loss=0.3354, ctc_loss=0.2524, cr_loss=0.4148, over 15943.00 frames. ], tot_loss[loss=0.3253, ctc_loss=0.2425, cr_loss=0.414, over 760804.40 frames. ], batch size: 35, lr: 2.88e-02, grad_scale: 32.0
2024-09-22 16:24:58,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=54777.333333333336, ans=0.035
2024-09-22 16:25:11,947 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.79 vs. limit=12.0
2024-09-22 16:25:22,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=54870.666666666664, ans=0.125
2024-09-22 16:25:43,233 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.68 vs. limit=15.0
2024-09-22 16:25:47,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=54917.333333333336, ans=0.2
2024-09-22 16:25:48,334 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=15.67 vs. limit=22.5
2024-09-22 16:25:49,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=54917.333333333336, ans=0.0
2024-09-22 16:25:52,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=54964.0, ans=0.125
2024-09-22 16:25:55,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=54964.0, ans=0.025
2024-09-22 16:25:59,088 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.52 vs. limit=10.0
2024-09-22 16:26:12,371 INFO [train.py:1198] (1/4) Epoch 4, batch 100, loss[loss=0.3328, ctc_loss=0.2476, cr_loss=0.4262, over 17198.00 frames. ], tot_loss[loss=0.3229, ctc_loss=0.2401, cr_loss=0.414, over 1342877.42 frames. ], batch size: 47, lr: 2.87e-02, grad_scale: 32.0
2024-09-22 16:26:23,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=55010.666666666664, ans=0.1
2024-09-22 16:26:44,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=55057.333333333336, ans=0.05
2024-09-22 16:26:50,367 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 1.670e+02 1.866e+02 2.190e+02 3.249e+02, threshold=3.731e+02, percent-clipped=0.0
2024-09-22 16:26:50,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=55104.0, ans=0.0
2024-09-22 16:27:35,224 INFO [train.py:1198] (1/4) Epoch 4, batch 150, loss[loss=0.2637, ctc_loss=0.188, cr_loss=0.3785, over 16958.00 frames. ], tot_loss[loss=0.3212, ctc_loss=0.2386, cr_loss=0.413, over 1788916.02 frames. ], batch size: 42, lr: 2.87e-02, grad_scale: 32.0
2024-09-22 16:27:49,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=55290.666666666664, ans=0.0
2024-09-22 16:27:54,875 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.64 vs. limit=15.0
2024-09-22 16:27:57,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=55290.666666666664, ans=0.1
2024-09-22 16:28:09,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=55337.333333333336, ans=0.125
2024-09-22 16:28:13,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=55337.333333333336, ans=0.09899494936611666
2024-09-22 16:28:23,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=55384.0, ans=0.0
2024-09-22 16:28:58,935 INFO [train.py:1198] (1/4) Epoch 4, batch 200, loss[loss=0.3081, ctc_loss=0.2277, cr_loss=0.402, over 17054.00 frames. ], tot_loss[loss=0.3198, ctc_loss=0.2374, cr_loss=0.4119, over 2144803.94 frames. ], batch size: 46, lr: 2.86e-02, grad_scale: 32.0
2024-09-22 16:29:33,618 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 1.559e+02 1.682e+02 1.957e+02 2.989e+02, threshold=3.363e+02, percent-clipped=0.0
2024-09-22 16:29:37,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=55570.666666666664, ans=0.0
2024-09-22 16:30:14,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=55664.0, ans=0.125
2024-09-22 16:30:17,536 INFO [train.py:1198] (1/4) Epoch 4, batch 250, loss[loss=0.3409, ctc_loss=0.2581, cr_loss=0.4142, over 17024.00 frames. ], tot_loss[loss=0.3217, ctc_loss=0.239, cr_loss=0.4137, over 2409220.21 frames. ], batch size: 44, lr: 2.86e-02, grad_scale: 32.0
2024-09-22 16:31:05,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=55804.0, ans=0.025
2024-09-22 16:31:19,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=55850.666666666664, ans=0.125
2024-09-22 16:31:33,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=55897.333333333336, ans=0.09899494936611666
2024-09-22 16:31:43,124 INFO [train.py:1198] (1/4) Epoch 4, batch 300, loss[loss=0.3678, ctc_loss=0.2806, cr_loss=0.436, over 15828.00 frames. ], tot_loss[loss=0.3232, ctc_loss=0.2402, cr_loss=0.4152, over 2618030.10 frames. ], batch size: 74, lr: 2.86e-02, grad_scale: 32.0
2024-09-22 16:31:44,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=55944.0, ans=0.125
2024-09-22 16:31:45,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=55944.0, ans=0.125
2024-09-22 16:32:20,152 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.659e+02 1.864e+02 2.226e+02 3.223e+02, threshold=3.728e+02, percent-clipped=0.0
2024-09-22 16:33:07,320 INFO [train.py:1198] (1/4) Epoch 4, batch 350, loss[loss=0.2699, ctc_loss=0.2003, cr_loss=0.348, over 17188.00 frames. ], tot_loss[loss=0.3224, ctc_loss=0.2395, cr_loss=0.4143, over 2780500.45 frames. ], batch size: 41, lr: 2.85e-02, grad_scale: 32.0
2024-09-22 16:33:12,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=56177.333333333336, ans=0.0
2024-09-22 16:33:12,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=56177.333333333336, ans=0.0
2024-09-22 16:33:34,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=56224.0, ans=0.025
2024-09-22 16:33:45,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=56270.666666666664, ans=0.125
2024-09-22 16:34:15,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=56364.0, ans=0.125
2024-09-22 16:34:23,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=56364.0, ans=0.1
2024-09-22 16:34:29,681 INFO [train.py:1198] (1/4) Epoch 4, batch 400, loss[loss=0.2604, ctc_loss=0.1908, cr_loss=0.348, over 16634.00 frames. ], tot_loss[loss=0.3224, ctc_loss=0.2395, cr_loss=0.4145, over 2905181.93 frames. ], batch size: 37, lr: 2.85e-02, grad_scale: 32.0
2024-09-22 16:34:41,349 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 16:34:46,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=56457.333333333336, ans=0.1
2024-09-22 16:34:47,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=56457.333333333336, ans=0.125
2024-09-22 16:35:03,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=56504.0, ans=0.1
2024-09-22 16:35:04,961 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.655e+02 1.851e+02 2.343e+02 4.879e+02, threshold=3.703e+02, percent-clipped=2.0
2024-09-22 16:35:27,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=56550.666666666664, ans=0.125
2024-09-22 16:35:38,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=56597.333333333336, ans=0.1
2024-09-22 16:35:52,501 INFO [train.py:1198] (1/4) Epoch 4, batch 450, loss[loss=0.3551, ctc_loss=0.2653, cr_loss=0.4491, over 16920.00 frames. ], tot_loss[loss=0.3216, ctc_loss=0.2388, cr_loss=0.4135, over 3006771.61 frames. ], batch size: 58, lr: 2.84e-02, grad_scale: 32.0
2024-09-22 16:35:57,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=56644.0, ans=0.1
2024-09-22 16:35:58,402 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.76 vs. limit=6.0
2024-09-22 16:36:22,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=56690.666666666664, ans=0.0
2024-09-22 16:36:38,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=56737.333333333336, ans=0.125
2024-09-22 16:36:41,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=56784.0, ans=0.04949747468305833
2024-09-22 16:37:09,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=56830.666666666664, ans=0.125
2024-09-22 16:37:14,898 INFO [train.py:1198] (1/4) Epoch 4, batch 500, loss[loss=0.3686, ctc_loss=0.2832, cr_loss=0.4274, over 16257.00 frames. ], tot_loss[loss=0.3218, ctc_loss=0.2391, cr_loss=0.4136, over 3082683.16 frames. ], batch size: 74, lr: 2.84e-02, grad_scale: 32.0
2024-09-22 16:37:28,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=56877.333333333336, ans=0.0
2024-09-22 16:37:39,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=56924.0, ans=0.125
2024-09-22 16:37:53,150 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 1.629e+02 1.949e+02 2.165e+02 3.477e+02, threshold=3.897e+02, percent-clipped=0.0
2024-09-22 16:37:53,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.min_abs, batch_count=56970.666666666664, ans=0.5
2024-09-22 16:37:53,990 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=21.34 vs. limit=22.5
2024-09-22 16:38:09,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=57017.333333333336, ans=0.2
2024-09-22 16:38:28,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=57064.0, ans=0.1
2024-09-22 16:38:30,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=57064.0, ans=0.125
2024-09-22 16:38:39,921 INFO [train.py:1198] (1/4) Epoch 4, batch 550, loss[loss=0.3625, ctc_loss=0.2676, cr_loss=0.4743, over 17208.00 frames. ], tot_loss[loss=0.3214, ctc_loss=0.2387, cr_loss=0.4139, over 3144410.44 frames. ], batch size: 55, lr: 2.83e-02, grad_scale: 32.0
2024-09-22 16:38:49,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=57110.666666666664, ans=0.125
2024-09-22 16:39:02,810 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=5.21 vs. limit=12.0
2024-09-22 16:39:37,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=57250.666666666664, ans=0.1
2024-09-22 16:39:47,383 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.58 vs. limit=15.0
2024-09-22 16:39:59,406 INFO [train.py:1198] (1/4) Epoch 4, batch 600, loss[loss=0.3424, ctc_loss=0.2552, cr_loss=0.4363, over 17004.00 frames. ], tot_loss[loss=0.322, ctc_loss=0.2391, cr_loss=0.4147, over 3196995.84 frames. ], batch size: 53, lr: 2.83e-02, grad_scale: 32.0
2024-09-22 16:40:31,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=57437.333333333336, ans=0.125
2024-09-22 16:40:34,388 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 1.589e+02 1.770e+02 2.208e+02 4.389e+02, threshold=3.540e+02, percent-clipped=1.0
2024-09-22 16:40:50,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=57484.0, ans=0.125
2024-09-22 16:40:58,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.min_abs, batch_count=57484.0, ans=0.5
2024-09-22 16:41:19,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=57530.666666666664, ans=0.0
2024-09-22 16:41:24,122 INFO [train.py:1198] (1/4) Epoch 4, batch 650, loss[loss=0.2982, ctc_loss=0.2176, cr_loss=0.403, over 17076.00 frames. ], tot_loss[loss=0.3237, ctc_loss=0.2405, cr_loss=0.416, over 3220994.62 frames. ], batch size: 46, lr: 2.83e-02, grad_scale: 32.0
2024-09-22 16:41:32,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=57577.333333333336, ans=0.125
2024-09-22 16:41:52,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=57624.0, ans=0.05
2024-09-22 16:42:03,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=57670.666666666664, ans=0.025
2024-09-22 16:42:29,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.03 vs. limit=15.0
2024-09-22 16:42:32,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=57764.0, ans=0.05
2024-09-22 16:42:42,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=57764.0, ans=0.125
2024-09-22 16:42:45,631 INFO [train.py:1198] (1/4) Epoch 4, batch 700, loss[loss=0.3199, ctc_loss=0.2378, cr_loss=0.4105, over 16984.00 frames. ], tot_loss[loss=0.3234, ctc_loss=0.2404, cr_loss=0.4152, over 3249665.33 frames. ], batch size: 56, lr: 2.82e-02, grad_scale: 32.0
2024-09-22 16:42:59,342 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=12.00 vs. limit=15.0
2024-09-22 16:43:23,507 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 1.612e+02 1.882e+02 2.294e+02 3.695e+02, threshold=3.764e+02, percent-clipped=3.0
2024-09-22 16:43:43,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=57950.666666666664, ans=0.125
2024-09-22 16:43:49,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=57950.666666666664, ans=0.0
2024-09-22 16:43:51,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=57997.333333333336, ans=0.0
2024-09-22 16:43:57,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=57997.333333333336, ans=0.125
2024-09-22 16:44:08,399 INFO [train.py:1198] (1/4) Epoch 4, batch 750, loss[loss=0.2998, ctc_loss=0.2222, cr_loss=0.3881, over 17300.00 frames. ], tot_loss[loss=0.3219, ctc_loss=0.2389, cr_loss=0.4148, over 3280037.04 frames. ], batch size: 51, lr: 2.82e-02, grad_scale: 32.0
2024-09-22 16:44:08,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=58044.0, ans=0.025
2024-09-22 16:44:18,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=58044.0, ans=0.1
2024-09-22 16:44:18,424 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.96 vs. limit=15.0
2024-09-22 16:44:22,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=58090.666666666664, ans=0.2
2024-09-22 16:44:51,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=58137.333333333336, ans=0.025
2024-09-22 16:44:51,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=58137.333333333336, ans=0.125
2024-09-22 16:44:51,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=58137.333333333336, ans=0.125
2024-09-22 16:44:52,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=58137.333333333336, ans=0.125
2024-09-22 16:45:02,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=58184.0, ans=0.0
2024-09-22 16:45:27,577 INFO [train.py:1198] (1/4) Epoch 4, batch 800, loss[loss=0.3589, ctc_loss=0.271, cr_loss=0.4394, over 16039.00 frames. ], tot_loss[loss=0.3208, ctc_loss=0.2379, cr_loss=0.4143, over 3304899.40 frames. ], batch size: 74, lr: 2.81e-02, grad_scale: 32.0
2024-09-22 16:45:43,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.27 vs. limit=22.5
2024-09-22 16:46:07,412 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 1.727e+02 1.869e+02 2.216e+02 3.268e+02, threshold=3.738e+02, percent-clipped=0.0
2024-09-22 16:46:23,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=58417.333333333336, ans=0.125
2024-09-22 16:46:26,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=58417.333333333336, ans=0.025
2024-09-22 16:46:49,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=58464.0, ans=0.1
2024-09-22 16:46:51,992 INFO [train.py:1198] (1/4) Epoch 4, batch 850, loss[loss=0.2844, ctc_loss=0.2051, cr_loss=0.3963, over 17030.00 frames. ], tot_loss[loss=0.3208, ctc_loss=0.2379, cr_loss=0.4142, over 3320081.84 frames. ], batch size: 39, lr: 2.81e-02, grad_scale: 32.0
2024-09-22 16:46:56,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=58510.666666666664, ans=0.125
2024-09-22 16:47:16,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=58557.333333333336, ans=0.0
2024-09-22 16:47:48,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=58650.666666666664, ans=0.0
2024-09-22 16:48:01,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=58697.333333333336, ans=0.0
2024-09-22 16:48:16,702 INFO [train.py:1198] (1/4) Epoch 4, batch 900, loss[loss=0.3404, ctc_loss=0.2516, cr_loss=0.444, over 16493.00 frames. ], tot_loss[loss=0.3198, ctc_loss=0.2371, cr_loss=0.4138, over 3330617.70 frames. ], batch size: 66, lr: 2.81e-02, grad_scale: 32.0
2024-09-22 16:48:32,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=58790.666666666664, ans=0.0
2024-09-22 16:48:36,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=58790.666666666664, ans=0.125
2024-09-22 16:48:38,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=58790.666666666664, ans=0.07
2024-09-22 16:48:52,230 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.664e+02 1.989e+02 2.596e+02 4.339e+02, threshold=3.979e+02, percent-clipped=2.0
2024-09-22 16:49:23,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=58930.666666666664, ans=0.0
2024-09-22 16:49:26,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=58930.666666666664, ans=0.0
2024-09-22 16:49:36,961 INFO [train.py:1198] (1/4) Epoch 4, batch 950, loss[loss=0.3389, ctc_loss=0.2504, cr_loss=0.4425, over 17066.00 frames. ], tot_loss[loss=0.3203, ctc_loss=0.2375, cr_loss=0.4141, over 3319679.83 frames. ], batch size: 46, lr: 2.80e-02, grad_scale: 32.0
2024-09-22 16:49:37,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=58977.333333333336, ans=0.0
2024-09-22 16:49:49,102 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.81 vs. limit=10.0
2024-09-22 16:49:49,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=58977.333333333336, ans=0.2
2024-09-22 16:49:59,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=59024.0, ans=0.025
2024-09-22 16:50:21,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.78 vs. limit=22.5
2024-09-22 16:50:29,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=59117.333333333336, ans=0.125
2024-09-22 16:51:01,450 INFO [train.py:1198] (1/4) Epoch 4, batch 1000, loss[loss=0.346, ctc_loss=0.2617, cr_loss=0.4219, over 16592.00 frames. ], tot_loss[loss=0.3215, ctc_loss=0.2385, cr_loss=0.4149, over 3317712.19 frames. ], batch size: 66, lr: 2.80e-02, grad_scale: 32.0
2024-09-22 16:51:25,777 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.82 vs. limit=15.0
2024-09-22 16:51:36,100 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.574e+02 1.735e+02 2.105e+02 3.870e+02, threshold=3.470e+02, percent-clipped=0.0
2024-09-22 16:51:39,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=59304.0, ans=0.0
2024-09-22 16:51:53,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=59350.666666666664, ans=0.2
2024-09-22 16:52:11,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=59397.333333333336, ans=0.125
2024-09-22 16:52:20,427 INFO [train.py:1198] (1/4) Epoch 4, batch 1050, loss[loss=0.3557, ctc_loss=0.273, cr_loss=0.4138, over 16736.00 frames. ], tot_loss[loss=0.3217, ctc_loss=0.2387, cr_loss=0.4151, over 3330875.80 frames. ], batch size: 61, lr: 2.79e-02, grad_scale: 32.0
2024-09-22 16:52:22,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=59444.0, ans=0.0
2024-09-22 16:52:23,860 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 16:53:09,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=59584.0, ans=0.125
2024-09-22 16:53:23,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.max_positive, batch_count=59584.0, ans=0.95
2024-09-22 16:53:44,831 INFO [train.py:1198] (1/4) Epoch 4, batch 1100, loss[loss=0.4017, ctc_loss=0.3192, cr_loss=0.4123, over 11216.00 frames. ], tot_loss[loss=0.323, ctc_loss=0.2398, cr_loss=0.4162, over 3323097.91 frames. ], batch size: 123, lr: 2.79e-02, grad_scale: 32.0
2024-09-22 16:53:49,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=59677.333333333336, ans=0.125
2024-09-22 16:54:20,014 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.292e+02 1.576e+02 1.850e+02 2.274e+02 3.544e+02, threshold=3.699e+02, percent-clipped=1.0
2024-09-22 16:54:26,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=59770.666666666664, ans=0.025
2024-09-22 16:54:55,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=59864.0, ans=0.04949747468305833
2024-09-22 16:55:04,528 INFO [train.py:1198] (1/4) Epoch 4, batch 1150, loss[loss=0.2933, ctc_loss=0.2167, cr_loss=0.3832, over 17273.00 frames. ], tot_loss[loss=0.3225, ctc_loss=0.2392, cr_loss=0.4163, over 3333040.58 frames. ], batch size: 42, lr: 2.78e-02, grad_scale: 32.0
2024-09-22 16:55:04,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=59910.666666666664, ans=0.05
2024-09-22 16:55:11,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=59910.666666666664, ans=0.125
2024-09-22 16:55:18,112 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=9.12 vs. limit=15.0
2024-09-22 16:55:19,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=59957.333333333336, ans=0.125
2024-09-22 16:55:28,856 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-22 16:55:42,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=60004.0, ans=0.125
2024-09-22 16:56:13,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=60097.333333333336, ans=0.125
2024-09-22 16:56:26,317 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.02 vs. limit=22.5
2024-09-22 16:56:28,658 INFO [train.py:1198] (1/4) Epoch 4, batch 1200, loss[loss=0.291, ctc_loss=0.2143, cr_loss=0.3832, over 17239.00 frames. ], tot_loss[loss=0.3201, ctc_loss=0.2372, cr_loss=0.4147, over 3347560.53 frames. ], batch size: 42, lr: 2.78e-02, grad_scale: 32.0
2024-09-22 16:57:01,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.08 vs. limit=6.0
2024-09-22 16:57:03,612 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 1.578e+02 1.757e+02 2.030e+02 3.618e+02, threshold=3.514e+02, percent-clipped=0.0
2024-09-22 16:57:04,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=60237.333333333336, ans=0.0
2024-09-22 16:57:19,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=60284.0, ans=0.0
2024-09-22 16:57:51,021 INFO [train.py:1198] (1/4) Epoch 4, batch 1250, loss[loss=0.3275, ctc_loss=0.2417, cr_loss=0.4289, over 17033.00 frames. ], tot_loss[loss=0.3206, ctc_loss=0.2375, cr_loss=0.4155, over 3357261.71 frames. ], batch size: 52, lr: 2.78e-02, grad_scale: 32.0
2024-09-22 16:57:56,204 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.49 vs. limit=10.0
2024-09-22 16:58:01,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.27 vs. limit=15.0
2024-09-22 16:58:12,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=60424.0, ans=0.025
2024-09-22 16:59:00,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=60564.0, ans=0.2
2024-09-22 16:59:12,919 INFO [train.py:1198] (1/4) Epoch 4, batch 1300, loss[loss=0.3082, ctc_loss=0.2252, cr_loss=0.415, over 17158.00 frames. ], tot_loss[loss=0.32, ctc_loss=0.237, cr_loss=0.4152, over 3363284.69 frames. ], batch size: 41, lr: 2.77e-02, grad_scale: 32.0
2024-09-22 16:59:42,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=60657.333333333336, ans=0.125
2024-09-22 16:59:43,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=60704.0, ans=0.0
2024-09-22 16:59:43,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=60704.0, ans=0.0
2024-09-22 16:59:47,774 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.72 vs. limit=12.0
2024-09-22 16:59:48,140 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.594e+02 1.780e+02 2.076e+02 4.160e+02, threshold=3.560e+02, percent-clipped=3.0
2024-09-22 16:59:48,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.60 vs. limit=15.0
2024-09-22 17:00:31,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=60844.0, ans=0.1
2024-09-22 17:00:32,385 INFO [train.py:1198] (1/4) Epoch 4, batch 1350, loss[loss=0.2888, ctc_loss=0.2119, cr_loss=0.3848, over 17152.00 frames. ], tot_loss[loss=0.3198, ctc_loss=0.2368, cr_loss=0.4149, over 3370850.07 frames. ], batch size: 45, lr: 2.77e-02, grad_scale: 32.0
2024-09-22 17:00:35,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=60844.0, ans=0.0
2024-09-22 17:00:35,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=60844.0, ans=0.025
2024-09-22 17:00:48,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=60844.0, ans=0.125
2024-09-22 17:01:04,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=60890.666666666664, ans=0.2
2024-09-22 17:01:13,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=60937.333333333336, ans=0.2
2024-09-22 17:01:30,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=60984.0, ans=0.2
2024-09-22 17:01:30,742 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.44 vs. limit=15.0
2024-09-22 17:01:57,167 INFO [train.py:1198] (1/4) Epoch 4, batch 1400, loss[loss=0.3232, ctc_loss=0.2389, cr_loss=0.4214, over 17234.00 frames. ], tot_loss[loss=0.3184, ctc_loss=0.2357, cr_loss=0.4135, over 3360870.09 frames. ], batch size: 50, lr: 2.76e-02, grad_scale: 32.0
2024-09-22 17:02:19,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=61124.0, ans=0.1
2024-09-22 17:02:24,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=61124.0, ans=0.2
2024-09-22 17:02:34,720 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 1.619e+02 1.850e+02 2.266e+02 3.949e+02, threshold=3.701e+02, percent-clipped=2.0
2024-09-22 17:03:16,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.15 vs. limit=15.0
2024-09-22 17:03:21,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=61310.666666666664, ans=0.0
2024-09-22 17:03:22,329 INFO [train.py:1198] (1/4) Epoch 4, batch 1450, loss[loss=0.2579, ctc_loss=0.1917, cr_loss=0.3309, over 17052.00 frames. ], tot_loss[loss=0.3171, ctc_loss=0.2345, cr_loss=0.4129, over 3364398.73 frames. ], batch size: 39, lr: 2.76e-02, grad_scale: 32.0
2024-09-22 17:03:35,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=61310.666666666664, ans=0.125
2024-09-22 17:03:36,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=61357.333333333336, ans=0.125
2024-09-22 17:03:41,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=61357.333333333336, ans=0.0
2024-09-22 17:03:46,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=61357.333333333336, ans=0.0
2024-09-22 17:04:06,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=61404.0, ans=0.1
2024-09-22 17:04:18,587 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.64 vs. limit=12.0
2024-09-22 17:04:37,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=61497.333333333336, ans=0.125
2024-09-22 17:04:41,897 INFO [train.py:1198] (1/4) Epoch 4, batch 1500, loss[loss=0.3162, ctc_loss=0.2349, cr_loss=0.4069, over 17296.00 frames. ], tot_loss[loss=0.3179, ctc_loss=0.2351, cr_loss=0.4139, over 3357239.86 frames. ], batch size: 46, lr: 2.76e-02, grad_scale: 32.0
2024-09-22 17:04:45,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=61544.0, ans=0.125
2024-09-22 17:04:55,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=7.21 vs. limit=15.0
2024-09-22 17:05:17,065 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.547e+02 1.744e+02 2.028e+02 3.491e+02, threshold=3.489e+02, percent-clipped=0.0
2024-09-22 17:05:23,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=61637.333333333336, ans=0.2
2024-09-22 17:05:33,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=61684.0, ans=0.125
2024-09-22 17:05:37,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=61684.0, ans=0.125
2024-09-22 17:06:02,092 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.30 vs. limit=15.0
2024-09-22 17:06:06,037 INFO [train.py:1198] (1/4) Epoch 4, batch 1550, loss[loss=0.3312, ctc_loss=0.2495, cr_loss=0.4086, over 17358.00 frames. ], tot_loss[loss=0.318, ctc_loss=0.2353, cr_loss=0.4137, over 3359779.61 frames. ], batch size: 48, lr: 2.75e-02, grad_scale: 32.0
2024-09-22 17:06:22,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=61824.0, ans=0.125
2024-09-22 17:06:50,045 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.69 vs. limit=5.0
2024-09-22 17:06:52,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=61917.333333333336, ans=0.035
2024-09-22 17:07:03,959 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.65 vs. limit=12.0
2024-09-22 17:07:26,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=62010.666666666664, ans=0.05
2024-09-22 17:07:27,834 INFO [train.py:1198] (1/4) Epoch 4, batch 1600, loss[loss=0.2849, ctc_loss=0.2058, cr_loss=0.3956, over 17353.00 frames. ], tot_loss[loss=0.3192, ctc_loss=0.2361, cr_loss=0.4156, over 3364742.37 frames. ], batch size: 48, lr: 2.75e-02, grad_scale: 32.0
2024-09-22 17:07:28,648 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=15.24 vs. limit=15.0
2024-09-22 17:07:37,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=62010.666666666664, ans=0.0
2024-09-22 17:08:05,555 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.652e+02 1.885e+02 2.249e+02 4.170e+02, threshold=3.770e+02, percent-clipped=2.0
2024-09-22 17:08:15,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=62104.0, ans=0.09899494936611666
2024-09-22 17:08:31,861 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.33 vs. limit=15.0
2024-09-22 17:08:48,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=62244.0, ans=0.125
2024-09-22 17:08:50,127 INFO [train.py:1198] (1/4) Epoch 4, batch 1650, loss[loss=0.2952, ctc_loss=0.2169, cr_loss=0.3913, over 17313.00 frames. ], tot_loss[loss=0.3192, ctc_loss=0.2361, cr_loss=0.4155, over 3356678.63 frames. ], batch size: 46, lr: 2.75e-02, grad_scale: 32.0
2024-09-22 17:09:20,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=62337.333333333336, ans=0.125
2024-09-22 17:09:25,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=62337.333333333336, ans=0.035
2024-09-22 17:09:53,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=62430.666666666664, ans=0.125
2024-09-22 17:09:55,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=62430.666666666664, ans=0.2
2024-09-22 17:10:09,473 INFO [train.py:1198] (1/4) Epoch 4, batch 1700, loss[loss=0.3146, ctc_loss=0.2358, cr_loss=0.3943, over 17225.00 frames. ], tot_loss[loss=0.3199, ctc_loss=0.2367, cr_loss=0.4157, over 3357716.57 frames. ], batch size: 50, lr: 2.74e-02, grad_scale: 32.0
2024-09-22 17:10:20,093 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.55 vs. limit=5.0
2024-09-22 17:10:22,505 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.58 vs. limit=22.5
2024-09-22 17:10:27,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=62524.0, ans=0.125
2024-09-22 17:10:40,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=62524.0, ans=0.125
2024-09-22 17:10:49,656 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.564e+02 1.856e+02 2.208e+02 3.257e+02, threshold=3.711e+02, percent-clipped=0.0
2024-09-22 17:11:01,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=62617.333333333336, ans=0.125
2024-09-22 17:11:04,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=62617.333333333336, ans=0.025
2024-09-22 17:11:34,080 INFO [train.py:1198] (1/4) Epoch 4, batch 1750, loss[loss=0.2785, ctc_loss=0.2031, cr_loss=0.3767, over 17067.00 frames. ], tot_loss[loss=0.3177, ctc_loss=0.2351, cr_loss=0.4133, over 3356321.39 frames. ], batch size: 46, lr: 2.74e-02, grad_scale: 32.0
2024-09-22 17:11:46,018 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=23.10 vs. limit=22.5
2024-09-22 17:11:51,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=62757.333333333336, ans=0.125
2024-09-22 17:11:58,922 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.26 vs. limit=15.0
2024-09-22 17:12:15,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=62804.0, ans=0.0
2024-09-22 17:12:19,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=62804.0, ans=0.025
2024-09-22 17:12:31,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=62850.666666666664, ans=0.125
2024-09-22 17:12:40,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=62897.333333333336, ans=0.125
2024-09-22 17:12:43,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=62897.333333333336, ans=0.1
2024-09-22 17:12:58,852 INFO [train.py:1198] (1/4) Epoch 4, batch 1800, loss[loss=0.2725, ctc_loss=0.2054, cr_loss=0.3356, over 17103.00 frames. ], tot_loss[loss=0.3164, ctc_loss=0.234, cr_loss=0.4119, over 3351229.02 frames. ], batch size: 40, lr: 2.73e-02, grad_scale: 32.0
2024-09-22 17:13:13,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=62990.666666666664, ans=0.125
2024-09-22 17:13:24,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=62990.666666666664, ans=0.125
2024-09-22 17:13:33,467 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.277e+02 1.628e+02 1.997e+02 2.584e+02 3.622e+02, threshold=3.995e+02, percent-clipped=0.0
2024-09-22 17:13:40,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=63037.333333333336, ans=0.0
2024-09-22 17:13:52,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=63084.0, ans=0.0
2024-09-22 17:14:16,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=63177.333333333336, ans=0.025
2024-09-22 17:14:17,201 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.11 vs. limit=15.0
2024-09-22 17:14:17,813 INFO [train.py:1198] (1/4) Epoch 4, batch 1850, loss[loss=0.2947, ctc_loss=0.2186, cr_loss=0.3806, over 17255.00 frames. ], tot_loss[loss=0.3151, ctc_loss=0.233, cr_loss=0.4104, over 3357490.19 frames. ], batch size: 44, lr: 2.73e-02, grad_scale: 32.0
2024-09-22 17:14:37,637 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.66 vs. limit=15.0
2024-09-22 17:14:44,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=63224.0, ans=0.0
2024-09-22 17:14:49,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=63270.666666666664, ans=0.0
2024-09-22 17:14:49,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=63270.666666666664, ans=0.125
2024-09-22 17:14:57,821 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.68 vs. limit=15.0
2024-09-22 17:14:58,153 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.85 vs. limit=15.0
2024-09-22 17:15:05,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=63317.333333333336, ans=0.1
2024-09-22 17:15:41,411 INFO [train.py:1198] (1/4) Epoch 4, batch 1900, loss[loss=0.3655, ctc_loss=0.273, cr_loss=0.4628, over 16798.00 frames. ], tot_loss[loss=0.315, ctc_loss=0.233, cr_loss=0.4099, over 3364646.16 frames. ], batch size: 61, lr: 2.73e-02, grad_scale: 32.0
2024-09-22 17:15:41,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=63410.666666666664, ans=0.0
2024-09-22 17:15:46,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=63410.666666666664, ans=0.0
2024-09-22 17:16:14,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.44 vs. limit=15.0
2024-09-22 17:16:16,406 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.606e+02 1.832e+02 2.161e+02 3.717e+02, threshold=3.664e+02, percent-clipped=0.0
2024-09-22 17:16:16,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=63504.0, ans=0.1
2024-09-22 17:16:26,742 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.33 vs. limit=10.0
2024-09-22 17:16:27,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=63550.666666666664, ans=0.125
2024-09-22 17:16:50,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=63597.333333333336, ans=0.2
2024-09-22 17:16:59,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=63644.0, ans=0.1
2024-09-22 17:17:01,343 INFO [train.py:1198] (1/4) Epoch 4, batch 1950, loss[loss=0.2451, ctc_loss=0.1757, cr_loss=0.3471, over 16947.00 frames. ], tot_loss[loss=0.3149, ctc_loss=0.2329, cr_loss=0.4102, over 3361072.91 frames. ], batch size: 42, lr: 2.72e-02, grad_scale: 32.0
2024-09-22 17:17:01,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=63644.0, ans=0.125
2024-09-22 17:17:20,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=63690.666666666664, ans=0.1
2024-09-22 17:17:21,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=63690.666666666664, ans=0.125
2024-09-22 17:17:21,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=63690.666666666664, ans=0.0
2024-09-22 17:17:26,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=63690.666666666664, ans=0.05
2024-09-22 17:17:39,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=63737.333333333336, ans=0.125
2024-09-22 17:18:11,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=63830.666666666664, ans=0.125
2024-09-22 17:18:15,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=63830.666666666664, ans=0.1
2024-09-22 17:18:21,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=63830.666666666664, ans=0.0
2024-09-22 17:18:26,131 INFO [train.py:1198] (1/4) Epoch 4, batch 2000, loss[loss=0.3046, ctc_loss=0.2266, cr_loss=0.3901, over 17231.00 frames. ], tot_loss[loss=0.3157, ctc_loss=0.2334, cr_loss=0.4118, over 3363648.73 frames. ], batch size: 50, lr: 2.72e-02, grad_scale: 64.0
2024-09-22 17:18:37,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=63877.333333333336, ans=0.125
2024-09-22 17:19:01,203 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.569e+02 1.789e+02 2.368e+02 3.802e+02, threshold=3.577e+02, percent-clipped=1.0
2024-09-22 17:19:09,554 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:19:14,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=64017.333333333336, ans=0.1
2024-09-22 17:19:17,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=64017.333333333336, ans=0.2
2024-09-22 17:19:20,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=64017.333333333336, ans=0.1
2024-09-22 17:19:28,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=64064.0, ans=0.0
2024-09-22 17:19:32,243 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=21.77 vs. limit=22.5
2024-09-22 17:19:36,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=64064.0, ans=0.125
2024-09-22 17:19:45,617 INFO [train.py:1198] (1/4) Epoch 4, batch 2050, loss[loss=0.3548, ctc_loss=0.265, cr_loss=0.4491, over 17288.00 frames. ], tot_loss[loss=0.3155, ctc_loss=0.2332, cr_loss=0.4117, over 3366900.56 frames. ], batch size: 49, lr: 2.71e-02, grad_scale: 64.0
2024-09-22 17:19:54,718 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.02 vs. limit=22.5
2024-09-22 17:19:59,085 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.45 vs. limit=10.0
2024-09-22 17:20:09,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=64157.333333333336, ans=0.1
2024-09-22 17:20:19,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=64204.0, ans=0.1
2024-09-22 17:20:31,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=64204.0, ans=0.025
2024-09-22 17:20:40,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=64250.666666666664, ans=0.0
2024-09-22 17:20:47,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=64250.666666666664, ans=0.0
2024-09-22 17:20:47,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=64250.666666666664, ans=0.0
2024-09-22 17:20:51,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=64297.333333333336, ans=0.125
2024-09-22 17:20:58,966 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.38 vs. limit=6.0
2024-09-22 17:21:01,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=64297.333333333336, ans=0.125
2024-09-22 17:21:07,689 INFO [train.py:1198] (1/4) Epoch 4, batch 2100, loss[loss=0.257, ctc_loss=0.1898, cr_loss=0.3362, over 16670.00 frames. ], tot_loss[loss=0.3153, ctc_loss=0.2332, cr_loss=0.4107, over 3367502.12 frames. ], batch size: 37, lr: 2.71e-02, grad_scale: 32.0
2024-09-22 17:21:09,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=64344.0, ans=0.0
2024-09-22 17:21:20,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=64344.0, ans=0.125
2024-09-22 17:21:37,500 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.46 vs. limit=15.0
2024-09-22 17:21:44,600 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 1.633e+02 1.973e+02 2.304e+02 3.408e+02, threshold=3.946e+02, percent-clipped=0.0
2024-09-22 17:22:00,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=64484.0, ans=0.125
2024-09-22 17:22:29,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=64577.333333333336, ans=0.125
2024-09-22 17:22:33,062 INFO [train.py:1198] (1/4) Epoch 4, batch 2150, loss[loss=0.2763, ctc_loss=0.2022, cr_loss=0.3702, over 17293.00 frames. ], tot_loss[loss=0.3144, ctc_loss=0.2323, cr_loss=0.4109, over 3370260.30 frames. ], batch size: 46, lr: 2.71e-02, grad_scale: 32.0
2024-09-22 17:23:12,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=64670.666666666664, ans=0.125
2024-09-22 17:23:15,329 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=10.44 vs. limit=15.0
2024-09-22 17:23:16,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=64670.666666666664, ans=0.125
2024-09-22 17:23:28,013 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.75 vs. limit=12.0
2024-09-22 17:23:52,466 INFO [train.py:1198] (1/4) Epoch 4, batch 2200, loss[loss=0.3285, ctc_loss=0.2457, cr_loss=0.4138, over 17222.00 frames. ], tot_loss[loss=0.315, ctc_loss=0.2328, cr_loss=0.4115, over 3370218.35 frames. ], batch size: 50, lr: 2.70e-02, grad_scale: 32.0
2024-09-22 17:23:54,922 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=7.56 vs. limit=15.0
2024-09-22 17:23:55,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=64810.666666666664, ans=0.0
2024-09-22 17:24:26,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=64904.0, ans=0.125
2024-09-22 17:24:29,089 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.691e+02 2.009e+02 2.410e+02 3.639e+02, threshold=4.017e+02, percent-clipped=0.0
2024-09-22 17:24:29,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=64904.0, ans=0.125
2024-09-22 17:24:53,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=64950.666666666664, ans=0.125
2024-09-22 17:25:11,834 INFO [train.py:1198] (1/4) Epoch 4, batch 2250, loss[loss=0.3653, ctc_loss=0.2865, cr_loss=0.3938, over 12017.00 frames. ], tot_loss[loss=0.3164, ctc_loss=0.2338, cr_loss=0.413, over 3355501.05 frames. ], batch size: 123, lr: 2.70e-02, grad_scale: 32.0
2024-09-22 17:25:32,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.08 vs. limit=12.0
2024-09-22 17:25:46,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=65137.333333333336, ans=0.0
2024-09-22 17:26:07,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.03 vs. limit=15.0
2024-09-22 17:26:16,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=65230.666666666664, ans=0.125
2024-09-22 17:26:33,874 INFO [train.py:1198] (1/4) Epoch 4, batch 2300, loss[loss=0.319, ctc_loss=0.2351, cr_loss=0.4192, over 17041.00 frames. ], tot_loss[loss=0.3163, ctc_loss=0.2337, cr_loss=0.4131, over 3354038.68 frames. ], batch size: 44, lr: 2.70e-02, grad_scale: 32.0
2024-09-22 17:26:38,048 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=11.55 vs. limit=15.0
2024-09-22 17:27:13,235 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.664e+02 1.971e+02 2.314e+02 3.882e+02, threshold=3.942e+02, percent-clipped=0.0
2024-09-22 17:27:16,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=65370.666666666664, ans=0.125
2024-09-22 17:27:49,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=65464.0, ans=0.0
2024-09-22 17:27:58,573 INFO [train.py:1198] (1/4) Epoch 4, batch 2350, loss[loss=0.3324, ctc_loss=0.2452, cr_loss=0.4361, over 16990.00 frames. ], tot_loss[loss=0.3168, ctc_loss=0.2341, cr_loss=0.4134, over 3347979.03 frames. ], batch size: 51, lr: 2.69e-02, grad_scale: 32.0
2024-09-22 17:27:58,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=65510.666666666664, ans=0.125
2024-09-22 17:28:32,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=65604.0, ans=0.0
2024-09-22 17:28:37,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=65604.0, ans=0.2
2024-09-22 17:28:38,102 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.13 vs. limit=15.0
2024-09-22 17:28:38,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=65604.0, ans=0.125
2024-09-22 17:28:42,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=65604.0, ans=0.04949747468305833
2024-09-22 17:28:48,513 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:28:59,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=65650.66666666667, ans=0.2
2024-09-22 17:29:07,821 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=17.07 vs. limit=22.5
2024-09-22 17:29:18,256 INFO [train.py:1198] (1/4) Epoch 4, batch 2400, loss[loss=0.3908, ctc_loss=0.3035, cr_loss=0.4366, over 11621.00 frames. ], tot_loss[loss=0.3169, ctc_loss=0.2342, cr_loss=0.4135, over 3342656.44 frames. ], batch size: 123, lr: 2.69e-02, grad_scale: 32.0
2024-09-22 17:29:31,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=65744.0, ans=0.125
2024-09-22 17:29:42,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=65790.66666666667, ans=0.0
2024-09-22 17:29:52,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=65837.33333333333, ans=0.125
2024-09-22 17:29:54,855 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.273e+02 1.593e+02 1.794e+02 2.177e+02 3.793e+02, threshold=3.589e+02, percent-clipped=0.0
2024-09-22 17:29:59,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=65837.33333333333, ans=0.0
2024-09-22 17:30:01,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=65837.33333333333, ans=0.025
2024-09-22 17:30:20,449 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.49 vs. limit=15.0
2024-09-22 17:30:24,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=65930.66666666667, ans=0.125
2024-09-22 17:30:40,198 INFO [train.py:1198] (1/4) Epoch 4, batch 2450, loss[loss=0.2995, ctc_loss=0.2214, cr_loss=0.3905, over 16965.00 frames. ], tot_loss[loss=0.316, ctc_loss=0.2336, cr_loss=0.4123, over 3332108.66 frames. ], batch size: 42, lr: 2.68e-02, grad_scale: 32.0
2024-09-22 17:30:58,059 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:31:09,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=66024.0, ans=0.125
2024-09-22 17:31:19,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=11.26 vs. limit=15.0
2024-09-22 17:31:22,382 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.35 vs. limit=15.0
2024-09-22 17:31:27,060 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=20.76 vs. limit=22.5
2024-09-22 17:31:30,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=66117.33333333333, ans=0.125
2024-09-22 17:31:33,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=66117.33333333333, ans=0.125
2024-09-22 17:31:34,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=66117.33333333333, ans=0.125
2024-09-22 17:31:44,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=66164.0, ans=0.125
2024-09-22 17:31:52,105 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.18 vs. limit=12.0
2024-09-22 17:32:01,361 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=12.01 vs. limit=15.0
2024-09-22 17:32:02,189 INFO [train.py:1198] (1/4) Epoch 4, batch 2500, loss[loss=0.3291, ctc_loss=0.2403, cr_loss=0.4444, over 17222.00 frames. ], tot_loss[loss=0.3147, ctc_loss=0.2324, cr_loss=0.4114, over 3344666.04 frames. ], batch size: 55, lr: 2.68e-02, grad_scale: 32.0
2024-09-22 17:32:41,646 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 1.715e+02 1.996e+02 2.438e+02 3.886e+02, threshold=3.992e+02, percent-clipped=3.0
2024-09-22 17:32:45,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=66304.0, ans=0.125
2024-09-22 17:32:49,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=66304.0, ans=0.0
2024-09-22 17:32:51,772 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=15.55 vs. limit=15.0
2024-09-22 17:33:20,174 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:33:20,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=66397.33333333333, ans=0.125
2024-09-22 17:33:24,679 INFO [train.py:1198] (1/4) Epoch 4, batch 2550, loss[loss=0.3226, ctc_loss=0.2417, cr_loss=0.4046, over 17302.00 frames. ], tot_loss[loss=0.3135, ctc_loss=0.2315, cr_loss=0.4103, over 3349519.58 frames. ], batch size: 51, lr: 2.68e-02, grad_scale: 32.0
2024-09-22 17:33:50,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=66490.66666666667, ans=0.125
2024-09-22 17:33:54,060 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.88 vs. limit=15.0
2024-09-22 17:34:15,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=66584.0, ans=0.0
2024-09-22 17:34:44,460 INFO [train.py:1198] (1/4) Epoch 4, batch 2600, loss[loss=0.3445, ctc_loss=0.256, cr_loss=0.4426, over 17239.00 frames. ], tot_loss[loss=0.3146, ctc_loss=0.2324, cr_loss=0.4109, over 3338881.62 frames. ], batch size: 55, lr: 2.67e-02, grad_scale: 32.0
2024-09-22 17:35:11,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=66724.0, ans=0.1
2024-09-22 17:35:25,830 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.648e+02 1.831e+02 2.212e+02 5.606e+02, threshold=3.662e+02, percent-clipped=1.0
2024-09-22 17:36:01,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=66864.0, ans=0.125
2024-09-22 17:36:08,650 INFO [train.py:1198] (1/4) Epoch 4, batch 2650, loss[loss=0.2833, ctc_loss=0.1989, cr_loss=0.4221, over 17006.00 frames. ], tot_loss[loss=0.3147, ctc_loss=0.2325, cr_loss=0.4111, over 3331854.23 frames. ], batch size: 39, lr: 2.67e-02, grad_scale: 32.0
2024-09-22 17:36:13,039 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=13.26 vs. limit=22.5
2024-09-22 17:36:24,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=66957.33333333333, ans=0.0
2024-09-22 17:36:53,022 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.76 vs. limit=15.0
2024-09-22 17:37:04,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.min_positive, batch_count=67050.66666666667, ans=0.025
2024-09-22 17:37:32,620 INFO [train.py:1198] (1/4) Epoch 4, batch 2700, loss[loss=0.283, ctc_loss=0.2045, cr_loss=0.3928, over 17192.00 frames. ], tot_loss[loss=0.3148, ctc_loss=0.2325, cr_loss=0.4111, over 3327959.88 frames. ], batch size: 41, lr: 2.67e-02, grad_scale: 32.0
2024-09-22 17:38:04,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=67237.33333333333, ans=0.125
2024-09-22 17:38:05,057 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=11.08 vs. limit=15.0
2024-09-22 17:38:09,012 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 1.678e+02 1.945e+02 2.373e+02 3.767e+02, threshold=3.890e+02, percent-clipped=1.0
2024-09-22 17:38:20,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=67284.0, ans=0.0
2024-09-22 17:38:52,039 INFO [train.py:1198] (1/4) Epoch 4, batch 2750, loss[loss=0.3094, ctc_loss=0.2269, cr_loss=0.4124, over 17265.00 frames. ], tot_loss[loss=0.3142, ctc_loss=0.2321, cr_loss=0.4105, over 3339959.97 frames. ], batch size: 44, lr: 2.66e-02, grad_scale: 32.0
2024-09-22 17:38:55,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=67377.33333333333, ans=0.125
2024-09-22 17:39:05,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=67377.33333333333, ans=0.07
2024-09-22 17:39:18,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=67424.0, ans=0.125
2024-09-22 17:39:31,439 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.22 vs. limit=6.0
2024-09-22 17:39:42,397 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=6.41 vs. limit=15.0
2024-09-22 17:39:43,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=67517.33333333333, ans=0.0
2024-09-22 17:39:52,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=67517.33333333333, ans=0.0
2024-09-22 17:40:06,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=67564.0, ans=0.125
2024-09-22 17:40:17,122 INFO [train.py:1198] (1/4) Epoch 4, batch 2800, loss[loss=0.3415, ctc_loss=0.2494, cr_loss=0.4604, over 17052.00 frames. ], tot_loss[loss=0.3136, ctc_loss=0.2315, cr_loss=0.4104, over 3333137.15 frames. ], batch size: 56, lr: 2.66e-02, grad_scale: 32.0
2024-09-22 17:40:53,564 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.604e+02 1.847e+02 2.270e+02 3.677e+02, threshold=3.693e+02, percent-clipped=0.0
2024-09-22 17:41:03,741 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.27 vs. limit=22.5
2024-09-22 17:41:21,015 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=22.45 vs. limit=22.5
2024-09-22 17:41:24,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.91 vs. limit=6.0
2024-09-22 17:41:38,544 INFO [train.py:1198] (1/4) Epoch 4, batch 2850, loss[loss=0.3238, ctc_loss=0.2391, cr_loss=0.4235, over 16618.00 frames. ], tot_loss[loss=0.3155, ctc_loss=0.2331, cr_loss=0.4124, over 3341817.37 frames. ], batch size: 66, lr: 2.65e-02, grad_scale: 32.0
2024-09-22 17:41:46,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=67844.0, ans=0.2
2024-09-22 17:42:02,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=67890.66666666667, ans=0.0
2024-09-22 17:42:14,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=67937.33333333333, ans=0.1
2024-09-22 17:42:16,477 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:42:21,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=67937.33333333333, ans=0.04949747468305833
2024-09-22 17:42:29,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=67984.0, ans=0.125
2024-09-22 17:42:30,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=67984.0, ans=0.125
2024-09-22 17:42:46,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=68030.66666666667, ans=0.2
2024-09-22 17:42:50,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=68030.66666666667, ans=0.2
2024-09-22 17:43:00,659 INFO [train.py:1198] (1/4) Epoch 4, batch 2900, loss[loss=0.2629, ctc_loss=0.1884, cr_loss=0.3722, over 17105.00 frames. ], tot_loss[loss=0.3139, ctc_loss=0.2317, cr_loss=0.411, over 3342511.91 frames. ], batch size: 40, lr: 2.65e-02, grad_scale: 32.0
2024-09-22 17:43:04,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=68077.33333333333, ans=0.125
2024-09-22 17:43:12,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=68077.33333333333, ans=0.1
2024-09-22 17:43:18,588 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:43:37,709 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.627e+02 1.915e+02 2.361e+02 4.224e+02, threshold=3.831e+02, percent-clipped=1.0
2024-09-22 17:43:47,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=68217.33333333333, ans=0.125
2024-09-22 17:43:47,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=68217.33333333333, ans=0.125
2024-09-22 17:43:56,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=68217.33333333333, ans=10.0
2024-09-22 17:44:07,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=68264.0, ans=0.07
2024-09-22 17:44:20,377 INFO [train.py:1198] (1/4) Epoch 4, batch 2950, loss[loss=0.2774, ctc_loss=0.1994, cr_loss=0.3898, over 17270.00 frames. ], tot_loss[loss=0.3136, ctc_loss=0.2314, cr_loss=0.411, over 3345013.47 frames. ], batch size: 42, lr: 2.65e-02, grad_scale: 32.0
2024-09-22 17:44:40,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2.whitening_limit, batch_count=68357.33333333333, ans=15.0
2024-09-22 17:45:13,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=68450.66666666667, ans=0.125
2024-09-22 17:45:18,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=68450.66666666667, ans=0.0
2024-09-22 17:45:19,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=13.11 vs. limit=15.0
2024-09-22 17:45:20,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=68450.66666666667, ans=0.125
2024-09-22 17:45:44,885 INFO [train.py:1198] (1/4) Epoch 4, batch 3000, loss[loss=0.3044, ctc_loss=0.2201, cr_loss=0.4219, over 17207.00 frames. ], tot_loss[loss=0.3131, ctc_loss=0.2309, cr_loss=0.4108, over 3348489.56 frames. ], batch size: 47, lr: 2.64e-02, grad_scale: 32.0
2024-09-22 17:45:44,886 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 17:46:00,416 INFO [train.py:1230] (1/4) Epoch 4, validation: loss=0.07263, ctc_loss=0.07263, cr_loss=7.17e-15, over 944034.00 frames. 
2024-09-22 17:46:00,417 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 17:46:11,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=68544.0, ans=0.125
2024-09-22 17:46:16,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=68590.66666666667, ans=0.125
2024-09-22 17:46:36,347 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.618e+02 1.808e+02 2.126e+02 4.273e+02, threshold=3.616e+02, percent-clipped=2.0
2024-09-22 17:46:46,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=68684.0, ans=0.0
2024-09-22 17:46:54,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=68684.0, ans=0.125
2024-09-22 17:47:16,554 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.53 vs. limit=22.5
2024-09-22 17:47:19,107 INFO [train.py:1198] (1/4) Epoch 4, batch 3050, loss[loss=0.3987, ctc_loss=0.31, cr_loss=0.4434, over 11857.00 frames. ], tot_loss[loss=0.3128, ctc_loss=0.2307, cr_loss=0.4105, over 3340687.20 frames. ], batch size: 123, lr: 2.64e-02, grad_scale: 32.0
2024-09-22 17:47:31,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=68777.33333333333, ans=0.125
2024-09-22 17:47:48,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=68824.0, ans=0.125
2024-09-22 17:48:02,084 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=15.18 vs. limit=15.0
2024-09-22 17:48:06,414 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.40 vs. limit=22.5
2024-09-22 17:48:42,984 INFO [train.py:1198] (1/4) Epoch 4, batch 3100, loss[loss=0.3236, ctc_loss=0.2429, cr_loss=0.4033, over 17218.00 frames. ], tot_loss[loss=0.3123, ctc_loss=0.2304, cr_loss=0.4096, over 3340563.67 frames. ], batch size: 55, lr: 2.64e-02, grad_scale: 32.0
2024-09-22 17:48:43,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=69010.66666666667, ans=0.125
2024-09-22 17:48:52,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=69010.66666666667, ans=0.1
2024-09-22 17:49:01,262 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.10 vs. limit=6.0
2024-09-22 17:49:11,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=69057.33333333333, ans=0.125
2024-09-22 17:49:19,305 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.562e+02 1.773e+02 2.272e+02 4.016e+02, threshold=3.545e+02, percent-clipped=1.0
2024-09-22 17:49:25,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=69104.0, ans=0.125
2024-09-22 17:49:46,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=69197.33333333333, ans=0.125
2024-09-22 17:50:01,897 INFO [train.py:1198] (1/4) Epoch 4, batch 3150, loss[loss=0.3402, ctc_loss=0.2491, cr_loss=0.4557, over 17296.00 frames. ], tot_loss[loss=0.312, ctc_loss=0.2299, cr_loss=0.4102, over 3346237.79 frames. ], batch size: 49, lr: 2.63e-02, grad_scale: 32.0
2024-09-22 17:50:18,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=69290.66666666667, ans=0.125
2024-09-22 17:50:40,730 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:50:56,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=69384.0, ans=0.2
2024-09-22 17:51:05,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=69430.66666666667, ans=0.5
2024-09-22 17:51:19,753 INFO [train.py:1198] (1/4) Epoch 4, batch 3200, loss[loss=0.3285, ctc_loss=0.2467, cr_loss=0.4089, over 14990.00 frames. ], tot_loss[loss=0.3114, ctc_loss=0.2294, cr_loss=0.4099, over 3346647.70 frames. ], batch size: 89, lr: 2.63e-02, grad_scale: 32.0
2024-09-22 17:51:24,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=69477.33333333333, ans=0.035
2024-09-22 17:51:38,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=69524.0, ans=0.125
2024-09-22 17:51:53,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=69570.66666666667, ans=0.1
2024-09-22 17:51:55,750 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.584e+02 1.798e+02 2.186e+02 3.575e+02, threshold=3.596e+02, percent-clipped=1.0
2024-09-22 17:51:57,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=69570.66666666667, ans=0.0
2024-09-22 17:51:57,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=69570.66666666667, ans=0.1
2024-09-22 17:51:59,043 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:52:00,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=69570.66666666667, ans=0.025
2024-09-22 17:52:37,931 INFO [train.py:1198] (1/4) Epoch 4, batch 3250, loss[loss=0.3392, ctc_loss=0.2564, cr_loss=0.4138, over 15802.00 frames. ], tot_loss[loss=0.3096, ctc_loss=0.2278, cr_loss=0.4086, over 3353000.33 frames. ], batch size: 74, lr: 2.63e-02, grad_scale: 32.0
2024-09-22 17:52:55,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=69757.33333333333, ans=0.125
2024-09-22 17:52:59,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=69757.33333333333, ans=0.0
2024-09-22 17:53:11,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=69804.0, ans=0.5
2024-09-22 17:53:14,552 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 17:53:19,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=69804.0, ans=0.1
2024-09-22 17:53:34,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=69850.66666666667, ans=0.0
2024-09-22 17:53:43,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=69897.33333333333, ans=0.125
2024-09-22 17:53:56,009 INFO [train.py:1198] (1/4) Epoch 4, batch 3300, loss[loss=0.2628, ctc_loss=0.1939, cr_loss=0.3447, over 16949.00 frames. ], tot_loss[loss=0.311, ctc_loss=0.229, cr_loss=0.41, over 3354293.27 frames. ], batch size: 42, lr: 2.62e-02, grad_scale: 32.0
2024-09-22 17:54:29,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=70037.33333333333, ans=0.025
2024-09-22 17:54:32,122 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 1.544e+02 1.731e+02 2.121e+02 3.523e+02, threshold=3.462e+02, percent-clipped=0.0
2024-09-22 17:55:00,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=70130.66666666667, ans=0.125
2024-09-22 17:55:03,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=70130.66666666667, ans=0.0
2024-09-22 17:55:13,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=70130.66666666667, ans=0.125
2024-09-22 17:55:19,190 INFO [train.py:1198] (1/4) Epoch 4, batch 3350, loss[loss=0.3075, ctc_loss=0.2238, cr_loss=0.418, over 17098.00 frames. ], tot_loss[loss=0.3102, ctc_loss=0.2282, cr_loss=0.4099, over 3359728.96 frames. ], batch size: 49, lr: 2.62e-02, grad_scale: 32.0
2024-09-22 17:55:39,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=70224.0, ans=0.125
2024-09-22 17:55:39,773 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.04 vs. limit=10.0
2024-09-22 17:55:44,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=70224.0, ans=0.125
2024-09-22 17:55:58,640 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.32 vs. limit=15.0
2024-09-22 17:56:18,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=70317.33333333333, ans=0.0
2024-09-22 17:56:21,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=70364.0, ans=0.125
2024-09-22 17:56:36,980 INFO [train.py:1198] (1/4) Epoch 4, batch 3400, loss[loss=0.2714, ctc_loss=0.2004, cr_loss=0.3549, over 17263.00 frames. ], tot_loss[loss=0.3102, ctc_loss=0.2282, cr_loss=0.4099, over 3358999.39 frames. ], batch size: 42, lr: 2.62e-02, grad_scale: 32.0
2024-09-22 17:56:51,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.51 vs. limit=22.5
2024-09-22 17:57:12,457 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.333e+02 1.518e+02 1.665e+02 2.014e+02 3.333e+02, threshold=3.330e+02, percent-clipped=0.0
2024-09-22 17:57:15,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.max_abs, batch_count=70504.0, ans=10.0
2024-09-22 17:57:25,575 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.82 vs. limit=15.0
2024-09-22 17:57:35,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=70550.66666666667, ans=0.04949747468305833
2024-09-22 17:57:54,208 INFO [train.py:1198] (1/4) Epoch 4, batch 3450, loss[loss=0.2951, ctc_loss=0.2159, cr_loss=0.3958, over 17363.00 frames. ], tot_loss[loss=0.3111, ctc_loss=0.229, cr_loss=0.4103, over 3356355.92 frames. ], batch size: 48, lr: 2.61e-02, grad_scale: 32.0
2024-09-22 17:58:24,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=70690.66666666667, ans=0.125
2024-09-22 17:58:33,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=70737.33333333333, ans=0.125
2024-09-22 17:58:35,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=70737.33333333333, ans=0.125
2024-09-22 17:58:38,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=70737.33333333333, ans=0.2
2024-09-22 17:58:44,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=70784.0, ans=0.0
2024-09-22 17:58:52,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=70784.0, ans=0.125
2024-09-22 17:59:15,855 INFO [train.py:1198] (1/4) Epoch 4, batch 3500, loss[loss=0.3673, ctc_loss=0.2722, cr_loss=0.4757, over 16680.00 frames. ], tot_loss[loss=0.3118, ctc_loss=0.2296, cr_loss=0.411, over 3359886.78 frames. ], batch size: 61, lr: 2.61e-02, grad_scale: 32.0
2024-09-22 17:59:25,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=70877.33333333333, ans=0.1
2024-09-22 17:59:34,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=70924.0, ans=0.125
2024-09-22 17:59:35,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.06 vs. limit=6.0
2024-09-22 17:59:47,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=70970.66666666667, ans=0.125
2024-09-22 17:59:53,260 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.551e+02 1.744e+02 2.057e+02 3.215e+02, threshold=3.488e+02, percent-clipped=0.0
2024-09-22 18:00:07,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=71017.33333333333, ans=0.2
2024-09-22 18:00:14,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=71017.33333333333, ans=0.1
2024-09-22 18:00:20,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=71064.0, ans=0.125
2024-09-22 18:00:24,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_positive, batch_count=71064.0, ans=0.05
2024-09-22 18:00:27,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=71064.0, ans=0.0
2024-09-22 18:00:33,938 INFO [train.py:1198] (1/4) Epoch 4, batch 3550, loss[loss=0.2762, ctc_loss=0.2041, cr_loss=0.3605, over 17262.00 frames. ], tot_loss[loss=0.3097, ctc_loss=0.2279, cr_loss=0.4093, over 3358461.30 frames. ], batch size: 44, lr: 2.61e-02, grad_scale: 16.0
2024-09-22 18:00:43,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=71110.66666666667, ans=0.0
2024-09-22 18:01:19,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=71250.66666666667, ans=0.0
2024-09-22 18:01:51,387 INFO [train.py:1198] (1/4) Epoch 4, batch 3600, loss[loss=0.2784, ctc_loss=0.2009, cr_loss=0.3874, over 17008.00 frames. ], tot_loss[loss=0.3115, ctc_loss=0.2293, cr_loss=0.4108, over 3347993.92 frames. ], batch size: 44, lr: 2.60e-02, grad_scale: 32.0
2024-09-22 18:01:59,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=71344.0, ans=0.125
2024-09-22 18:02:18,083 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.77 vs. limit=15.0
2024-09-22 18:02:18,367 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=7.51 vs. limit=15.0
2024-09-22 18:02:24,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=71437.33333333333, ans=0.1
2024-09-22 18:02:28,362 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.219e+02 1.639e+02 2.046e+02 2.808e+02 4.325e+02, threshold=4.093e+02, percent-clipped=8.0
2024-09-22 18:02:30,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=71437.33333333333, ans=0.125
2024-09-22 18:03:03,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=71530.66666666667, ans=0.125
2024-09-22 18:03:09,092 INFO [train.py:1198] (1/4) Epoch 4, batch 3650, loss[loss=0.3023, ctc_loss=0.2159, cr_loss=0.4323, over 17196.00 frames. ], tot_loss[loss=0.3118, ctc_loss=0.2295, cr_loss=0.4111, over 3336468.55 frames. ], batch size: 45, lr: 2.60e-02, grad_scale: 32.0
2024-09-22 18:03:23,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=71624.0, ans=0.2
2024-09-22 18:03:29,652 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 18:04:27,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=71810.66666666667, ans=0.0
2024-09-22 18:04:28,661 INFO [train.py:1198] (1/4) Epoch 4, batch 3700, loss[loss=0.2943, ctc_loss=0.2145, cr_loss=0.3987, over 17003.00 frames. ], tot_loss[loss=0.3129, ctc_loss=0.2304, cr_loss=0.4123, over 3340200.45 frames. ], batch size: 44, lr: 2.60e-02, grad_scale: 32.0
2024-09-22 18:04:30,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=71810.66666666667, ans=0.125
2024-09-22 18:04:44,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=71857.33333333333, ans=0.025
2024-09-22 18:04:47,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=71857.33333333333, ans=0.025
2024-09-22 18:05:07,669 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.267e+02 1.579e+02 1.797e+02 2.044e+02 5.255e+02, threshold=3.594e+02, percent-clipped=1.0
2024-09-22 18:05:48,767 INFO [train.py:1198] (1/4) Epoch 4, batch 3750, loss[loss=0.304, ctc_loss=0.2192, cr_loss=0.4242, over 16957.00 frames. ], tot_loss[loss=0.3151, ctc_loss=0.2322, cr_loss=0.4148, over 3333726.14 frames. ], batch size: 42, lr: 2.59e-02, grad_scale: 32.0
2024-09-22 18:06:06,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=72090.66666666667, ans=0.0
2024-09-22 18:06:17,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=72090.66666666667, ans=0.0
2024-09-22 18:06:49,081 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.14 vs. limit=22.5
2024-09-22 18:06:52,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=72230.66666666667, ans=0.0
2024-09-22 18:07:06,330 INFO [train.py:1198] (1/4) Epoch 4, batch 3800, loss[loss=0.3003, ctc_loss=0.2193, cr_loss=0.4049, over 17016.00 frames. ], tot_loss[loss=0.3166, ctc_loss=0.2335, cr_loss=0.4153, over 3318228.06 frames. ], batch size: 44, lr: 2.59e-02, grad_scale: 32.0
2024-09-22 18:07:11,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=72277.33333333333, ans=0.1
2024-09-22 18:07:44,153 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.589e+02 1.725e+02 2.068e+02 4.482e+02, threshold=3.450e+02, percent-clipped=2.0
2024-09-22 18:07:55,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=72417.33333333333, ans=0.125
2024-09-22 18:07:58,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=72417.33333333333, ans=0.1
2024-09-22 18:08:25,284 INFO [train.py:1198] (1/4) Epoch 4, batch 3850, loss[loss=0.3831, ctc_loss=0.295, cr_loss=0.4402, over 11654.00 frames. ], tot_loss[loss=0.3172, ctc_loss=0.2346, cr_loss=0.4128, over 3262015.53 frames. ], batch size: 123, lr: 2.59e-02, grad_scale: 32.0
2024-09-22 18:08:30,838 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=12.26 vs. limit=15.0
2024-09-22 18:08:34,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=72510.66666666667, ans=0.125
2024-09-22 18:08:41,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=72557.33333333333, ans=0.2
2024-09-22 18:08:46,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=14.87 vs. limit=22.5
2024-09-22 18:09:04,291 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=72604.0, ans=0.07
2024-09-22 18:10:26,643 INFO [train.py:1198] (1/4) Epoch 5, batch 0, loss[loss=0.3252, ctc_loss=0.2426, cr_loss=0.4127, over 17015.00 frames. ], tot_loss[loss=0.3252, ctc_loss=0.2426, cr_loss=0.4127, over 17015.00 frames. ], batch size: 52, lr: 2.40e-02, grad_scale: 32.0
2024-09-22 18:10:26,644 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 18:10:42,120 INFO [train.py:1230] (1/4) Epoch 5, validation: loss=0.07551, ctc_loss=0.07551, cr_loss=6.915e-15, over 944034.00 frames. 
2024-09-22 18:10:42,120 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 18:10:51,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=72725.33333333333, ans=0.1
2024-09-22 18:10:54,115 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.55 vs. limit=15.0
2024-09-22 18:10:59,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=72772.0, ans=0.125
2024-09-22 18:11:06,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=72772.0, ans=0.1
2024-09-22 18:11:10,261 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.61 vs. limit=15.0
2024-09-22 18:11:13,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.57 vs. limit=6.0
2024-09-22 18:11:27,091 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.660e+02 1.848e+02 2.232e+02 4.613e+02, threshold=3.696e+02, percent-clipped=4.0
2024-09-22 18:11:37,514 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.59 vs. limit=6.0
2024-09-22 18:11:48,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=72912.0, ans=0.1
2024-09-22 18:12:02,273 INFO [train.py:1198] (1/4) Epoch 5, batch 50, loss[loss=0.3012, ctc_loss=0.2206, cr_loss=0.4032, over 17026.00 frames. ], tot_loss[loss=0.3172, ctc_loss=0.2333, cr_loss=0.4192, over 749270.37 frames. ], batch size: 44, lr: 2.40e-02, grad_scale: 32.0
2024-09-22 18:12:04,567 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.93 vs. limit=15.0
2024-09-22 18:12:29,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=73005.33333333333, ans=0.2
2024-09-22 18:12:37,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=73052.0, ans=0.2
2024-09-22 18:12:57,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=73098.66666666667, ans=0.125
2024-09-22 18:13:10,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=73145.33333333333, ans=0.2
2024-09-22 18:13:19,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=73145.33333333333, ans=0.0
2024-09-22 18:13:23,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=73145.33333333333, ans=0.125
2024-09-22 18:13:27,566 INFO [train.py:1198] (1/4) Epoch 5, batch 100, loss[loss=0.3254, ctc_loss=0.2415, cr_loss=0.4195, over 15900.00 frames. ], tot_loss[loss=0.3126, ctc_loss=0.2294, cr_loss=0.416, over 1321954.25 frames. ], batch size: 74, lr: 2.40e-02, grad_scale: 32.0
2024-09-22 18:13:29,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=73192.0, ans=0.125
2024-09-22 18:13:40,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=73192.0, ans=0.0
2024-09-22 18:14:12,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 1.507e+02 1.769e+02 2.148e+02 4.396e+02, threshold=3.538e+02, percent-clipped=1.0
2024-09-22 18:14:18,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=73332.0, ans=0.0
2024-09-22 18:14:30,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=73378.66666666667, ans=15.0
2024-09-22 18:14:46,987 INFO [train.py:1198] (1/4) Epoch 5, batch 150, loss[loss=0.258, ctc_loss=0.1878, cr_loss=0.351, over 17186.00 frames. ], tot_loss[loss=0.3092, ctc_loss=0.2267, cr_loss=0.4122, over 1769470.83 frames. ], batch size: 47, lr: 2.40e-02, grad_scale: 32.0
2024-09-22 18:15:26,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=73518.66666666667, ans=0.125
2024-09-22 18:15:29,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=73518.66666666667, ans=15.0
2024-09-22 18:16:12,904 INFO [train.py:1198] (1/4) Epoch 5, batch 200, loss[loss=0.3617, ctc_loss=0.2709, cr_loss=0.4543, over 14896.00 frames. ], tot_loss[loss=0.309, ctc_loss=0.2265, cr_loss=0.4123, over 2109548.42 frames. ], batch size: 90, lr: 2.39e-02, grad_scale: 32.0
2024-09-22 18:16:13,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=6.18 vs. limit=12.0
2024-09-22 18:16:14,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=73658.66666666667, ans=0.2
2024-09-22 18:16:34,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.56 vs. limit=6.0
2024-09-22 18:16:57,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.454e+02 1.719e+02 2.223e+02 3.373e+02, threshold=3.438e+02, percent-clipped=0.0
2024-09-22 18:17:19,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=73845.33333333333, ans=0.0
2024-09-22 18:17:32,286 INFO [train.py:1198] (1/4) Epoch 5, batch 250, loss[loss=0.2889, ctc_loss=0.2061, cr_loss=0.4142, over 16772.00 frames. ], tot_loss[loss=0.3091, ctc_loss=0.2267, cr_loss=0.4121, over 2376463.79 frames. ], batch size: 37, lr: 2.39e-02, grad_scale: 32.0
2024-09-22 18:17:32,959 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.63 vs. limit=15.0
2024-09-22 18:18:20,029 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=13.76 vs. limit=15.0
2024-09-22 18:18:43,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=74078.66666666667, ans=0.125
2024-09-22 18:18:57,276 INFO [train.py:1198] (1/4) Epoch 5, batch 300, loss[loss=0.2406, ctc_loss=0.1788, cr_loss=0.3092, over 17062.00 frames. ], tot_loss[loss=0.308, ctc_loss=0.2258, cr_loss=0.4113, over 2604482.28 frames. ], batch size: 39, lr: 2.39e-02, grad_scale: 32.0
2024-09-22 18:18:57,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=74125.33333333333, ans=0.2
2024-09-22 18:19:07,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=74125.33333333333, ans=0.0
2024-09-22 18:19:09,167 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.44 vs. limit=15.0
2024-09-22 18:19:15,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=74172.0, ans=0.1
2024-09-22 18:19:27,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=74218.66666666667, ans=0.0
2024-09-22 18:19:41,768 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.262e+02 1.584e+02 1.929e+02 2.265e+02 4.720e+02, threshold=3.859e+02, percent-clipped=2.0
2024-09-22 18:19:49,872 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 18:19:57,998 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=16.10 vs. limit=22.5
2024-09-22 18:20:19,148 INFO [train.py:1198] (1/4) Epoch 5, batch 350, loss[loss=0.2932, ctc_loss=0.2163, cr_loss=0.3844, over 17321.00 frames. ], tot_loss[loss=0.3096, ctc_loss=0.2271, cr_loss=0.4124, over 2767459.09 frames. ], batch size: 51, lr: 2.38e-02, grad_scale: 32.0
2024-09-22 18:20:19,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=74358.66666666667, ans=0.0
2024-09-22 18:20:22,922 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.21 vs. limit=6.0
2024-09-22 18:20:52,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=74452.0, ans=0.07
2024-09-22 18:21:13,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=74498.66666666667, ans=0.125
2024-09-22 18:21:41,486 INFO [train.py:1198] (1/4) Epoch 5, batch 400, loss[loss=0.3229, ctc_loss=0.2396, cr_loss=0.4164, over 16927.00 frames. ], tot_loss[loss=0.3072, ctc_loss=0.2251, cr_loss=0.4105, over 2897082.27 frames. ], batch size: 58, lr: 2.38e-02, grad_scale: 32.0
2024-09-22 18:21:44,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=74592.0, ans=0.015
2024-09-22 18:21:44,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=74592.0, ans=0.125
2024-09-22 18:21:54,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=74592.0, ans=0.125
2024-09-22 18:22:10,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=74638.66666666667, ans=0.0
2024-09-22 18:22:28,037 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 1.516e+02 1.701e+02 1.953e+02 3.306e+02, threshold=3.402e+02, percent-clipped=0.0
2024-09-22 18:22:33,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=74732.0, ans=0.09899494936611666
2024-09-22 18:23:05,460 INFO [train.py:1198] (1/4) Epoch 5, batch 450, loss[loss=0.2778, ctc_loss=0.2018, cr_loss=0.3798, over 17184.00 frames. ], tot_loss[loss=0.3059, ctc_loss=0.224, cr_loss=0.4096, over 3007146.04 frames. ], batch size: 41, lr: 2.38e-02, grad_scale: 32.0
2024-09-22 18:23:18,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=74825.33333333333, ans=0.0
2024-09-22 18:23:27,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=74872.0, ans=0.2
2024-09-22 18:23:38,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=74918.66666666667, ans=0.125
2024-09-22 18:24:10,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=75012.0, ans=0.1
2024-09-22 18:24:22,003 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=14.79 vs. limit=15.0
2024-09-22 18:24:26,622 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.61 vs. limit=22.5
2024-09-22 18:24:27,402 INFO [train.py:1198] (1/4) Epoch 5, batch 500, loss[loss=0.2811, ctc_loss=0.206, cr_loss=0.3753, over 17086.00 frames. ], tot_loss[loss=0.3052, ctc_loss=0.2234, cr_loss=0.409, over 3080683.32 frames. ], batch size: 49, lr: 2.37e-02, grad_scale: 32.0
2024-09-22 18:24:44,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=75105.33333333333, ans=0.1
2024-09-22 18:24:48,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=75105.33333333333, ans=0.0
2024-09-22 18:24:55,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.max_abs, batch_count=75105.33333333333, ans=10.0
2024-09-22 18:25:02,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.min_positive, batch_count=75152.0, ans=0.025
2024-09-22 18:25:07,275 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=18.23 vs. limit=22.5
2024-09-22 18:25:13,994 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.528e+02 1.760e+02 2.096e+02 4.266e+02, threshold=3.519e+02, percent-clipped=4.0
2024-09-22 18:25:44,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=75245.33333333333, ans=0.1
2024-09-22 18:25:45,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=75245.33333333333, ans=0.125
2024-09-22 18:25:51,523 INFO [train.py:1198] (1/4) Epoch 5, batch 550, loss[loss=0.2703, ctc_loss=0.197, cr_loss=0.3665, over 17250.00 frames. ], tot_loss[loss=0.3053, ctc_loss=0.2234, cr_loss=0.4094, over 3147936.02 frames. ], batch size: 44, lr: 2.37e-02, grad_scale: 32.0
2024-09-22 18:26:15,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=75338.66666666667, ans=0.2
2024-09-22 18:26:18,852 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 18:26:40,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=75432.0, ans=0.05
2024-09-22 18:26:44,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.33 vs. limit=10.0
2024-09-22 18:26:52,344 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.82 vs. limit=12.0
2024-09-22 18:27:10,588 INFO [train.py:1198] (1/4) Epoch 5, batch 600, loss[loss=0.2852, ctc_loss=0.2078, cr_loss=0.3868, over 17305.00 frames. ], tot_loss[loss=0.3062, ctc_loss=0.2242, cr_loss=0.4101, over 3193520.44 frames. ], batch size: 49, lr: 2.37e-02, grad_scale: 32.0
2024-09-22 18:27:12,989 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.13 vs. limit=22.5
2024-09-22 18:27:34,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=75572.0, ans=0.0
2024-09-22 18:27:44,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=75618.66666666667, ans=0.025
2024-09-22 18:27:48,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=75618.66666666667, ans=0.125
2024-09-22 18:27:57,743 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 1.538e+02 1.800e+02 2.212e+02 3.356e+02, threshold=3.600e+02, percent-clipped=0.0
2024-09-22 18:28:35,278 INFO [train.py:1198] (1/4) Epoch 5, batch 650, loss[loss=0.2534, ctc_loss=0.1848, cr_loss=0.3429, over 17088.00 frames. ], tot_loss[loss=0.3044, ctc_loss=0.2227, cr_loss=0.4089, over 3238433.98 frames. ], batch size: 43, lr: 2.36e-02, grad_scale: 32.0
2024-09-22 18:28:37,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=75758.66666666667, ans=0.0
2024-09-22 18:28:43,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=75758.66666666667, ans=0.0
2024-09-22 18:29:09,027 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=11.16 vs. limit=15.0
2024-09-22 18:29:11,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=75852.0, ans=0.125
2024-09-22 18:29:54,260 INFO [train.py:1198] (1/4) Epoch 5, batch 700, loss[loss=0.2893, ctc_loss=0.2088, cr_loss=0.4025, over 16772.00 frames. ], tot_loss[loss=0.3037, ctc_loss=0.2221, cr_loss=0.4079, over 3264507.57 frames. ], batch size: 37, lr: 2.36e-02, grad_scale: 32.0
2024-09-22 18:30:14,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=76038.66666666667, ans=0.125
2024-09-22 18:30:42,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=76085.33333333333, ans=0.125
2024-09-22 18:30:43,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 1.521e+02 1.705e+02 2.105e+02 2.881e+02, threshold=3.410e+02, percent-clipped=0.0
2024-09-22 18:31:00,122 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=5.35 vs. limit=12.0
2024-09-22 18:31:03,794 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=13.68 vs. limit=22.5
2024-09-22 18:31:18,063 INFO [train.py:1198] (1/4) Epoch 5, batch 750, loss[loss=0.27, ctc_loss=0.1959, cr_loss=0.3705, over 17133.00 frames. ], tot_loss[loss=0.3038, ctc_loss=0.2222, cr_loss=0.4084, over 3288632.19 frames. ], batch size: 40, lr: 2.36e-02, grad_scale: 32.0
2024-09-22 18:31:21,900 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.67 vs. limit=12.0
2024-09-22 18:31:24,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=76225.33333333333, ans=0.125
2024-09-22 18:31:39,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=76272.0, ans=0.125
2024-09-22 18:32:32,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=76412.0, ans=0.125
2024-09-22 18:32:37,282 INFO [train.py:1198] (1/4) Epoch 5, batch 800, loss[loss=0.3227, ctc_loss=0.2374, cr_loss=0.4265, over 17026.00 frames. ], tot_loss[loss=0.3044, ctc_loss=0.2226, cr_loss=0.4091, over 3311704.93 frames. ], batch size: 51, lr: 2.36e-02, grad_scale: 32.0
2024-09-22 18:32:39,549 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.12 vs. limit=6.0
2024-09-22 18:32:44,825 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 18:33:08,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=76505.33333333333, ans=0.1
2024-09-22 18:33:11,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=76505.33333333333, ans=0.0
2024-09-22 18:33:17,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=76552.0, ans=0.1
2024-09-22 18:33:19,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=76552.0, ans=0.1
2024-09-22 18:33:26,740 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 1.675e+02 1.893e+02 2.259e+02 3.192e+02, threshold=3.786e+02, percent-clipped=0.0
2024-09-22 18:33:55,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=76645.33333333333, ans=0.125
2024-09-22 18:34:01,757 INFO [train.py:1198] (1/4) Epoch 5, batch 850, loss[loss=0.2778, ctc_loss=0.1999, cr_loss=0.3896, over 17092.00 frames. ], tot_loss[loss=0.3042, ctc_loss=0.2224, cr_loss=0.4088, over 3323593.26 frames. ], batch size: 43, lr: 2.35e-02, grad_scale: 32.0
2024-09-22 18:34:10,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=76692.0, ans=0.1
2024-09-22 18:34:29,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.34 vs. limit=15.0
2024-09-22 18:35:23,901 INFO [train.py:1198] (1/4) Epoch 5, batch 900, loss[loss=0.3004, ctc_loss=0.2228, cr_loss=0.3875, over 17163.00 frames. ], tot_loss[loss=0.3034, ctc_loss=0.222, cr_loss=0.4071, over 3323812.43 frames. ], batch size: 45, lr: 2.35e-02, grad_scale: 32.0
2024-09-22 18:35:32,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=76925.33333333333, ans=0.1
2024-09-22 18:35:35,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=76925.33333333333, ans=0.125
2024-09-22 18:36:10,876 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.238e+02 1.424e+02 1.574e+02 1.806e+02 2.908e+02, threshold=3.147e+02, percent-clipped=0.0
2024-09-22 18:36:30,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=77112.0, ans=0.125
2024-09-22 18:36:35,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=77112.0, ans=0.1
2024-09-22 18:36:35,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=77112.0, ans=0.125
2024-09-22 18:36:40,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=9.11 vs. limit=15.0
2024-09-22 18:36:46,127 INFO [train.py:1198] (1/4) Epoch 5, batch 950, loss[loss=0.3204, ctc_loss=0.236, cr_loss=0.4224, over 17223.00 frames. ], tot_loss[loss=0.3038, ctc_loss=0.2222, cr_loss=0.408, over 3330221.62 frames. ], batch size: 50, lr: 2.35e-02, grad_scale: 32.0
2024-09-22 18:36:51,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=77158.66666666667, ans=0.125
2024-09-22 18:36:55,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=77158.66666666667, ans=0.025
2024-09-22 18:37:11,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=77205.33333333333, ans=0.125
2024-09-22 18:37:20,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=77252.0, ans=0.125
2024-09-22 18:37:25,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=77252.0, ans=0.1
2024-09-22 18:37:27,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=77252.0, ans=0.1
2024-09-22 18:37:33,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=5.01 vs. limit=5.0
2024-09-22 18:37:49,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=77298.66666666667, ans=0.125
2024-09-22 18:37:51,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=77345.33333333333, ans=0.125
2024-09-22 18:37:51,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.47 vs. limit=10.0
2024-09-22 18:38:10,579 INFO [train.py:1198] (1/4) Epoch 5, batch 1000, loss[loss=0.3373, ctc_loss=0.2505, cr_loss=0.4336, over 14848.00 frames. ], tot_loss[loss=0.3044, ctc_loss=0.2229, cr_loss=0.4079, over 3328126.92 frames. ], batch size: 89, lr: 2.34e-02, grad_scale: 32.0
2024-09-22 18:38:13,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=77392.0, ans=0.0
2024-09-22 18:38:14,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.58 vs. limit=10.0
2024-09-22 18:38:55,016 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.650e+02 1.785e+02 2.141e+02 3.125e+02, threshold=3.569e+02, percent-clipped=0.0
2024-09-22 18:39:21,436 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=5.80 vs. limit=15.0
2024-09-22 18:39:27,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=77578.66666666667, ans=0.125
2024-09-22 18:39:30,025 INFO [train.py:1198] (1/4) Epoch 5, batch 1050, loss[loss=0.2552, ctc_loss=0.1844, cr_loss=0.3537, over 17191.00 frames. ], tot_loss[loss=0.3049, ctc_loss=0.223, cr_loss=0.4095, over 3335572.10 frames. ], batch size: 41, lr: 2.34e-02, grad_scale: 32.0
2024-09-22 18:39:30,822 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.35 vs. limit=15.0
2024-09-22 18:40:15,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=77718.66666666667, ans=0.125
2024-09-22 18:40:37,866 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.12 vs. limit=15.0
2024-09-22 18:40:39,398 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.72 vs. limit=15.0
2024-09-22 18:40:54,538 INFO [train.py:1198] (1/4) Epoch 5, batch 1100, loss[loss=0.3037, ctc_loss=0.2237, cr_loss=0.4001, over 17142.00 frames. ], tot_loss[loss=0.3037, ctc_loss=0.2221, cr_loss=0.408, over 3335786.12 frames. ], batch size: 40, lr: 2.34e-02, grad_scale: 16.0
2024-09-22 18:41:31,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=77952.0, ans=0.125
2024-09-22 18:41:32,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=77952.0, ans=0.125
2024-09-22 18:41:40,427 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.520e+02 1.761e+02 2.071e+02 3.558e+02, threshold=3.523e+02, percent-clipped=0.0
2024-09-22 18:41:42,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.52 vs. limit=15.0
2024-09-22 18:41:56,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=78045.33333333333, ans=0.05
2024-09-22 18:41:59,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=78045.33333333333, ans=0.0
2024-09-22 18:42:02,124 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=11.29 vs. limit=15.0
2024-09-22 18:42:13,887 INFO [train.py:1198] (1/4) Epoch 5, batch 1150, loss[loss=0.2902, ctc_loss=0.2148, cr_loss=0.377, over 16729.00 frames. ], tot_loss[loss=0.3035, ctc_loss=0.222, cr_loss=0.4077, over 3334773.15 frames. ], batch size: 61, lr: 2.34e-02, grad_scale: 16.0
2024-09-22 18:42:30,419 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.58 vs. limit=15.0
2024-09-22 18:42:37,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=78138.66666666667, ans=0.125
2024-09-22 18:42:43,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=78138.66666666667, ans=0.2
2024-09-22 18:42:51,967 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.81 vs. limit=15.0
2024-09-22 18:43:03,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=78185.33333333333, ans=0.125
2024-09-22 18:43:23,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=78278.66666666667, ans=0.2
2024-09-22 18:43:26,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=78278.66666666667, ans=0.0
2024-09-22 18:43:27,633 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.46 vs. limit=10.0
2024-09-22 18:43:33,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=78278.66666666667, ans=0.0
2024-09-22 18:43:37,775 INFO [train.py:1198] (1/4) Epoch 5, batch 1200, loss[loss=0.3338, ctc_loss=0.2475, cr_loss=0.4313, over 17006.00 frames. ], tot_loss[loss=0.3035, ctc_loss=0.222, cr_loss=0.4076, over 3331900.78 frames. ], batch size: 53, lr: 2.33e-02, grad_scale: 32.0
2024-09-22 18:43:43,633 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.06 vs. limit=15.0
2024-09-22 18:44:05,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=78372.0, ans=0.1
2024-09-22 18:44:05,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=78372.0, ans=0.05
2024-09-22 18:44:13,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=78418.66666666667, ans=0.125
2024-09-22 18:44:24,262 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.563e+02 1.694e+02 1.943e+02 2.938e+02, threshold=3.387e+02, percent-clipped=0.0
2024-09-22 18:44:35,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=78465.33333333333, ans=0.025
2024-09-22 18:44:57,619 INFO [train.py:1198] (1/4) Epoch 5, batch 1250, loss[loss=0.3023, ctc_loss=0.2256, cr_loss=0.3832, over 16942.00 frames. ], tot_loss[loss=0.303, ctc_loss=0.2216, cr_loss=0.4067, over 3341872.87 frames. ], batch size: 58, lr: 2.33e-02, grad_scale: 32.0
2024-09-22 18:45:08,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=78558.66666666667, ans=0.125
2024-09-22 18:45:17,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=78605.33333333333, ans=0.125
2024-09-22 18:45:28,278 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.95 vs. limit=22.5
2024-09-22 18:45:51,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.25 vs. limit=10.0
2024-09-22 18:45:56,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=78698.66666666667, ans=0.025
2024-09-22 18:46:22,170 INFO [train.py:1198] (1/4) Epoch 5, batch 1300, loss[loss=0.3129, ctc_loss=0.2298, cr_loss=0.4155, over 16776.00 frames. ], tot_loss[loss=0.302, ctc_loss=0.2207, cr_loss=0.4065, over 3341845.02 frames. ], batch size: 61, lr: 2.33e-02, grad_scale: 32.0
2024-09-22 18:46:25,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=78792.0, ans=0.125
2024-09-22 18:46:29,258 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=13.72 vs. limit=22.5
2024-09-22 18:46:30,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=78792.0, ans=0.0
2024-09-22 18:47:08,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.495e+02 1.790e+02 2.182e+02 4.439e+02, threshold=3.579e+02, percent-clipped=1.0
2024-09-22 18:47:22,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=78932.0, ans=0.125
2024-09-22 18:47:30,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=78978.66666666667, ans=0.2
2024-09-22 18:47:33,923 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=20.13 vs. limit=22.5
2024-09-22 18:47:39,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=78978.66666666667, ans=0.2
2024-09-22 18:47:41,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=78978.66666666667, ans=0.2
2024-09-22 18:47:43,981 INFO [train.py:1198] (1/4) Epoch 5, batch 1350, loss[loss=0.3061, ctc_loss=0.2294, cr_loss=0.3835, over 17156.00 frames. ], tot_loss[loss=0.3021, ctc_loss=0.2208, cr_loss=0.4065, over 3335265.53 frames. ], batch size: 48, lr: 2.32e-02, grad_scale: 32.0
2024-09-22 18:48:16,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=79118.66666666667, ans=0.125
2024-09-22 18:49:05,571 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.77 vs. limit=6.0
2024-09-22 18:49:05,830 INFO [train.py:1198] (1/4) Epoch 5, batch 1400, loss[loss=0.3219, ctc_loss=0.2374, cr_loss=0.4225, over 17305.00 frames. ], tot_loss[loss=0.3036, ctc_loss=0.2221, cr_loss=0.4072, over 3337767.58 frames. ], batch size: 49, lr: 2.32e-02, grad_scale: 32.0
2024-09-22 18:49:21,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=79305.33333333333, ans=0.125
2024-09-22 18:49:24,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=79305.33333333333, ans=0.1
2024-09-22 18:49:28,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=79305.33333333333, ans=0.125
2024-09-22 18:49:54,449 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.617e+02 1.825e+02 2.247e+02 3.972e+02, threshold=3.649e+02, percent-clipped=1.0
2024-09-22 18:50:08,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=79398.66666666667, ans=0.07
2024-09-22 18:50:15,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=79445.33333333333, ans=0.125
2024-09-22 18:50:18,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=79445.33333333333, ans=0.125
2024-09-22 18:50:19,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.73 vs. limit=12.0
2024-09-22 18:50:24,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=79445.33333333333, ans=0.1
2024-09-22 18:50:30,041 INFO [train.py:1198] (1/4) Epoch 5, batch 1450, loss[loss=0.2821, ctc_loss=0.2019, cr_loss=0.4009, over 17214.00 frames. ], tot_loss[loss=0.3025, ctc_loss=0.221, cr_loss=0.4075, over 3345728.64 frames. ], batch size: 47, lr: 2.32e-02, grad_scale: 32.0
2024-09-22 18:51:14,222 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.80 vs. limit=15.0
2024-09-22 18:51:19,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=79632.0, ans=0.125
2024-09-22 18:51:19,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=79632.0, ans=0.125
2024-09-22 18:51:34,964 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.15 vs. limit=22.5
2024-09-22 18:51:49,952 INFO [train.py:1198] (1/4) Epoch 5, batch 1500, loss[loss=0.2932, ctc_loss=0.2141, cr_loss=0.3956, over 17301.00 frames. ], tot_loss[loss=0.3037, ctc_loss=0.2218, cr_loss=0.4095, over 3349893.86 frames. ], batch size: 49, lr: 2.32e-02, grad_scale: 32.0
2024-09-22 18:52:07,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=79772.0, ans=0.025
2024-09-22 18:52:32,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=79818.66666666667, ans=0.0
2024-09-22 18:52:40,768 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.573e+02 1.905e+02 2.453e+02 3.573e+02, threshold=3.810e+02, percent-clipped=0.0
2024-09-22 18:53:04,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=79912.0, ans=0.125
2024-09-22 18:53:14,049 INFO [train.py:1198] (1/4) Epoch 5, batch 1550, loss[loss=0.2755, ctc_loss=0.1972, cr_loss=0.3915, over 17099.00 frames. ], tot_loss[loss=0.3017, ctc_loss=0.2202, cr_loss=0.4076, over 3360232.81 frames. ], batch size: 40, lr: 2.31e-02, grad_scale: 32.0
2024-09-22 18:53:17,889 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.27 vs. limit=10.0
2024-09-22 18:53:44,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=80052.0, ans=0.125
2024-09-22 18:53:53,219 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.02 vs. limit=6.0
2024-09-22 18:54:15,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=80098.66666666667, ans=0.125
2024-09-22 18:54:16,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=80145.33333333333, ans=0.1
2024-09-22 18:54:33,899 INFO [train.py:1198] (1/4) Epoch 5, batch 1600, loss[loss=0.3352, ctc_loss=0.245, cr_loss=0.4509, over 16598.00 frames. ], tot_loss[loss=0.3003, ctc_loss=0.2188, cr_loss=0.4072, over 3368780.92 frames. ], batch size: 66, lr: 2.31e-02, grad_scale: 32.0
2024-09-22 18:54:42,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=80192.0, ans=0.125
2024-09-22 18:54:53,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=80238.66666666667, ans=0.125
2024-09-22 18:54:58,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=80238.66666666667, ans=0.125
2024-09-22 18:55:22,232 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.455e+02 1.591e+02 1.855e+02 3.051e+02, threshold=3.183e+02, percent-clipped=0.0
2024-09-22 18:55:22,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=80332.0, ans=0.07
2024-09-22 18:55:58,301 INFO [train.py:1198] (1/4) Epoch 5, batch 1650, loss[loss=0.2953, ctc_loss=0.2169, cr_loss=0.3919, over 16749.00 frames. ], tot_loss[loss=0.2993, ctc_loss=0.2181, cr_loss=0.4059, over 3370024.43 frames. ], batch size: 37, lr: 2.31e-02, grad_scale: 32.0
2024-09-22 18:55:58,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=80425.33333333333, ans=0.125
2024-09-22 18:56:00,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=80425.33333333333, ans=0.0
2024-09-22 18:56:02,323 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.99 vs. limit=22.5
2024-09-22 18:56:27,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=80472.0, ans=0.0
2024-09-22 18:56:51,207 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.06 vs. limit=15.0
2024-09-22 18:56:52,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=80565.33333333333, ans=0.1
2024-09-22 18:56:54,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=80565.33333333333, ans=0.0
2024-09-22 18:56:54,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=80565.33333333333, ans=0.0
2024-09-22 18:56:58,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=80565.33333333333, ans=0.1
2024-09-22 18:56:59,567 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.69 vs. limit=15.0
2024-09-22 18:57:05,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=80612.0, ans=0.125
2024-09-22 18:57:08,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=80612.0, ans=0.125
2024-09-22 18:57:19,846 INFO [train.py:1198] (1/4) Epoch 5, batch 1700, loss[loss=0.2932, ctc_loss=0.2122, cr_loss=0.405, over 17049.00 frames. ], tot_loss[loss=0.2981, ctc_loss=0.2171, cr_loss=0.4053, over 3375154.08 frames. ], batch size: 46, lr: 2.30e-02, grad_scale: 32.0
2024-09-22 18:57:27,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=80658.66666666667, ans=0.125
2024-09-22 18:57:38,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=80705.33333333333, ans=0.0
2024-09-22 18:57:57,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2.whitening_limit, batch_count=80752.0, ans=15.0
2024-09-22 18:58:08,258 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.556e+02 1.832e+02 2.128e+02 4.427e+02, threshold=3.664e+02, percent-clipped=3.0
2024-09-22 18:58:22,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=80798.66666666667, ans=0.125
2024-09-22 18:58:30,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=80845.33333333333, ans=0.2
2024-09-22 18:58:31,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=80845.33333333333, ans=0.125
2024-09-22 18:58:41,927 INFO [train.py:1198] (1/4) Epoch 5, batch 1750, loss[loss=0.3323, ctc_loss=0.2437, cr_loss=0.4433, over 17105.00 frames. ], tot_loss[loss=0.2986, ctc_loss=0.2173, cr_loss=0.4064, over 3376040.67 frames. ], batch size: 49, lr: 2.30e-02, grad_scale: 32.0
2024-09-22 18:58:45,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=80892.0, ans=0.0
2024-09-22 18:59:12,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=80985.33333333333, ans=0.2
2024-09-22 18:59:36,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=81032.0, ans=0.125
2024-09-22 19:00:03,798 INFO [train.py:1198] (1/4) Epoch 5, batch 1800, loss[loss=0.3437, ctc_loss=0.2492, cr_loss=0.4727, over 16568.00 frames. ], tot_loss[loss=0.2994, ctc_loss=0.218, cr_loss=0.407, over 3370396.78 frames. ], batch size: 66, lr: 2.30e-02, grad_scale: 32.0
2024-09-22 19:00:08,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=81125.33333333333, ans=0.0
2024-09-22 19:00:12,546 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=12.94 vs. limit=15.0
2024-09-22 19:00:12,882 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=13.16 vs. limit=22.5
2024-09-22 19:00:18,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=81172.0, ans=10.0
2024-09-22 19:00:19,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=81172.0, ans=0.025
2024-09-22 19:00:22,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=81172.0, ans=0.1
2024-09-22 19:00:33,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=81172.0, ans=0.05
2024-09-22 19:00:34,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=81172.0, ans=0.0
2024-09-22 19:00:38,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=81218.66666666667, ans=0.0
2024-09-22 19:00:52,231 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.571e+02 1.779e+02 2.119e+02 4.116e+02, threshold=3.559e+02, percent-clipped=1.0
2024-09-22 19:01:03,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=81265.33333333333, ans=0.2
2024-09-22 19:01:25,611 INFO [train.py:1198] (1/4) Epoch 5, batch 1850, loss[loss=0.2863, ctc_loss=0.209, cr_loss=0.3864, over 16306.00 frames. ], tot_loss[loss=0.2999, ctc_loss=0.2184, cr_loss=0.4072, over 3364243.27 frames. ], batch size: 36, lr: 2.30e-02, grad_scale: 32.0
2024-09-22 19:01:48,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=81405.33333333333, ans=0.2
2024-09-22 19:01:51,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=81405.33333333333, ans=0.2
2024-09-22 19:02:07,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=81452.0, ans=0.125
2024-09-22 19:02:46,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=81545.33333333333, ans=0.125
2024-09-22 19:02:50,453 INFO [train.py:1198] (1/4) Epoch 5, batch 1900, loss[loss=0.3177, ctc_loss=0.2301, cr_loss=0.4383, over 17040.00 frames. ], tot_loss[loss=0.3017, ctc_loss=0.22, cr_loss=0.4084, over 3355495.18 frames. ], batch size: 52, lr: 2.29e-02, grad_scale: 32.0
2024-09-22 19:02:52,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=81592.0, ans=0.125
2024-09-22 19:03:09,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=81638.66666666667, ans=0.125
2024-09-22 19:03:35,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=81685.33333333333, ans=0.1
2024-09-22 19:03:36,543 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.506e+02 1.844e+02 2.270e+02 3.771e+02, threshold=3.688e+02, percent-clipped=2.0
2024-09-22 19:03:41,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=81732.0, ans=0.1
2024-09-22 19:03:46,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=81732.0, ans=0.04949747468305833
2024-09-22 19:03:49,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=81732.0, ans=0.125
2024-09-22 19:03:49,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=81732.0, ans=0.035
2024-09-22 19:03:57,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=81778.66666666667, ans=0.125
2024-09-22 19:04:05,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=81778.66666666667, ans=0.07
2024-09-22 19:04:10,003 INFO [train.py:1198] (1/4) Epoch 5, batch 1950, loss[loss=0.308, ctc_loss=0.2296, cr_loss=0.3923, over 16712.00 frames. ], tot_loss[loss=0.3018, ctc_loss=0.2199, cr_loss=0.4093, over 3359270.50 frames. ], batch size: 61, lr: 2.29e-02, grad_scale: 32.0
2024-09-22 19:04:33,209 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.78 vs. limit=15.0
2024-09-22 19:04:36,518 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.65 vs. limit=15.0
2024-09-22 19:04:46,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=81918.66666666667, ans=0.125
2024-09-22 19:05:15,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=82012.0, ans=0.0
2024-09-22 19:05:26,400 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.26 vs. limit=12.0
2024-09-22 19:05:34,769 INFO [train.py:1198] (1/4) Epoch 5, batch 2000, loss[loss=0.2996, ctc_loss=0.214, cr_loss=0.4284, over 17030.00 frames. ], tot_loss[loss=0.3008, ctc_loss=0.219, cr_loss=0.4091, over 3368354.39 frames. ], batch size: 44, lr: 2.29e-02, grad_scale: 32.0
2024-09-22 19:05:43,251 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 19:05:54,931 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=7.41 vs. limit=15.0
2024-09-22 19:06:04,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.41 vs. limit=12.0
2024-09-22 19:06:19,054 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.76 vs. limit=22.5
2024-09-22 19:06:21,337 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.273e+02 1.502e+02 1.753e+02 2.170e+02 3.000e+02, threshold=3.507e+02, percent-clipped=0.0
2024-09-22 19:06:29,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=82198.66666666667, ans=0.125
2024-09-22 19:06:48,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=82245.33333333333, ans=0.125
2024-09-22 19:06:54,492 INFO [train.py:1198] (1/4) Epoch 5, batch 2050, loss[loss=0.2702, ctc_loss=0.1993, cr_loss=0.3545, over 17163.00 frames. ], tot_loss[loss=0.299, ctc_loss=0.2176, cr_loss=0.4069, over 3370865.88 frames. ], batch size: 45, lr: 2.28e-02, grad_scale: 32.0
2024-09-22 19:06:58,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=82292.0, ans=0.1
2024-09-22 19:07:31,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=82385.33333333333, ans=0.1
2024-09-22 19:08:18,229 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.68 vs. limit=10.0
2024-09-22 19:08:18,626 INFO [train.py:1198] (1/4) Epoch 5, batch 2100, loss[loss=0.2703, ctc_loss=0.1974, cr_loss=0.3648, over 17078.00 frames. ], tot_loss[loss=0.297, ctc_loss=0.2161, cr_loss=0.4045, over 3369076.71 frames. ], batch size: 43, lr: 2.28e-02, grad_scale: 32.0
2024-09-22 19:08:25,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=82525.33333333333, ans=0.125
2024-09-22 19:08:35,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=82572.0, ans=22.5
2024-09-22 19:09:00,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.51 vs. limit=12.0
2024-09-22 19:09:04,411 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.583e+02 1.843e+02 2.179e+02 3.617e+02, threshold=3.686e+02, percent-clipped=2.0
2024-09-22 19:09:06,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=82665.33333333333, ans=0.2
2024-09-22 19:09:15,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=82665.33333333333, ans=0.125
2024-09-22 19:09:19,543 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.20 vs. limit=22.5
2024-09-22 19:09:40,263 INFO [train.py:1198] (1/4) Epoch 5, batch 2150, loss[loss=0.3729, ctc_loss=0.2763, cr_loss=0.4831, over 16395.00 frames. ], tot_loss[loss=0.2973, ctc_loss=0.2163, cr_loss=0.4049, over 3369989.30 frames. ], batch size: 66, lr: 2.28e-02, grad_scale: 32.0
2024-09-22 19:09:44,481 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.40 vs. limit=15.0
2024-09-22 19:09:58,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=82805.33333333333, ans=0.0
2024-09-22 19:10:13,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=82852.0, ans=0.0
2024-09-22 19:10:13,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=82852.0, ans=0.0
2024-09-22 19:10:21,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.74 vs. limit=15.0
2024-09-22 19:10:29,908 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.37 vs. limit=6.0
2024-09-22 19:10:54,767 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.00 vs. limit=15.0
2024-09-22 19:11:02,148 INFO [train.py:1198] (1/4) Epoch 5, batch 2200, loss[loss=0.2611, ctc_loss=0.1859, cr_loss=0.3761, over 17030.00 frames. ], tot_loss[loss=0.297, ctc_loss=0.2159, cr_loss=0.4052, over 3367255.83 frames. ], batch size: 39, lr: 2.28e-02, grad_scale: 32.0
2024-09-22 19:11:48,592 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 1.603e+02 1.776e+02 2.386e+02 3.569e+02, threshold=3.552e+02, percent-clipped=0.0
2024-09-22 19:12:12,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=83178.66666666667, ans=0.025
2024-09-22 19:12:17,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=83178.66666666667, ans=0.0
2024-09-22 19:12:24,734 INFO [train.py:1198] (1/4) Epoch 5, batch 2250, loss[loss=0.3167, ctc_loss=0.2331, cr_loss=0.4184, over 16491.00 frames. ], tot_loss[loss=0.2969, ctc_loss=0.216, cr_loss=0.4045, over 3370113.80 frames. ], batch size: 66, lr: 2.27e-02, grad_scale: 32.0
2024-09-22 19:12:29,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=83225.33333333333, ans=0.025
2024-09-22 19:12:39,320 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=21.20 vs. limit=22.5
2024-09-22 19:12:42,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=83272.0, ans=0.1
2024-09-22 19:12:43,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=83272.0, ans=0.0
2024-09-22 19:13:12,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.38 vs. limit=15.0
2024-09-22 19:13:40,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=83412.0, ans=0.125
2024-09-22 19:13:46,565 INFO [train.py:1198] (1/4) Epoch 5, batch 2300, loss[loss=0.3063, ctc_loss=0.224, cr_loss=0.4115, over 17019.00 frames. ], tot_loss[loss=0.2958, ctc_loss=0.215, cr_loss=0.4038, over 3373721.69 frames. ], batch size: 44, lr: 2.27e-02, grad_scale: 32.0
2024-09-22 19:13:48,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=83458.66666666667, ans=0.125
2024-09-22 19:14:04,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=83505.33333333333, ans=0.0
2024-09-22 19:14:09,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=83505.33333333333, ans=0.2
2024-09-22 19:14:34,650 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.515e+02 1.751e+02 2.046e+02 3.052e+02, threshold=3.503e+02, percent-clipped=0.0
2024-09-22 19:14:35,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=83598.66666666667, ans=0.125
2024-09-22 19:14:41,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=83598.66666666667, ans=0.1
2024-09-22 19:15:05,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=83645.33333333333, ans=0.0
2024-09-22 19:15:07,862 INFO [train.py:1198] (1/4) Epoch 5, batch 2350, loss[loss=0.3423, ctc_loss=0.2534, cr_loss=0.4445, over 16431.00 frames. ], tot_loss[loss=0.2957, ctc_loss=0.2149, cr_loss=0.4038, over 3374901.53 frames. ], batch size: 66, lr: 2.27e-02, grad_scale: 32.0
2024-09-22 19:16:30,649 INFO [train.py:1198] (1/4) Epoch 5, batch 2400, loss[loss=0.2614, ctc_loss=0.1909, cr_loss=0.3528, over 17033.00 frames. ], tot_loss[loss=0.2971, ctc_loss=0.2163, cr_loss=0.4038, over 3369518.79 frames. ], batch size: 39, lr: 2.27e-02, grad_scale: 32.0
2024-09-22 19:16:34,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=83925.33333333333, ans=0.0
2024-09-22 19:16:37,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=83925.33333333333, ans=0.0
2024-09-22 19:17:07,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.36 vs. limit=10.0
2024-09-22 19:17:15,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer_ff3.min_abs, batch_count=84018.66666666667, ans=0.2
2024-09-22 19:17:19,651 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.495e+02 1.658e+02 1.967e+02 2.763e+02, threshold=3.315e+02, percent-clipped=0.0
2024-09-22 19:17:39,122 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.47 vs. limit=15.0
2024-09-22 19:17:49,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.min_positive, batch_count=84112.0, ans=0.05
2024-09-22 19:17:55,426 INFO [train.py:1198] (1/4) Epoch 5, batch 2450, loss[loss=0.2934, ctc_loss=0.2128, cr_loss=0.4028, over 17065.00 frames. ], tot_loss[loss=0.2955, ctc_loss=0.215, cr_loss=0.4028, over 3374379.26 frames. ], batch size: 52, lr: 2.26e-02, grad_scale: 32.0
2024-09-22 19:18:05,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=84158.66666666667, ans=0.125
2024-09-22 19:18:42,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=84298.66666666667, ans=0.125
2024-09-22 19:18:56,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=84298.66666666667, ans=0.2
2024-09-22 19:19:15,415 INFO [train.py:1198] (1/4) Epoch 5, batch 2500, loss[loss=0.3133, ctc_loss=0.2298, cr_loss=0.4176, over 17006.00 frames. ], tot_loss[loss=0.2963, ctc_loss=0.2156, cr_loss=0.4037, over 3367671.31 frames. ], batch size: 53, lr: 2.26e-02, grad_scale: 32.0
2024-09-22 19:19:41,359 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=5.17 vs. limit=12.0
2024-09-22 19:19:58,661 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 19:20:00,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=84485.33333333333, ans=0.1
2024-09-22 19:20:04,525 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 1.488e+02 1.690e+02 1.912e+02 3.424e+02, threshold=3.381e+02, percent-clipped=1.0
2024-09-22 19:20:17,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=84532.0, ans=0.125
2024-09-22 19:20:24,242 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.14 vs. limit=10.0
2024-09-22 19:20:33,487 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.41 vs. limit=15.0
2024-09-22 19:20:34,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=84578.66666666667, ans=0.125
2024-09-22 19:20:36,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=84578.66666666667, ans=0.125
2024-09-22 19:20:40,872 INFO [train.py:1198] (1/4) Epoch 5, batch 2550, loss[loss=0.3091, ctc_loss=0.2272, cr_loss=0.4093, over 16997.00 frames. ], tot_loss[loss=0.2979, ctc_loss=0.217, cr_loss=0.4046, over 3359726.36 frames. ], batch size: 53, lr: 2.26e-02, grad_scale: 32.0
2024-09-22 19:21:36,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=84765.33333333333, ans=0.1
2024-09-22 19:21:56,253 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.06 vs. limit=15.0
2024-09-22 19:22:03,009 INFO [train.py:1198] (1/4) Epoch 5, batch 2600, loss[loss=0.2937, ctc_loss=0.2147, cr_loss=0.395, over 17226.00 frames. ], tot_loss[loss=0.2963, ctc_loss=0.2158, cr_loss=0.4026, over 3353984.15 frames. ], batch size: 50, lr: 2.25e-02, grad_scale: 32.0
2024-09-22 19:22:03,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=84858.66666666667, ans=0.05
2024-09-22 19:22:24,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.54 vs. limit=15.0
2024-09-22 19:22:29,893 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=9.62 vs. limit=15.0
2024-09-22 19:22:51,550 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.583e+02 1.835e+02 2.132e+02 3.981e+02, threshold=3.669e+02, percent-clipped=1.0
2024-09-22 19:23:24,864 INFO [train.py:1198] (1/4) Epoch 5, batch 2650, loss[loss=0.3035, ctc_loss=0.2171, cr_loss=0.4318, over 17061.00 frames. ], tot_loss[loss=0.2973, ctc_loss=0.2164, cr_loss=0.4044, over 3352802.34 frames. ], batch size: 46, lr: 2.25e-02, grad_scale: 32.0
2024-09-22 19:23:25,745 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.93 vs. limit=15.0
2024-09-22 19:23:36,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=85092.0, ans=0.2
2024-09-22 19:23:41,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=85138.66666666667, ans=0.125
2024-09-22 19:23:47,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=85138.66666666667, ans=0.125
2024-09-22 19:23:50,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=85138.66666666667, ans=0.0
2024-09-22 19:23:58,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=85185.33333333333, ans=0.2
2024-09-22 19:24:45,647 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=2.92 vs. limit=15.0
2024-09-22 19:24:46,094 INFO [train.py:1198] (1/4) Epoch 5, batch 2700, loss[loss=0.2905, ctc_loss=0.213, cr_loss=0.3876, over 17297.00 frames. ], tot_loss[loss=0.2977, ctc_loss=0.2168, cr_loss=0.4046, over 3353245.40 frames. ], batch size: 49, lr: 2.25e-02, grad_scale: 32.0
2024-09-22 19:24:56,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=85325.33333333333, ans=0.1
2024-09-22 19:25:34,610 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.500e+02 1.639e+02 1.809e+02 2.394e+02, threshold=3.278e+02, percent-clipped=0.0
2024-09-22 19:26:06,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=85558.66666666667, ans=0.025
2024-09-22 19:26:08,131 INFO [train.py:1198] (1/4) Epoch 5, batch 2750, loss[loss=0.2462, ctc_loss=0.1749, cr_loss=0.3563, over 17121.00 frames. ], tot_loss[loss=0.2977, ctc_loss=0.2165, cr_loss=0.406, over 3365115.24 frames. ], batch size: 40, lr: 2.25e-02, grad_scale: 32.0
2024-09-22 19:26:22,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=85605.33333333333, ans=0.025
2024-09-22 19:26:33,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=85605.33333333333, ans=0.0
2024-09-22 19:26:59,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=85698.66666666667, ans=0.125
2024-09-22 19:27:04,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=85698.66666666667, ans=0.125
2024-09-22 19:27:31,979 INFO [train.py:1198] (1/4) Epoch 5, batch 2800, loss[loss=0.3299, ctc_loss=0.242, cr_loss=0.4396, over 15074.00 frames. ], tot_loss[loss=0.2974, ctc_loss=0.2162, cr_loss=0.4056, over 3365583.46 frames. ], batch size: 89, lr: 2.24e-02, grad_scale: 32.0
2024-09-22 19:27:34,879 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.71 vs. limit=15.0
2024-09-22 19:27:41,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=85792.0, ans=0.0
2024-09-22 19:27:46,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=85838.66666666667, ans=0.125
2024-09-22 19:28:04,938 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.98 vs. limit=15.0
2024-09-22 19:28:18,241 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.484e+02 1.665e+02 1.912e+02 3.153e+02, threshold=3.329e+02, percent-clipped=0.0
2024-09-22 19:28:25,478 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.16 vs. limit=15.0
2024-09-22 19:28:51,698 INFO [train.py:1198] (1/4) Epoch 5, batch 2850, loss[loss=0.3658, ctc_loss=0.2749, cr_loss=0.4542, over 14841.00 frames. ], tot_loss[loss=0.2971, ctc_loss=0.2162, cr_loss=0.4049, over 3364062.36 frames. ], batch size: 89, lr: 2.24e-02, grad_scale: 32.0
2024-09-22 19:29:35,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=86118.66666666667, ans=0.125
2024-09-22 19:30:16,018 INFO [train.py:1198] (1/4) Epoch 5, batch 2900, loss[loss=0.3242, ctc_loss=0.2361, cr_loss=0.4404, over 17145.00 frames. ], tot_loss[loss=0.2968, ctc_loss=0.2158, cr_loss=0.405, over 3359668.37 frames. ], batch size: 48, lr: 2.24e-02, grad_scale: 32.0
2024-09-22 19:30:16,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=86258.66666666667, ans=0.025
2024-09-22 19:30:17,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=86258.66666666667, ans=0.125
2024-09-22 19:30:57,949 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.76 vs. limit=12.0
2024-09-22 19:31:01,933 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.588e+02 1.741e+02 2.202e+02 4.410e+02, threshold=3.483e+02, percent-clipped=1.0
2024-09-22 19:31:07,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=86398.66666666667, ans=0.125
2024-09-22 19:31:35,099 INFO [train.py:1198] (1/4) Epoch 5, batch 2950, loss[loss=0.3838, ctc_loss=0.2952, cr_loss=0.4429, over 11816.00 frames. ], tot_loss[loss=0.2981, ctc_loss=0.217, cr_loss=0.4055, over 3346067.37 frames. ], batch size: 123, lr: 2.24e-02, grad_scale: 32.0
2024-09-22 19:31:36,187 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.28 vs. limit=15.0
2024-09-22 19:31:37,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=86492.0, ans=0.0
2024-09-22 19:31:42,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=86492.0, ans=0.0
2024-09-22 19:31:58,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=86538.66666666667, ans=0.125
2024-09-22 19:32:21,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=86585.33333333333, ans=0.0
2024-09-22 19:32:40,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=86632.0, ans=0.1
2024-09-22 19:32:50,423 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=7.69 vs. limit=15.0
2024-09-22 19:32:59,037 INFO [train.py:1198] (1/4) Epoch 5, batch 3000, loss[loss=0.2572, ctc_loss=0.1835, cr_loss=0.3683, over 17279.00 frames. ], tot_loss[loss=0.297, ctc_loss=0.2161, cr_loss=0.4044, over 3353653.77 frames. ], batch size: 42, lr: 2.23e-02, grad_scale: 32.0
2024-09-22 19:32:59,037 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 19:33:14,595 INFO [train.py:1230] (1/4) Epoch 5, validation: loss=0.06642, ctc_loss=0.06642, cr_loss=7.381e-15, over 944034.00 frames. 
2024-09-22 19:33:14,596 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 19:33:27,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=86725.33333333333, ans=0.5
2024-09-22 19:33:30,681 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=10.14 vs. limit=15.0
2024-09-22 19:33:36,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=86772.0, ans=0.125
2024-09-22 19:33:41,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=86772.0, ans=0.125
2024-09-22 19:33:58,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=86818.66666666667, ans=0.125
2024-09-22 19:34:00,034 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.524e+02 1.804e+02 2.236e+02 6.139e+02, threshold=3.607e+02, percent-clipped=4.0
2024-09-22 19:34:22,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=86912.0, ans=0.2
2024-09-22 19:34:35,392 INFO [train.py:1198] (1/4) Epoch 5, batch 3050, loss[loss=0.267, ctc_loss=0.194, cr_loss=0.3647, over 17287.00 frames. ], tot_loss[loss=0.2975, ctc_loss=0.2166, cr_loss=0.4047, over 3350041.07 frames. ], batch size: 42, lr: 2.23e-02, grad_scale: 32.0
2024-09-22 19:34:48,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=86958.66666666667, ans=0.125
2024-09-22 19:34:49,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=87005.33333333333, ans=0.2
2024-09-22 19:35:06,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=87052.0, ans=0.125
2024-09-22 19:35:53,027 INFO [train.py:1198] (1/4) Epoch 5, batch 3100, loss[loss=0.344, ctc_loss=0.2547, cr_loss=0.4464, over 16814.00 frames. ], tot_loss[loss=0.2968, ctc_loss=0.216, cr_loss=0.4042, over 3354293.07 frames. ], batch size: 61, lr: 2.23e-02, grad_scale: 64.0
2024-09-22 19:36:01,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=87192.0, ans=0.0
2024-09-22 19:36:04,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=87192.0, ans=0.2
2024-09-22 19:36:05,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=87192.0, ans=10.0
2024-09-22 19:36:08,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=87238.66666666667, ans=0.1
2024-09-22 19:36:10,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=87238.66666666667, ans=10.0
2024-09-22 19:36:13,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=87238.66666666667, ans=0.125
2024-09-22 19:36:37,948 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.555e+02 1.774e+02 2.101e+02 3.567e+02, threshold=3.549e+02, percent-clipped=0.0
2024-09-22 19:36:42,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=87332.0, ans=0.0
2024-09-22 19:36:54,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=87332.0, ans=0.0
2024-09-22 19:37:13,091 INFO [train.py:1198] (1/4) Epoch 5, batch 3150, loss[loss=0.3046, ctc_loss=0.2211, cr_loss=0.4174, over 17289.00 frames. ], tot_loss[loss=0.2978, ctc_loss=0.2168, cr_loss=0.4052, over 3347919.35 frames. ], batch size: 49, lr: 2.23e-02, grad_scale: 64.0
2024-09-22 19:37:40,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.01 vs. limit=15.0
2024-09-22 19:38:01,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=87565.33333333333, ans=0.0
2024-09-22 19:38:06,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=87565.33333333333, ans=0.2
2024-09-22 19:38:31,402 INFO [train.py:1198] (1/4) Epoch 5, batch 3200, loss[loss=0.2791, ctc_loss=0.2018, cr_loss=0.3864, over 17253.00 frames. ], tot_loss[loss=0.2976, ctc_loss=0.2165, cr_loss=0.4053, over 3355551.53 frames. ], batch size: 44, lr: 2.22e-02, grad_scale: 32.0
2024-09-22 19:38:31,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=87658.66666666667, ans=0.025
2024-09-22 19:38:44,665 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.44 vs. limit=10.0
2024-09-22 19:38:53,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=87705.33333333333, ans=0.125
2024-09-22 19:38:56,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=87705.33333333333, ans=0.1
2024-09-22 19:39:18,460 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.520e+02 1.734e+02 1.974e+02 3.517e+02, threshold=3.467e+02, percent-clipped=0.0
2024-09-22 19:39:31,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=87798.66666666667, ans=0.025
2024-09-22 19:39:49,578 INFO [train.py:1198] (1/4) Epoch 5, batch 3250, loss[loss=0.3659, ctc_loss=0.2724, cr_loss=0.4675, over 14804.00 frames. ], tot_loss[loss=0.2972, ctc_loss=0.2162, cr_loss=0.4049, over 3352634.18 frames. ], batch size: 89, lr: 2.22e-02, grad_scale: 32.0
2024-09-22 19:39:49,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=87892.0, ans=0.0
2024-09-22 19:40:30,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=87985.33333333333, ans=0.125
2024-09-22 19:40:46,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=88032.0, ans=0.1
2024-09-22 19:41:08,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=88125.33333333333, ans=0.05
2024-09-22 19:41:09,579 INFO [train.py:1198] (1/4) Epoch 5, batch 3300, loss[loss=0.2824, ctc_loss=0.206, cr_loss=0.3821, over 17228.00 frames. ], tot_loss[loss=0.2958, ctc_loss=0.2151, cr_loss=0.4035, over 3356519.48 frames. ], batch size: 50, lr: 2.22e-02, grad_scale: 32.0
2024-09-22 19:41:33,723 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=24.32 vs. limit=22.5
2024-09-22 19:41:41,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=88218.66666666667, ans=0.125
2024-09-22 19:41:58,205 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.544e+02 1.772e+02 2.234e+02 4.094e+02, threshold=3.543e+02, percent-clipped=4.0
2024-09-22 19:42:03,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=88265.33333333333, ans=0.125
2024-09-22 19:42:15,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=88312.0, ans=0.125
2024-09-22 19:42:18,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=88312.0, ans=0.0
2024-09-22 19:42:26,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=88312.0, ans=0.1
2024-09-22 19:42:29,376 INFO [train.py:1198] (1/4) Epoch 5, batch 3350, loss[loss=0.3457, ctc_loss=0.254, cr_loss=0.4585, over 16925.00 frames. ], tot_loss[loss=0.2959, ctc_loss=0.2152, cr_loss=0.4037, over 3361673.02 frames. ], batch size: 58, lr: 2.22e-02, grad_scale: 32.0
2024-09-22 19:42:32,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=88358.66666666667, ans=0.125
2024-09-22 19:42:37,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=88358.66666666667, ans=0.025
2024-09-22 19:43:02,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=88452.0, ans=0.1
2024-09-22 19:43:11,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=88452.0, ans=0.125
2024-09-22 19:43:36,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=88545.33333333333, ans=0.05
2024-09-22 19:43:41,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=88545.33333333333, ans=0.025
2024-09-22 19:43:47,476 INFO [train.py:1198] (1/4) Epoch 5, batch 3400, loss[loss=0.3123, ctc_loss=0.2311, cr_loss=0.4058, over 17025.00 frames. ], tot_loss[loss=0.2979, ctc_loss=0.2166, cr_loss=0.4061, over 3364396.38 frames. ], batch size: 56, lr: 2.21e-02, grad_scale: 32.0
2024-09-22 19:43:47,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=88592.0, ans=0.125
2024-09-22 19:44:23,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=88685.33333333333, ans=0.125
2024-09-22 19:44:30,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=88685.33333333333, ans=0.0
2024-09-22 19:44:34,720 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.562e+02 1.776e+02 2.153e+02 3.268e+02, threshold=3.552e+02, percent-clipped=0.0
2024-09-22 19:44:36,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=88732.0, ans=0.0
2024-09-22 19:44:38,487 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.46 vs. limit=15.0
2024-09-22 19:44:50,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=88778.66666666667, ans=0.07
2024-09-22 19:44:52,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=15.23 vs. limit=22.5
2024-09-22 19:45:05,685 INFO [train.py:1198] (1/4) Epoch 5, batch 3450, loss[loss=0.3107, ctc_loss=0.2214, cr_loss=0.4462, over 17235.00 frames. ], tot_loss[loss=0.2989, ctc_loss=0.2174, cr_loss=0.4074, over 3360554.91 frames. ], batch size: 50, lr: 2.21e-02, grad_scale: 32.0
2024-09-22 19:45:13,079 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.75 vs. limit=15.0
2024-09-22 19:45:45,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=88918.66666666667, ans=0.035
2024-09-22 19:45:50,848 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=7.85 vs. limit=15.0
2024-09-22 19:45:53,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=88965.33333333333, ans=0.125
2024-09-22 19:45:58,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=88965.33333333333, ans=0.1
2024-09-22 19:46:25,705 INFO [train.py:1198] (1/4) Epoch 5, batch 3500, loss[loss=0.3086, ctc_loss=0.2267, cr_loss=0.4094, over 17223.00 frames. ], tot_loss[loss=0.2994, ctc_loss=0.2178, cr_loss=0.408, over 3357080.04 frames. ], batch size: 47, lr: 2.21e-02, grad_scale: 32.0
2024-09-22 19:46:49,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=89105.33333333333, ans=0.0
2024-09-22 19:47:08,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=89152.0, ans=0.0
2024-09-22 19:47:14,315 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.466e+02 1.595e+02 1.829e+02 3.245e+02, threshold=3.189e+02, percent-clipped=0.0
2024-09-22 19:47:19,888 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.95 vs. limit=22.5
2024-09-22 19:47:29,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=89245.33333333333, ans=0.1
2024-09-22 19:47:30,370 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.83 vs. limit=12.0
2024-09-22 19:47:45,031 INFO [train.py:1198] (1/4) Epoch 5, batch 3550, loss[loss=0.2959, ctc_loss=0.2153, cr_loss=0.4029, over 17288.00 frames. ], tot_loss[loss=0.298, ctc_loss=0.2166, cr_loss=0.4067, over 3347999.89 frames. ], batch size: 49, lr: 2.21e-02, grad_scale: 32.0
2024-09-22 19:47:49,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=89292.0, ans=0.125
2024-09-22 19:48:04,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=89338.66666666667, ans=0.2
2024-09-22 19:48:08,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=89338.66666666667, ans=0.1
2024-09-22 19:48:28,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=89385.33333333333, ans=0.07
2024-09-22 19:48:30,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=89432.0, ans=0.1
2024-09-22 19:48:30,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=89432.0, ans=0.0
2024-09-22 19:48:36,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=89432.0, ans=0.025
2024-09-22 19:48:42,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=89432.0, ans=0.0
2024-09-22 19:48:47,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=89478.66666666667, ans=0.125
2024-09-22 19:49:00,250 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.66 vs. limit=10.0
2024-09-22 19:49:02,263 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=11.75 vs. limit=15.0
2024-09-22 19:49:02,836 INFO [train.py:1198] (1/4) Epoch 5, batch 3600, loss[loss=0.2613, ctc_loss=0.1859, cr_loss=0.3769, over 17050.00 frames. ], tot_loss[loss=0.296, ctc_loss=0.215, cr_loss=0.4048, over 3351556.24 frames. ], batch size: 39, lr: 2.20e-02, grad_scale: 32.0
2024-09-22 19:49:17,026 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 19:49:21,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=89572.0, ans=0.2
2024-09-22 19:49:24,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=89572.0, ans=0.1
2024-09-22 19:49:36,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=89618.66666666667, ans=0.1
2024-09-22 19:49:46,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=89618.66666666667, ans=0.1
2024-09-22 19:49:49,384 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.445e+02 1.594e+02 1.731e+02 2.971e+02, threshold=3.187e+02, percent-clipped=0.0
2024-09-22 19:50:09,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=89712.0, ans=0.1
2024-09-22 19:50:22,563 INFO [train.py:1198] (1/4) Epoch 5, batch 3650, loss[loss=0.2684, ctc_loss=0.1928, cr_loss=0.378, over 17090.00 frames. ], tot_loss[loss=0.295, ctc_loss=0.2142, cr_loss=0.4042, over 3356519.17 frames. ], batch size: 43, lr: 2.20e-02, grad_scale: 32.0
2024-09-22 19:50:23,698 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.61 vs. limit=22.5
2024-09-22 19:50:29,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=89758.66666666667, ans=0.1
2024-09-22 19:51:06,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=89852.0, ans=0.125
2024-09-22 19:51:35,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=89945.33333333333, ans=0.125
2024-09-22 19:51:41,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=89992.0, ans=0.125
2024-09-22 19:51:43,156 INFO [train.py:1198] (1/4) Epoch 5, batch 3700, loss[loss=0.2487, ctc_loss=0.1771, cr_loss=0.3582, over 17033.00 frames. ], tot_loss[loss=0.2956, ctc_loss=0.2147, cr_loss=0.4047, over 3363181.44 frames. ], batch size: 39, lr: 2.20e-02, grad_scale: 32.0
2024-09-22 19:51:51,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=89992.0, ans=0.125
2024-09-22 19:52:29,540 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.562e+02 1.758e+02 2.028e+02 3.638e+02, threshold=3.517e+02, percent-clipped=1.0
2024-09-22 19:52:40,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=90132.0, ans=0.125
2024-09-22 19:52:42,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=90132.0, ans=0.5
2024-09-22 19:52:54,277 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=18.79 vs. limit=22.5
2024-09-22 19:52:56,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=90178.66666666667, ans=0.2
2024-09-22 19:53:01,152 INFO [train.py:1198] (1/4) Epoch 5, batch 3750, loss[loss=0.3576, ctc_loss=0.2668, cr_loss=0.4537, over 14762.00 frames. ], tot_loss[loss=0.2965, ctc_loss=0.2156, cr_loss=0.4049, over 3348621.18 frames. ], batch size: 89, lr: 2.20e-02, grad_scale: 32.0
2024-09-22 19:53:12,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=90225.33333333333, ans=0.125
2024-09-22 19:53:25,991 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.33 vs. limit=15.0
2024-09-22 19:53:36,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=90318.66666666667, ans=0.025
2024-09-22 19:53:44,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=90318.66666666667, ans=0.1
2024-09-22 19:53:50,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=90365.33333333333, ans=0.125
2024-09-22 19:54:09,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=90412.0, ans=0.0
2024-09-22 19:54:19,782 INFO [train.py:1198] (1/4) Epoch 5, batch 3800, loss[loss=0.3241, ctc_loss=0.2362, cr_loss=0.4394, over 15833.00 frames. ], tot_loss[loss=0.297, ctc_loss=0.216, cr_loss=0.4049, over 3343356.72 frames. ], batch size: 74, lr: 2.19e-02, grad_scale: 32.0
2024-09-22 19:54:37,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=90505.33333333333, ans=0.125
2024-09-22 19:54:37,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=90505.33333333333, ans=0.2
2024-09-22 19:54:43,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=90505.33333333333, ans=0.07
2024-09-22 19:54:50,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=90552.0, ans=0.1
2024-09-22 19:54:53,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=90552.0, ans=0.0
2024-09-22 19:55:06,889 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.208e+02 1.578e+02 1.852e+02 2.152e+02 4.120e+02, threshold=3.704e+02, percent-clipped=2.0
2024-09-22 19:55:07,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=90598.66666666667, ans=0.0
2024-09-22 19:55:12,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=90598.66666666667, ans=10.0
2024-09-22 19:55:34,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=90645.33333333333, ans=0.125
2024-09-22 19:55:38,742 INFO [train.py:1198] (1/4) Epoch 5, batch 3850, loss[loss=0.2979, ctc_loss=0.2191, cr_loss=0.3943, over 17285.00 frames. ], tot_loss[loss=0.2992, ctc_loss=0.2179, cr_loss=0.4062, over 3324589.77 frames. ], batch size: 51, lr: 2.19e-02, grad_scale: 32.0
2024-09-22 19:55:42,557 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=16.60 vs. limit=15.0
2024-09-22 19:56:04,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=90738.66666666667, ans=0.025
2024-09-22 19:57:39,897 INFO [train.py:1198] (1/4) Epoch 6, batch 0, loss[loss=0.3271, ctc_loss=0.2484, cr_loss=0.3933, over 16985.00 frames. ], tot_loss[loss=0.3271, ctc_loss=0.2484, cr_loss=0.3933, over 16985.00 frames. ], batch size: 53, lr: 2.04e-02, grad_scale: 32.0
2024-09-22 19:57:39,897 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 19:57:55,110 INFO [train.py:1230] (1/4) Epoch 6, validation: loss=0.06886, ctc_loss=0.06886, cr_loss=9.986e-15, over 944034.00 frames. 
2024-09-22 19:57:55,111 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 19:58:13,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=90953.33333333333, ans=0.0
2024-09-22 19:58:18,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=90953.33333333333, ans=0.125
2024-09-22 19:58:40,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=91000.0, ans=0.125
2024-09-22 19:58:50,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=91046.66666666667, ans=0.125
2024-09-22 19:58:51,238 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 1.570e+02 1.853e+02 2.174e+02 4.194e+02, threshold=3.706e+02, percent-clipped=2.0
2024-09-22 19:58:54,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=91046.66666666667, ans=0.025
2024-09-22 19:59:08,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=91093.33333333333, ans=0.1
2024-09-22 19:59:10,500 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=6.63 vs. limit=15.0
2024-09-22 19:59:19,095 INFO [train.py:1198] (1/4) Epoch 6, batch 50, loss[loss=0.3047, ctc_loss=0.2205, cr_loss=0.4212, over 17134.00 frames. ], tot_loss[loss=0.295, ctc_loss=0.2135, cr_loss=0.4073, over 760524.59 frames. ], batch size: 48, lr: 2.04e-02, grad_scale: 32.0
2024-09-22 19:59:34,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=91140.0, ans=0.1
2024-09-22 19:59:59,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=91233.33333333333, ans=0.2
2024-09-22 20:00:18,439 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=6.10 vs. limit=15.0
2024-09-22 20:00:41,134 INFO [train.py:1198] (1/4) Epoch 6, batch 100, loss[loss=0.2754, ctc_loss=0.1992, cr_loss=0.3807, over 17098.00 frames. ], tot_loss[loss=0.2915, ctc_loss=0.2107, cr_loss=0.4036, over 1336194.06 frames. ], batch size: 40, lr: 2.04e-02, grad_scale: 32.0
2024-09-22 20:01:05,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=91420.0, ans=0.125
2024-09-22 20:01:05,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=91420.0, ans=10.0
2024-09-22 20:01:23,577 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.42 vs. limit=15.0
2024-09-22 20:01:27,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=91513.33333333333, ans=0.125
2024-09-22 20:01:35,327 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 1.399e+02 1.624e+02 1.941e+02 3.446e+02, threshold=3.247e+02, percent-clipped=0.0
2024-09-22 20:01:57,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=91560.0, ans=0.0
2024-09-22 20:02:00,845 INFO [train.py:1198] (1/4) Epoch 6, batch 150, loss[loss=0.2414, ctc_loss=0.1749, cr_loss=0.3327, over 16703.00 frames. ], tot_loss[loss=0.2929, ctc_loss=0.2122, cr_loss=0.4036, over 1778684.02 frames. ], batch size: 37, lr: 2.04e-02, grad_scale: 32.0
2024-09-22 20:02:09,715 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.77 vs. limit=15.0
2024-09-22 20:02:21,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn2.whiten.whitening_limit, batch_count=91653.33333333333, ans=22.5
2024-09-22 20:02:40,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=91700.0, ans=0.05
2024-09-22 20:02:46,088 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.02 vs. limit=15.0
2024-09-22 20:02:51,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=91746.66666666667, ans=0.1
2024-09-22 20:03:04,393 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.40 vs. limit=22.5
2024-09-22 20:03:21,941 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.08 vs. limit=22.5
2024-09-22 20:03:23,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=91793.33333333333, ans=0.0
2024-09-22 20:03:25,936 INFO [train.py:1198] (1/4) Epoch 6, batch 200, loss[loss=0.3092, ctc_loss=0.2304, cr_loss=0.3943, over 15890.00 frames. ], tot_loss[loss=0.2945, ctc_loss=0.2133, cr_loss=0.4061, over 2130164.40 frames. ], batch size: 74, lr: 2.03e-02, grad_scale: 32.0
2024-09-22 20:03:32,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=91840.0, ans=0.1
2024-09-22 20:03:36,649 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.75 vs. limit=22.5
2024-09-22 20:03:45,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=91886.66666666667, ans=0.1
2024-09-22 20:03:48,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer_ff2.min_abs, batch_count=91886.66666666667, ans=0.1
2024-09-22 20:03:52,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=91886.66666666667, ans=0.2
2024-09-22 20:03:57,617 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=8.69 vs. limit=15.0
2024-09-22 20:04:12,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=91933.33333333333, ans=0.2
2024-09-22 20:04:20,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=91980.0, ans=0.1
2024-09-22 20:04:21,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=91980.0, ans=0.05
2024-09-22 20:04:25,624 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 1.556e+02 1.826e+02 2.254e+02 3.362e+02, threshold=3.652e+02, percent-clipped=2.0
2024-09-22 20:04:27,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=91980.0, ans=0.125
2024-09-22 20:04:30,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=91980.0, ans=0.0
2024-09-22 20:04:37,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=92026.66666666667, ans=0.025
2024-09-22 20:04:51,203 INFO [train.py:1198] (1/4) Epoch 6, batch 250, loss[loss=0.2102, ctc_loss=0.146, cr_loss=0.321, over 17068.00 frames. ], tot_loss[loss=0.2925, ctc_loss=0.2116, cr_loss=0.4041, over 2396415.21 frames. ], batch size: 39, lr: 2.03e-02, grad_scale: 32.0
2024-09-22 20:05:05,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=92120.0, ans=0.125
2024-09-22 20:05:34,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=92166.66666666667, ans=0.025
2024-09-22 20:05:39,115 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=15.23 vs. limit=15.0
2024-09-22 20:05:39,235 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.14 vs. limit=10.0
2024-09-22 20:05:40,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=92213.33333333333, ans=0.5
2024-09-22 20:05:41,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=92213.33333333333, ans=0.1
2024-09-22 20:05:49,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=92213.33333333333, ans=0.2
2024-09-22 20:06:02,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=92260.0, ans=0.0
2024-09-22 20:06:04,562 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.04 vs. limit=22.5
2024-09-22 20:06:10,521 INFO [train.py:1198] (1/4) Epoch 6, batch 300, loss[loss=0.2975, ctc_loss=0.2179, cr_loss=0.3977, over 16748.00 frames. ], tot_loss[loss=0.2931, ctc_loss=0.2123, cr_loss=0.4037, over 2600464.34 frames. ], batch size: 61, lr: 2.03e-02, grad_scale: 32.0
2024-09-22 20:06:37,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=92353.33333333333, ans=0.0
2024-09-22 20:07:04,351 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.501e+02 1.679e+02 1.995e+02 3.588e+02, threshold=3.358e+02, percent-clipped=0.0
2024-09-22 20:07:18,134 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.32 vs. limit=10.0
2024-09-22 20:07:32,318 INFO [train.py:1198] (1/4) Epoch 6, batch 350, loss[loss=0.2558, ctc_loss=0.1772, cr_loss=0.393, over 17313.00 frames. ], tot_loss[loss=0.2924, ctc_loss=0.2117, cr_loss=0.4036, over 2773457.06 frames. ], batch size: 46, lr: 2.03e-02, grad_scale: 32.0
2024-09-22 20:07:39,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=92540.0, ans=0.1
2024-09-22 20:08:10,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=92633.33333333333, ans=0.0
2024-09-22 20:08:46,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=92726.66666666667, ans=0.0
2024-09-22 20:08:55,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=92773.33333333333, ans=0.0
2024-09-22 20:08:57,315 INFO [train.py:1198] (1/4) Epoch 6, batch 400, loss[loss=0.2196, ctc_loss=0.151, cr_loss=0.3434, over 17036.00 frames. ], tot_loss[loss=0.2931, ctc_loss=0.2122, cr_loss=0.4042, over 2896517.61 frames. ], batch size: 39, lr: 2.02e-02, grad_scale: 32.0
2024-09-22 20:09:12,717 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.37 vs. limit=6.0
2024-09-22 20:09:15,730 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.32 vs. limit=22.5
2024-09-22 20:09:33,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=92866.66666666667, ans=0.125
2024-09-22 20:09:51,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=92913.33333333333, ans=0.025
2024-09-22 20:09:54,060 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.481e+02 1.649e+02 1.890e+02 2.985e+02, threshold=3.299e+02, percent-clipped=0.0
2024-09-22 20:10:19,708 INFO [train.py:1198] (1/4) Epoch 6, batch 450, loss[loss=0.2831, ctc_loss=0.2002, cr_loss=0.4145, over 17059.00 frames. ], tot_loss[loss=0.2944, ctc_loss=0.2131, cr_loss=0.4064, over 2992094.16 frames. ], batch size: 46, lr: 2.02e-02, grad_scale: 32.0
2024-09-22 20:10:23,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=93006.66666666667, ans=0.0
2024-09-22 20:10:26,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=93006.66666666667, ans=0.0
2024-09-22 20:10:53,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=93100.0, ans=0.95
2024-09-22 20:11:13,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=93146.66666666667, ans=0.0
2024-09-22 20:11:20,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=93146.66666666667, ans=0.125
2024-09-22 20:11:29,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=93193.33333333333, ans=0.0
2024-09-22 20:11:39,001 INFO [train.py:1198] (1/4) Epoch 6, batch 500, loss[loss=0.2681, ctc_loss=0.1901, cr_loss=0.3901, over 17064.00 frames. ], tot_loss[loss=0.2946, ctc_loss=0.2131, cr_loss=0.4076, over 3076911.01 frames. ], batch size: 46, lr: 2.02e-02, grad_scale: 32.0
2024-09-22 20:11:39,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=93240.0, ans=0.125
2024-09-22 20:11:48,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=93240.0, ans=0.0
2024-09-22 20:11:51,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=93240.0, ans=0.0
2024-09-22 20:12:29,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=93380.0, ans=0.125
2024-09-22 20:12:37,312 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.459e+02 1.650e+02 1.980e+02 2.967e+02, threshold=3.301e+02, percent-clipped=0.0
2024-09-22 20:12:59,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=93426.66666666667, ans=0.125
2024-09-22 20:13:01,017 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.00 vs. limit=12.0
2024-09-22 20:13:05,202 INFO [train.py:1198] (1/4) Epoch 6, batch 550, loss[loss=0.3189, ctc_loss=0.2329, cr_loss=0.4302, over 16955.00 frames. ], tot_loss[loss=0.2925, ctc_loss=0.2114, cr_loss=0.4054, over 3144265.55 frames. ], batch size: 58, lr: 2.02e-02, grad_scale: 32.0
2024-09-22 20:13:22,401 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.74 vs. limit=6.0
2024-09-22 20:13:27,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=93520.0, ans=0.0
2024-09-22 20:13:29,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=93520.0, ans=0.0
2024-09-22 20:13:48,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=93566.66666666667, ans=0.95
2024-09-22 20:14:30,621 INFO [train.py:1198] (1/4) Epoch 6, batch 600, loss[loss=0.2658, ctc_loss=0.1928, cr_loss=0.3649, over 17041.00 frames. ], tot_loss[loss=0.2922, ctc_loss=0.2113, cr_loss=0.4046, over 3192800.17 frames. ], batch size: 39, lr: 2.02e-02, grad_scale: 32.0
2024-09-22 20:14:32,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=93706.66666666667, ans=0.0
2024-09-22 20:14:49,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=93753.33333333333, ans=0.0
2024-09-22 20:15:20,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=93846.66666666667, ans=0.0
2024-09-22 20:15:24,958 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.461e+02 1.595e+02 1.846e+02 3.407e+02, threshold=3.191e+02, percent-clipped=1.0
2024-09-22 20:15:46,767 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.55 vs. limit=15.0
2024-09-22 20:15:49,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=93940.0, ans=0.1
2024-09-22 20:15:50,710 INFO [train.py:1198] (1/4) Epoch 6, batch 650, loss[loss=0.2805, ctc_loss=0.2064, cr_loss=0.3709, over 17019.00 frames. ], tot_loss[loss=0.2917, ctc_loss=0.2109, cr_loss=0.4039, over 3228854.92 frames. ], batch size: 44, lr: 2.01e-02, grad_scale: 32.0
2024-09-22 20:16:26,058 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.10 vs. limit=15.0
2024-09-22 20:16:27,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=94033.33333333333, ans=0.1
2024-09-22 20:17:09,771 INFO [train.py:1198] (1/4) Epoch 6, batch 700, loss[loss=0.3096, ctc_loss=0.2217, cr_loss=0.4395, over 17020.00 frames. ], tot_loss[loss=0.2916, ctc_loss=0.2107, cr_loss=0.4043, over 3255828.59 frames. ], batch size: 52, lr: 2.01e-02, grad_scale: 32.0
2024-09-22 20:17:56,765 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 20:18:09,123 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.435e+02 1.629e+02 1.890e+02 2.825e+02, threshold=3.258e+02, percent-clipped=0.0
2024-09-22 20:18:20,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=94360.0, ans=0.035
2024-09-22 20:18:28,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=94360.0, ans=0.2
2024-09-22 20:18:34,804 INFO [train.py:1198] (1/4) Epoch 6, batch 750, loss[loss=0.2478, ctc_loss=0.1786, cr_loss=0.3459, over 16958.00 frames. ], tot_loss[loss=0.2901, ctc_loss=0.2095, cr_loss=0.403, over 3284979.93 frames. ], batch size: 42, lr: 2.01e-02, grad_scale: 32.0
2024-09-22 20:18:54,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=94453.33333333333, ans=0.1
2024-09-22 20:18:58,745 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.92 vs. limit=22.5
2024-09-22 20:19:16,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=94500.0, ans=0.5
2024-09-22 20:19:21,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=94500.0, ans=0.0
2024-09-22 20:19:30,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=94546.66666666667, ans=0.0
2024-09-22 20:19:58,980 INFO [train.py:1198] (1/4) Epoch 6, batch 800, loss[loss=0.2603, ctc_loss=0.188, cr_loss=0.3613, over 17058.00 frames. ], tot_loss[loss=0.2901, ctc_loss=0.2097, cr_loss=0.4023, over 3295306.41 frames. ], batch size: 43, lr: 2.01e-02, grad_scale: 32.0
2024-09-22 20:20:10,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer_ff2.min_abs, batch_count=94640.0, ans=0.1
2024-09-22 20:20:18,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=94686.66666666667, ans=0.025
2024-09-22 20:20:19,638 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 20:20:37,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=94733.33333333333, ans=0.025
2024-09-22 20:20:40,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=94733.33333333333, ans=0.0
2024-09-22 20:20:41,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer_na.min_abs, batch_count=94733.33333333333, ans=0.02
2024-09-22 20:20:53,105 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.473e+02 1.596e+02 1.875e+02 3.402e+02, threshold=3.192e+02, percent-clipped=2.0
2024-09-22 20:21:06,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=94826.66666666667, ans=0.125
2024-09-22 20:21:18,626 INFO [train.py:1198] (1/4) Epoch 6, batch 850, loss[loss=0.287, ctc_loss=0.2055, cr_loss=0.4073, over 17326.00 frames. ], tot_loss[loss=0.2892, ctc_loss=0.2089, cr_loss=0.4016, over 3308417.77 frames. ], batch size: 51, lr: 2.00e-02, grad_scale: 32.0
2024-09-22 20:21:28,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=94873.33333333333, ans=0.125
2024-09-22 20:21:35,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=94920.0, ans=0.125
2024-09-22 20:21:36,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=94920.0, ans=0.0
2024-09-22 20:21:44,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=94920.0, ans=0.125
2024-09-22 20:21:46,570 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=9.03 vs. limit=15.0
2024-09-22 20:21:54,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=94966.66666666667, ans=0.07
2024-09-22 20:22:08,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=95013.33333333333, ans=0.125
2024-09-22 20:22:42,727 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 20:22:43,771 INFO [train.py:1198] (1/4) Epoch 6, batch 900, loss[loss=0.3589, ctc_loss=0.2749, cr_loss=0.4199, over 11945.00 frames. ], tot_loss[loss=0.2882, ctc_loss=0.2081, cr_loss=0.4005, over 3317697.79 frames. ], batch size: 123, lr: 2.00e-02, grad_scale: 32.0
2024-09-22 20:23:21,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=95200.0, ans=0.125
2024-09-22 20:23:24,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=95200.0, ans=0.125
2024-09-22 20:23:37,720 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.462e+02 1.630e+02 1.916e+02 2.984e+02, threshold=3.259e+02, percent-clipped=0.0
2024-09-22 20:23:42,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=95246.66666666667, ans=0.125
2024-09-22 20:23:51,917 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.13 vs. limit=15.0
2024-09-22 20:24:00,280 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.39 vs. limit=15.0
2024-09-22 20:24:05,929 INFO [train.py:1198] (1/4) Epoch 6, batch 950, loss[loss=0.3307, ctc_loss=0.2441, cr_loss=0.4332, over 17160.00 frames. ], tot_loss[loss=0.287, ctc_loss=0.2072, cr_loss=0.3988, over 3326285.40 frames. ], batch size: 48, lr: 2.00e-02, grad_scale: 32.0
2024-09-22 20:24:10,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=95340.0, ans=0.0
2024-09-22 20:24:29,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=95386.66666666667, ans=0.0
2024-09-22 20:25:28,488 INFO [train.py:1198] (1/4) Epoch 6, batch 1000, loss[loss=0.2697, ctc_loss=0.1929, cr_loss=0.3839, over 17169.00 frames. ], tot_loss[loss=0.2872, ctc_loss=0.2072, cr_loss=0.4002, over 3335506.23 frames. ], batch size: 45, lr: 2.00e-02, grad_scale: 32.0
2024-09-22 20:26:18,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=95713.33333333333, ans=0.09899494936611666
2024-09-22 20:26:22,800 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.403e+02 1.529e+02 1.831e+02 2.517e+02, threshold=3.058e+02, percent-clipped=0.0
2024-09-22 20:26:30,337 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.57 vs. limit=10.0
2024-09-22 20:26:43,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=95760.0, ans=0.0
2024-09-22 20:26:48,475 INFO [train.py:1198] (1/4) Epoch 6, batch 1050, loss[loss=0.2716, ctc_loss=0.1946, cr_loss=0.3852, over 17007.00 frames. ], tot_loss[loss=0.2874, ctc_loss=0.2071, cr_loss=0.4012, over 3346904.34 frames. ], batch size: 44, lr: 2.00e-02, grad_scale: 32.0
2024-09-22 20:27:17,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=95853.33333333333, ans=0.0
2024-09-22 20:27:43,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=95946.66666666667, ans=0.07
2024-09-22 20:27:59,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=95993.33333333333, ans=0.125
2024-09-22 20:28:13,098 INFO [train.py:1198] (1/4) Epoch 6, batch 1100, loss[loss=0.2502, ctc_loss=0.1806, cr_loss=0.3483, over 17080.00 frames. ], tot_loss[loss=0.2884, ctc_loss=0.208, cr_loss=0.4023, over 3350253.94 frames. ], batch size: 43, lr: 1.99e-02, grad_scale: 32.0
2024-09-22 20:28:16,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=96040.0, ans=0.0
2024-09-22 20:28:24,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=96040.0, ans=0.125
2024-09-22 20:28:30,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=96086.66666666667, ans=0.125
2024-09-22 20:28:48,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=96133.33333333333, ans=0.125
2024-09-22 20:28:59,126 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 20:28:59,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=96133.33333333333, ans=0.0
2024-09-22 20:29:11,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=96180.0, ans=0.0
2024-09-22 20:29:12,626 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.277e+02 1.448e+02 1.600e+02 1.818e+02 3.191e+02, threshold=3.201e+02, percent-clipped=3.0
2024-09-22 20:29:12,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=96180.0, ans=0.125
2024-09-22 20:29:14,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=96180.0, ans=0.1
2024-09-22 20:29:20,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=96226.66666666667, ans=0.0
2024-09-22 20:29:21,705 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=7.64 vs. limit=15.0
2024-09-22 20:29:33,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=96226.66666666667, ans=0.125
2024-09-22 20:29:38,029 INFO [train.py:1198] (1/4) Epoch 6, batch 1150, loss[loss=0.2959, ctc_loss=0.2194, cr_loss=0.3825, over 17074.00 frames. ], tot_loss[loss=0.2883, ctc_loss=0.2079, cr_loss=0.4024, over 3350635.14 frames. ], batch size: 46, lr: 1.99e-02, grad_scale: 32.0
2024-09-22 20:29:53,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=96320.0, ans=0.0
2024-09-22 20:30:09,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=96366.66666666667, ans=0.1
2024-09-22 20:30:12,127 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.95 vs. limit=22.5
2024-09-22 20:30:22,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=96366.66666666667, ans=0.2
2024-09-22 20:30:38,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=96413.33333333333, ans=0.125
2024-09-22 20:30:48,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=96460.0, ans=0.0
2024-09-22 20:30:57,273 INFO [train.py:1198] (1/4) Epoch 6, batch 1200, loss[loss=0.2787, ctc_loss=0.2001, cr_loss=0.3932, over 17077.00 frames. ], tot_loss[loss=0.288, ctc_loss=0.2077, cr_loss=0.4017, over 3348483.21 frames. ], batch size: 43, lr: 1.99e-02, grad_scale: 32.0
2024-09-22 20:31:02,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=96506.66666666667, ans=0.04949747468305833
2024-09-22 20:31:06,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.54 vs. limit=6.0
2024-09-22 20:31:18,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=96553.33333333333, ans=0.125
2024-09-22 20:31:18,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=96553.33333333333, ans=0.125
2024-09-22 20:31:33,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=96600.0, ans=0.0
2024-09-22 20:31:45,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=96646.66666666667, ans=0.125
2024-09-22 20:31:51,564 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.470e+02 1.636e+02 2.013e+02 4.309e+02, threshold=3.271e+02, percent-clipped=1.0
2024-09-22 20:31:54,062 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.60 vs. limit=6.0
2024-09-22 20:32:01,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_positive, batch_count=96693.33333333333, ans=0.05
2024-09-22 20:32:12,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=96693.33333333333, ans=0.0
2024-09-22 20:32:17,037 INFO [train.py:1198] (1/4) Epoch 6, batch 1250, loss[loss=0.3152, ctc_loss=0.231, cr_loss=0.4214, over 15157.00 frames. ], tot_loss[loss=0.2889, ctc_loss=0.2083, cr_loss=0.4031, over 3349760.82 frames. ], batch size: 89, lr: 1.99e-02, grad_scale: 32.0
2024-09-22 20:32:36,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=96786.66666666667, ans=0.0
2024-09-22 20:32:46,656 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=13.74 vs. limit=22.5
2024-09-22 20:32:51,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=96786.66666666667, ans=0.0
2024-09-22 20:33:12,396 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.39 vs. limit=6.0
2024-09-22 20:33:43,910 INFO [train.py:1198] (1/4) Epoch 6, batch 1300, loss[loss=0.266, ctc_loss=0.1913, cr_loss=0.3735, over 17061.00 frames. ], tot_loss[loss=0.2878, ctc_loss=0.2074, cr_loss=0.4019, over 3358797.64 frames. ], batch size: 46, lr: 1.99e-02, grad_scale: 32.0
2024-09-22 20:34:12,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=97020.0, ans=0.1
2024-09-22 20:34:42,526 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.454e+02 1.645e+02 1.943e+02 2.545e+02, threshold=3.291e+02, percent-clipped=0.0
2024-09-22 20:34:42,975 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 20:34:47,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=97113.33333333333, ans=0.1
2024-09-22 20:34:47,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=97113.33333333333, ans=0.025
2024-09-22 20:34:47,870 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.32 vs. limit=15.0
2024-09-22 20:34:52,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=97160.0, ans=0.125
2024-09-22 20:34:57,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=97160.0, ans=0.125
2024-09-22 20:35:02,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=97160.0, ans=0.125
2024-09-22 20:35:06,834 INFO [train.py:1198] (1/4) Epoch 6, batch 1350, loss[loss=0.2466, ctc_loss=0.1757, cr_loss=0.3543, over 16683.00 frames. ], tot_loss[loss=0.2879, ctc_loss=0.2076, cr_loss=0.4016, over 3354016.32 frames. ], batch size: 37, lr: 1.98e-02, grad_scale: 32.0
2024-09-22 20:35:40,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=97300.0, ans=0.125
2024-09-22 20:36:07,544 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=14.45 vs. limit=15.0
2024-09-22 20:36:17,048 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.91 vs. limit=15.0
2024-09-22 20:36:25,756 INFO [train.py:1198] (1/4) Epoch 6, batch 1400, loss[loss=0.269, ctc_loss=0.1871, cr_loss=0.4095, over 17031.00 frames. ], tot_loss[loss=0.2867, ctc_loss=0.2066, cr_loss=0.4002, over 3350071.67 frames. ], batch size: 44, lr: 1.98e-02, grad_scale: 32.0
2024-09-22 20:36:45,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=97486.66666666667, ans=0.2
2024-09-22 20:37:11,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=97533.33333333333, ans=0.125
2024-09-22 20:37:15,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=97580.0, ans=0.0
2024-09-22 20:37:24,530 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.419e+02 1.584e+02 1.972e+02 3.775e+02, threshold=3.168e+02, percent-clipped=1.0
2024-09-22 20:37:35,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=97626.66666666667, ans=0.2
2024-09-22 20:37:41,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.64 vs. limit=10.0
2024-09-22 20:37:43,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=97626.66666666667, ans=0.0
2024-09-22 20:37:48,278 INFO [train.py:1198] (1/4) Epoch 6, batch 1450, loss[loss=0.3143, ctc_loss=0.2251, cr_loss=0.4458, over 17001.00 frames. ], tot_loss[loss=0.287, ctc_loss=0.2071, cr_loss=0.3997, over 3341359.79 frames. ], batch size: 53, lr: 1.98e-02, grad_scale: 32.0
2024-09-22 20:37:48,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=97673.33333333333, ans=0.125
2024-09-22 20:38:17,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=97720.0, ans=0.0
2024-09-22 20:38:26,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=97766.66666666667, ans=0.125
2024-09-22 20:38:51,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=97813.33333333333, ans=0.125
2024-09-22 20:38:53,369 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.25 vs. limit=15.0
2024-09-22 20:39:12,832 INFO [train.py:1198] (1/4) Epoch 6, batch 1500, loss[loss=0.2997, ctc_loss=0.2188, cr_loss=0.4047, over 16999.00 frames. ], tot_loss[loss=0.2871, ctc_loss=0.2071, cr_loss=0.3998, over 3342207.44 frames. ], batch size: 53, lr: 1.98e-02, grad_scale: 32.0
2024-09-22 20:39:16,815 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.12 vs. limit=22.5
2024-09-22 20:39:27,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=97953.33333333333, ans=0.2
2024-09-22 20:39:29,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=97953.33333333333, ans=0.125
2024-09-22 20:40:09,174 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.478e+02 1.585e+02 1.791e+02 2.453e+02, threshold=3.170e+02, percent-clipped=0.0
2024-09-22 20:40:11,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=98046.66666666667, ans=0.5
2024-09-22 20:40:33,019 INFO [train.py:1198] (1/4) Epoch 6, batch 1550, loss[loss=0.2807, ctc_loss=0.1965, cr_loss=0.4208, over 17307.00 frames. ], tot_loss[loss=0.2862, ctc_loss=0.2064, cr_loss=0.3992, over 3349334.78 frames. ], batch size: 49, lr: 1.98e-02, grad_scale: 32.0
2024-09-22 20:40:39,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=98140.0, ans=0.0
2024-09-22 20:41:01,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=98186.66666666667, ans=0.125
2024-09-22 20:41:19,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=98280.0, ans=0.1
2024-09-22 20:41:27,782 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.25 vs. limit=15.0
2024-09-22 20:41:28,031 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.68 vs. limit=10.0
2024-09-22 20:41:39,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=98326.66666666667, ans=0.125
2024-09-22 20:41:51,963 INFO [train.py:1198] (1/4) Epoch 6, batch 1600, loss[loss=0.2574, ctc_loss=0.1845, cr_loss=0.3644, over 17100.00 frames. ], tot_loss[loss=0.2862, ctc_loss=0.2063, cr_loss=0.3998, over 3353139.73 frames. ], batch size: 43, lr: 1.97e-02, grad_scale: 32.0
2024-09-22 20:41:52,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=98373.33333333333, ans=0.0
2024-09-22 20:42:05,581 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.51 vs. limit=22.5
2024-09-22 20:42:14,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=98420.0, ans=0.125
2024-09-22 20:42:18,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=98420.0, ans=0.125
2024-09-22 20:42:21,834 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.98 vs. limit=15.0
2024-09-22 20:42:48,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=98513.33333333333, ans=0.1
2024-09-22 20:42:52,417 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.461e+02 1.657e+02 2.022e+02 3.350e+02, threshold=3.314e+02, percent-clipped=2.0
2024-09-22 20:42:59,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=98560.0, ans=0.2
2024-09-22 20:43:02,943 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.11 vs. limit=22.5
2024-09-22 20:43:08,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=98560.0, ans=0.025
2024-09-22 20:43:10,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=98560.0, ans=0.0
2024-09-22 20:43:10,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=98560.0, ans=0.1
2024-09-22 20:43:10,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=98560.0, ans=0.2
2024-09-22 20:43:16,606 INFO [train.py:1198] (1/4) Epoch 6, batch 1650, loss[loss=0.3101, ctc_loss=0.2207, cr_loss=0.4469, over 17302.00 frames. ], tot_loss[loss=0.2863, ctc_loss=0.2063, cr_loss=0.3996, over 3346523.69 frames. ], batch size: 49, lr: 1.97e-02, grad_scale: 32.0
2024-09-22 20:43:46,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=98653.33333333333, ans=0.95
2024-09-22 20:43:56,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=98700.0, ans=0.0
2024-09-22 20:44:08,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=98746.66666666667, ans=0.125
2024-09-22 20:44:23,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=98793.33333333333, ans=0.0
2024-09-22 20:44:32,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=98793.33333333333, ans=0.0
2024-09-22 20:44:40,254 INFO [train.py:1198] (1/4) Epoch 6, batch 1700, loss[loss=0.2306, ctc_loss=0.1608, cr_loss=0.349, over 17108.00 frames. ], tot_loss[loss=0.2874, ctc_loss=0.2072, cr_loss=0.4011, over 3346545.44 frames. ], batch size: 40, lr: 1.97e-02, grad_scale: 32.0
2024-09-22 20:44:46,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=98840.0, ans=0.125
2024-09-22 20:45:06,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=98886.66666666667, ans=0.0
2024-09-22 20:45:33,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=98980.0, ans=0.0
2024-09-22 20:45:36,068 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.362e+02 1.512e+02 1.797e+02 3.142e+02, threshold=3.023e+02, percent-clipped=0.0
2024-09-22 20:46:00,096 INFO [train.py:1198] (1/4) Epoch 6, batch 1750, loss[loss=0.3217, ctc_loss=0.2314, cr_loss=0.4519, over 16956.00 frames. ], tot_loss[loss=0.2882, ctc_loss=0.2079, cr_loss=0.4018, over 3351865.21 frames. ], batch size: 58, lr: 1.97e-02, grad_scale: 32.0
2024-09-22 20:46:16,827 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.16 vs. limit=15.0
2024-09-22 20:46:19,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=99120.0, ans=0.125
2024-09-22 20:46:33,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=99166.66666666667, ans=0.125
2024-09-22 20:46:44,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=99166.66666666667, ans=0.025
2024-09-22 20:46:45,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=99166.66666666667, ans=0.04949747468305833
2024-09-22 20:46:56,127 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 20:46:59,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=99213.33333333333, ans=0.025
2024-09-22 20:47:05,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=99260.0, ans=0.125
2024-09-22 20:47:23,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=99306.66666666667, ans=0.125
2024-09-22 20:47:24,698 INFO [train.py:1198] (1/4) Epoch 6, batch 1800, loss[loss=0.2913, ctc_loss=0.2132, cr_loss=0.3905, over 17203.00 frames. ], tot_loss[loss=0.2878, ctc_loss=0.2075, cr_loss=0.4015, over 3359265.16 frames. ], batch size: 47, lr: 1.96e-02, grad_scale: 32.0
2024-09-22 20:47:47,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=99353.33333333333, ans=0.0
2024-09-22 20:48:22,608 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 1.401e+02 1.510e+02 1.768e+02 2.829e+02, threshold=3.019e+02, percent-clipped=0.0
2024-09-22 20:48:43,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=99493.33333333333, ans=0.125
2024-09-22 20:48:43,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=99493.33333333333, ans=0.1
2024-09-22 20:48:45,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=99540.0, ans=0.125
2024-09-22 20:48:46,313 INFO [train.py:1198] (1/4) Epoch 6, batch 1850, loss[loss=0.2741, ctc_loss=0.1913, cr_loss=0.4141, over 17257.00 frames. ], tot_loss[loss=0.2872, ctc_loss=0.2069, cr_loss=0.4016, over 3361883.28 frames. ], batch size: 44, lr: 1.96e-02, grad_scale: 32.0
2024-09-22 20:49:05,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=99586.66666666667, ans=0.125
2024-09-22 20:49:08,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=99586.66666666667, ans=0.0
2024-09-22 20:49:08,876 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 20:49:27,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=99633.33333333333, ans=0.125
2024-09-22 20:49:42,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=99680.0, ans=0.025
2024-09-22 20:49:42,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=99680.0, ans=0.125
2024-09-22 20:50:09,094 INFO [train.py:1198] (1/4) Epoch 6, batch 1900, loss[loss=0.2841, ctc_loss=0.2026, cr_loss=0.4071, over 17267.00 frames. ], tot_loss[loss=0.2855, ctc_loss=0.2055, cr_loss=0.3998, over 3361935.38 frames. ], batch size: 44, lr: 1.96e-02, grad_scale: 32.0
2024-09-22 20:50:23,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=99820.0, ans=0.125
2024-09-22 20:51:05,355 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.450e+02 1.646e+02 2.008e+02 3.199e+02, threshold=3.291e+02, percent-clipped=1.0
2024-09-22 20:51:29,421 INFO [train.py:1198] (1/4) Epoch 6, batch 1950, loss[loss=0.2915, ctc_loss=0.21, cr_loss=0.4073, over 17098.00 frames. ], tot_loss[loss=0.2862, ctc_loss=0.206, cr_loss=0.401, over 3360977.16 frames. ], batch size: 49, lr: 1.96e-02, grad_scale: 32.0
2024-09-22 20:52:07,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=100100.0, ans=0.0
2024-09-22 20:52:23,962 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.50 vs. limit=22.5
2024-09-22 20:52:43,210 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.85 vs. limit=6.0
2024-09-22 20:52:53,473 INFO [train.py:1198] (1/4) Epoch 6, batch 2000, loss[loss=0.2408, ctc_loss=0.1705, cr_loss=0.3512, over 16720.00 frames. ], tot_loss[loss=0.2856, ctc_loss=0.2056, cr_loss=0.4001, over 3356792.68 frames. ], batch size: 37, lr: 1.96e-02, grad_scale: 32.0
2024-09-22 20:52:57,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=100240.0, ans=0.2
2024-09-22 20:53:05,598 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.31 vs. limit=10.0
2024-09-22 20:53:47,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=100380.0, ans=0.125
2024-09-22 20:53:50,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=100380.0, ans=0.125
2024-09-22 20:53:54,724 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 1.463e+02 1.652e+02 2.000e+02 4.397e+02, threshold=3.304e+02, percent-clipped=3.0
2024-09-22 20:54:17,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=100473.33333333333, ans=0.125
2024-09-22 20:54:18,663 INFO [train.py:1198] (1/4) Epoch 6, batch 2050, loss[loss=0.27, ctc_loss=0.1917, cr_loss=0.3918, over 17317.00 frames. ], tot_loss[loss=0.2859, ctc_loss=0.2058, cr_loss=0.4006, over 3358702.97 frames. ], batch size: 51, lr: 1.95e-02, grad_scale: 32.0
2024-09-22 20:54:27,380 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.72 vs. limit=12.0
2024-09-22 20:54:30,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=100473.33333333333, ans=0.125
2024-09-22 20:54:34,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=100520.0, ans=0.0
2024-09-22 20:54:36,806 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.43 vs. limit=15.0
2024-09-22 20:55:14,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=100613.33333333333, ans=0.125
2024-09-22 20:55:19,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=100613.33333333333, ans=0.125
2024-09-22 20:55:19,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=100613.33333333333, ans=0.125
2024-09-22 20:55:31,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=100660.0, ans=0.125
2024-09-22 20:55:33,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=100660.0, ans=0.2
2024-09-22 20:55:37,972 INFO [train.py:1198] (1/4) Epoch 6, batch 2100, loss[loss=0.2608, ctc_loss=0.1839, cr_loss=0.3845, over 17040.00 frames. ], tot_loss[loss=0.2877, ctc_loss=0.2072, cr_loss=0.4024, over 3348438.86 frames. ], batch size: 39, lr: 1.95e-02, grad_scale: 32.0
2024-09-22 20:55:41,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=100706.66666666667, ans=0.125
2024-09-22 20:55:47,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=100706.66666666667, ans=0.125
2024-09-22 20:55:55,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=100753.33333333333, ans=0.0
2024-09-22 20:56:26,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=100846.66666666667, ans=0.07
2024-09-22 20:56:30,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=100846.66666666667, ans=0.2
2024-09-22 20:56:33,955 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.439e+02 1.570e+02 1.892e+02 4.315e+02, threshold=3.139e+02, percent-clipped=1.0
2024-09-22 20:56:45,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=100893.33333333333, ans=0.0
2024-09-22 20:56:58,027 INFO [train.py:1198] (1/4) Epoch 6, batch 2150, loss[loss=0.2559, ctc_loss=0.1841, cr_loss=0.3591, over 17024.00 frames. ], tot_loss[loss=0.2868, ctc_loss=0.2065, cr_loss=0.4018, over 3357434.67 frames. ], batch size: 44, lr: 1.95e-02, grad_scale: 32.0
2024-09-22 20:57:57,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=101080.0, ans=0.125
2024-09-22 20:58:11,481 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.60 vs. limit=12.0
2024-09-22 20:58:12,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=101126.66666666667, ans=0.125
2024-09-22 20:58:20,078 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.15 vs. limit=15.0
2024-09-22 20:58:25,370 INFO [train.py:1198] (1/4) Epoch 6, batch 2200, loss[loss=0.247, ctc_loss=0.1721, cr_loss=0.3744, over 17014.00 frames. ], tot_loss[loss=0.2887, ctc_loss=0.2081, cr_loss=0.403, over 3348201.74 frames. ], batch size: 44, lr: 1.95e-02, grad_scale: 32.0
2024-09-22 20:58:30,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=101173.33333333333, ans=0.125
2024-09-22 20:58:48,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=101220.0, ans=0.0
2024-09-22 20:58:54,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=101220.0, ans=0.125
2024-09-22 20:59:13,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=101313.33333333333, ans=0.0
2024-09-22 20:59:22,925 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.501e+02 1.682e+02 2.031e+02 3.137e+02, threshold=3.364e+02, percent-clipped=0.0
2024-09-22 20:59:26,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=101313.33333333333, ans=0.0
2024-09-22 20:59:37,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=101360.0, ans=0.125
2024-09-22 20:59:41,813 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.29 vs. limit=15.0
2024-09-22 20:59:46,926 INFO [train.py:1198] (1/4) Epoch 6, batch 2250, loss[loss=0.2213, ctc_loss=0.1521, cr_loss=0.3458, over 17089.00 frames. ], tot_loss[loss=0.2876, ctc_loss=0.2073, cr_loss=0.4016, over 3351112.34 frames. ], batch size: 43, lr: 1.95e-02, grad_scale: 32.0
2024-09-22 20:59:48,108 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.20 vs. limit=10.0
2024-09-22 21:00:05,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=101453.33333333333, ans=0.05
2024-09-22 21:00:23,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=101500.0, ans=0.0
2024-09-22 21:00:38,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=101546.66666666667, ans=0.1
2024-09-22 21:01:06,134 INFO [train.py:1198] (1/4) Epoch 6, batch 2300, loss[loss=0.2388, ctc_loss=0.1713, cr_loss=0.3378, over 16972.00 frames. ], tot_loss[loss=0.2855, ctc_loss=0.2055, cr_loss=0.3995, over 3350965.33 frames. ], batch size: 42, lr: 1.94e-02, grad_scale: 32.0
2024-09-22 21:01:59,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=101780.0, ans=0.0
2024-09-22 21:02:06,859 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.444e+02 1.711e+02 1.913e+02 2.754e+02, threshold=3.422e+02, percent-clipped=0.0
2024-09-22 21:02:07,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=101780.0, ans=0.2
2024-09-22 21:02:10,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=101780.0, ans=0.125
2024-09-22 21:02:30,422 INFO [train.py:1198] (1/4) Epoch 6, batch 2350, loss[loss=0.2707, ctc_loss=0.1919, cr_loss=0.3935, over 17102.00 frames. ], tot_loss[loss=0.285, ctc_loss=0.2052, cr_loss=0.399, over 3347359.34 frames. ], batch size: 49, lr: 1.94e-02, grad_scale: 32.0
2024-09-22 21:02:42,019 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.92 vs. limit=15.0
2024-09-22 21:03:21,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.min_positive, batch_count=102013.33333333333, ans=0.025
2024-09-22 21:03:23,322 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.47 vs. limit=15.0
2024-09-22 21:03:41,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=102060.0, ans=0.0
2024-09-22 21:03:55,324 INFO [train.py:1198] (1/4) Epoch 6, batch 2400, loss[loss=0.2781, ctc_loss=0.2002, cr_loss=0.3896, over 17221.00 frames. ], tot_loss[loss=0.2841, ctc_loss=0.2044, cr_loss=0.3984, over 3345687.79 frames. ], batch size: 50, lr: 1.94e-02, grad_scale: 32.0
2024-09-22 21:04:06,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=102106.66666666667, ans=0.125
2024-09-22 21:04:09,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=102153.33333333333, ans=0.025
2024-09-22 21:04:09,901 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:04:22,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=102153.33333333333, ans=0.125
2024-09-22 21:04:27,356 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.03 vs. limit=15.0
2024-09-22 21:04:32,519 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=9.36 vs. limit=22.5
2024-09-22 21:04:50,184 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.414e+02 1.609e+02 1.881e+02 2.919e+02, threshold=3.217e+02, percent-clipped=0.0
2024-09-22 21:05:00,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=102293.33333333333, ans=0.1
2024-09-22 21:05:03,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=102293.33333333333, ans=0.1
2024-09-22 21:05:14,225 INFO [train.py:1198] (1/4) Epoch 6, batch 2450, loss[loss=0.2539, ctc_loss=0.1832, cr_loss=0.3538, over 17178.00 frames. ], tot_loss[loss=0.2834, ctc_loss=0.2038, cr_loss=0.3982, over 3351501.55 frames. ], batch size: 41, lr: 1.94e-02, grad_scale: 32.0
2024-09-22 21:05:54,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=102433.33333333333, ans=0.125
2024-09-22 21:06:13,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=102480.0, ans=0.2
2024-09-22 21:06:33,760 INFO [train.py:1198] (1/4) Epoch 6, batch 2500, loss[loss=0.3173, ctc_loss=0.2281, cr_loss=0.4459, over 16733.00 frames. ], tot_loss[loss=0.2837, ctc_loss=0.204, cr_loss=0.3988, over 3357467.60 frames. ], batch size: 61, lr: 1.94e-02, grad_scale: 32.0
2024-09-22 21:06:55,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=102620.0, ans=0.125
2024-09-22 21:07:19,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=102666.66666666667, ans=0.125
2024-09-22 21:07:34,920 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.454e+02 1.581e+02 1.860e+02 2.771e+02, threshold=3.162e+02, percent-clipped=0.0
2024-09-22 21:07:46,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=102760.0, ans=0.2
2024-09-22 21:07:56,042 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.23 vs. limit=15.0
2024-09-22 21:08:01,445 INFO [train.py:1198] (1/4) Epoch 6, batch 2550, loss[loss=0.254, ctc_loss=0.1805, cr_loss=0.3673, over 16969.00 frames. ], tot_loss[loss=0.2841, ctc_loss=0.2042, cr_loss=0.3996, over 3356204.81 frames. ], batch size: 42, lr: 1.93e-02, grad_scale: 32.0
2024-09-22 21:08:08,839 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.55 vs. limit=15.0
2024-09-22 21:08:24,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=102853.33333333333, ans=0.0
2024-09-22 21:08:31,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.69 vs. limit=12.0
2024-09-22 21:08:44,832 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.53 vs. limit=22.5
2024-09-22 21:08:49,668 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.46 vs. limit=15.0
2024-09-22 21:08:50,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=102946.66666666667, ans=0.1
2024-09-22 21:09:19,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=102993.33333333333, ans=0.125
2024-09-22 21:09:23,042 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=9.54 vs. limit=22.5
2024-09-22 21:09:23,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn2.whiten.whitening_limit, batch_count=103040.0, ans=22.5
2024-09-22 21:09:23,605 INFO [train.py:1198] (1/4) Epoch 6, batch 2600, loss[loss=0.2574, ctc_loss=0.1843, cr_loss=0.3653, over 17208.00 frames. ], tot_loss[loss=0.285, ctc_loss=0.2049, cr_loss=0.4003, over 3362173.71 frames. ], batch size: 47, lr: 1.93e-02, grad_scale: 32.0
2024-09-22 21:09:58,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=103133.33333333333, ans=0.1
2024-09-22 21:10:18,671 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.459e+02 1.706e+02 2.072e+02 3.287e+02, threshold=3.412e+02, percent-clipped=1.0
2024-09-22 21:10:31,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=103226.66666666667, ans=0.2
2024-09-22 21:10:42,337 INFO [train.py:1198] (1/4) Epoch 6, batch 2650, loss[loss=0.3035, ctc_loss=0.2183, cr_loss=0.4259, over 16887.00 frames. ], tot_loss[loss=0.2838, ctc_loss=0.2039, cr_loss=0.3993, over 3368519.41 frames. ], batch size: 58, lr: 1.93e-02, grad_scale: 32.0
2024-09-22 21:10:44,402 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:11:00,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=103320.0, ans=0.125
2024-09-22 21:11:11,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=103320.0, ans=0.125
2024-09-22 21:11:25,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=103366.66666666667, ans=0.1
2024-09-22 21:12:00,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=103460.0, ans=0.0
2024-09-22 21:12:06,888 INFO [train.py:1198] (1/4) Epoch 6, batch 2700, loss[loss=0.2859, ctc_loss=0.2048, cr_loss=0.4055, over 17050.00 frames. ], tot_loss[loss=0.2828, ctc_loss=0.2032, cr_loss=0.3984, over 3366415.56 frames. ], batch size: 56, lr: 1.93e-02, grad_scale: 32.0
2024-09-22 21:12:34,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=103553.33333333333, ans=0.125
2024-09-22 21:13:05,402 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.442e+02 1.580e+02 1.786e+02 2.700e+02, threshold=3.159e+02, percent-clipped=0.0
2024-09-22 21:13:06,038 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.60 vs. limit=6.0
2024-09-22 21:13:31,579 INFO [train.py:1198] (1/4) Epoch 6, batch 2750, loss[loss=0.2837, ctc_loss=0.2014, cr_loss=0.4113, over 16738.00 frames. ], tot_loss[loss=0.2839, ctc_loss=0.2039, cr_loss=0.3999, over 3365935.43 frames. ], batch size: 61, lr: 1.93e-02, grad_scale: 32.0
2024-09-22 21:13:34,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=103740.0, ans=0.1
2024-09-22 21:13:54,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=103786.66666666667, ans=0.125
2024-09-22 21:14:48,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=103926.66666666667, ans=0.0
2024-09-22 21:14:51,040 INFO [train.py:1198] (1/4) Epoch 6, batch 2800, loss[loss=0.2922, ctc_loss=0.2079, cr_loss=0.4217, over 17354.00 frames. ], tot_loss[loss=0.2841, ctc_loss=0.2042, cr_loss=0.3996, over 3351470.23 frames. ], batch size: 48, lr: 1.92e-02, grad_scale: 32.0
2024-09-22 21:15:21,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=104066.66666666667, ans=0.125
2024-09-22 21:15:46,539 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.391e+02 1.557e+02 1.769e+02 3.866e+02, threshold=3.114e+02, percent-clipped=1.0
2024-09-22 21:15:46,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=104113.33333333333, ans=0.035
2024-09-22 21:16:03,105 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=9.47 vs. limit=15.0
2024-09-22 21:16:04,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=104160.0, ans=0.125
2024-09-22 21:16:10,246 INFO [train.py:1198] (1/4) Epoch 6, batch 2850, loss[loss=0.2909, ctc_loss=0.2075, cr_loss=0.4171, over 17102.00 frames. ], tot_loss[loss=0.2843, ctc_loss=0.2043, cr_loss=0.3997, over 3351475.51 frames. ], batch size: 49, lr: 1.92e-02, grad_scale: 32.0
2024-09-22 21:16:10,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=104206.66666666667, ans=0.025
2024-09-22 21:16:18,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=8.10 vs. limit=15.0
2024-09-22 21:16:21,535 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:16:23,741 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.47 vs. limit=15.0
2024-09-22 21:16:31,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=104253.33333333333, ans=10.0
2024-09-22 21:17:12,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=104346.66666666667, ans=0.125
2024-09-22 21:17:28,025 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=26.89 vs. limit=22.5
2024-09-22 21:17:35,031 INFO [train.py:1198] (1/4) Epoch 6, batch 2900, loss[loss=0.2576, ctc_loss=0.1796, cr_loss=0.3899, over 17041.00 frames. ], tot_loss[loss=0.2844, ctc_loss=0.2044, cr_loss=0.3995, over 3353466.49 frames. ], batch size: 39, lr: 1.92e-02, grad_scale: 32.0
2024-09-22 21:17:53,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=104486.66666666667, ans=0.0
2024-09-22 21:17:56,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=104486.66666666667, ans=0.1
2024-09-22 21:18:01,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=104486.66666666667, ans=10.0
2024-09-22 21:18:03,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=104486.66666666667, ans=0.125
2024-09-22 21:18:09,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=104533.33333333333, ans=0.125
2024-09-22 21:18:17,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=104533.33333333333, ans=0.2
2024-09-22 21:18:31,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=104580.0, ans=0.0
2024-09-22 21:18:35,985 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.508e+02 1.666e+02 1.915e+02 2.988e+02, threshold=3.332e+02, percent-clipped=0.0
2024-09-22 21:18:37,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=104580.0, ans=0.125
2024-09-22 21:18:59,988 INFO [train.py:1198] (1/4) Epoch 6, batch 2950, loss[loss=0.3187, ctc_loss=0.2341, cr_loss=0.4232, over 16793.00 frames. ], tot_loss[loss=0.2839, ctc_loss=0.2039, cr_loss=0.3997, over 3359980.70 frames. ], batch size: 61, lr: 1.92e-02, grad_scale: 32.0
2024-09-22 21:19:23,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.02 vs. limit=15.0
2024-09-22 21:19:34,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2.whitening_limit, batch_count=104766.66666666667, ans=15.0
2024-09-22 21:20:19,065 INFO [train.py:1198] (1/4) Epoch 6, batch 3000, loss[loss=0.3039, ctc_loss=0.2228, cr_loss=0.4055, over 14811.00 frames. ], tot_loss[loss=0.2837, ctc_loss=0.2038, cr_loss=0.3993, over 3348991.71 frames. ], batch size: 89, lr: 1.92e-02, grad_scale: 32.0
2024-09-22 21:20:19,065 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 21:20:30,217 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([3.7103, 4.7126, 4.0189, 4.4675, 4.4475, 3.7388, 4.0484, 3.7475],
       device='cuda:1')
2024-09-22 21:20:34,502 INFO [train.py:1230] (1/4) Epoch 6, validation: loss=0.06097, ctc_loss=0.06097, cr_loss=6.736e-15, over 944034.00 frames. 
2024-09-22 21:20:34,503 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 21:20:36,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=5.43 vs. limit=12.0
2024-09-22 21:20:55,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=104953.33333333333, ans=0.0
2024-09-22 21:21:12,369 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:21:13,850 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=105000.0, ans=0.125
2024-09-22 21:21:17,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=105000.0, ans=0.1
2024-09-22 21:21:28,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=105046.66666666667, ans=0.125
2024-09-22 21:21:29,318 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.397e+02 1.557e+02 1.721e+02 3.093e+02, threshold=3.113e+02, percent-clipped=0.0
2024-09-22 21:21:31,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=105046.66666666667, ans=0.07
2024-09-22 21:21:32,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=105046.66666666667, ans=0.1
2024-09-22 21:21:35,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=105093.33333333333, ans=0.1
2024-09-22 21:21:47,805 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.95 vs. limit=15.0
2024-09-22 21:21:52,864 INFO [train.py:1198] (1/4) Epoch 6, batch 3050, loss[loss=0.251, ctc_loss=0.1801, cr_loss=0.3544, over 17080.00 frames. ], tot_loss[loss=0.2842, ctc_loss=0.2041, cr_loss=0.4003, over 3359583.13 frames. ], batch size: 43, lr: 1.92e-02, grad_scale: 32.0
2024-09-22 21:22:11,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=105186.66666666667, ans=0.0
2024-09-22 21:22:12,104 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.42 vs. limit=6.0
2024-09-22 21:22:21,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=105186.66666666667, ans=0.125
2024-09-22 21:22:24,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=105233.33333333333, ans=0.05
2024-09-22 21:23:12,835 INFO [train.py:1198] (1/4) Epoch 6, batch 3100, loss[loss=0.3338, ctc_loss=0.2426, cr_loss=0.456, over 15938.00 frames. ], tot_loss[loss=0.2858, ctc_loss=0.2055, cr_loss=0.4015, over 3346191.82 frames. ], batch size: 74, lr: 1.91e-02, grad_scale: 32.0
2024-09-22 21:23:26,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=105420.0, ans=0.0
2024-09-22 21:23:49,648 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=5.42 vs. limit=15.0
2024-09-22 21:24:09,900 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.436e+02 1.623e+02 1.870e+02 2.887e+02, threshold=3.246e+02, percent-clipped=0.0
2024-09-22 21:24:17,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=105560.0, ans=0.125
2024-09-22 21:24:23,083 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=12.31 vs. limit=15.0
2024-09-22 21:24:25,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=105560.0, ans=0.2
2024-09-22 21:24:33,411 INFO [train.py:1198] (1/4) Epoch 6, batch 3150, loss[loss=0.2593, ctc_loss=0.1837, cr_loss=0.378, over 17256.00 frames. ], tot_loss[loss=0.284, ctc_loss=0.204, cr_loss=0.3999, over 3354557.27 frames. ], batch size: 44, lr: 1.91e-02, grad_scale: 32.0
2024-09-22 21:25:26,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=105746.66666666667, ans=0.09899494936611666
2024-09-22 21:25:44,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=105793.33333333333, ans=0.025
2024-09-22 21:25:53,519 INFO [train.py:1198] (1/4) Epoch 6, batch 3200, loss[loss=0.2842, ctc_loss=0.2029, cr_loss=0.4065, over 17339.00 frames. ], tot_loss[loss=0.2848, ctc_loss=0.2046, cr_loss=0.401, over 3353207.81 frames. ], batch size: 48, lr: 1.91e-02, grad_scale: 32.0
2024-09-22 21:25:55,813 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=15.34 vs. limit=15.0
2024-09-22 21:26:07,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=105886.66666666667, ans=0.1
2024-09-22 21:26:17,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=105886.66666666667, ans=0.125
2024-09-22 21:26:21,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=105886.66666666667, ans=0.125
2024-09-22 21:26:29,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=105933.33333333333, ans=0.0
2024-09-22 21:26:35,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=105933.33333333333, ans=0.125
2024-09-22 21:26:46,491 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.94 vs. limit=15.0
2024-09-22 21:26:50,292 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.472e+02 1.715e+02 1.976e+02 3.064e+02, threshold=3.429e+02, percent-clipped=0.0
2024-09-22 21:26:59,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=106026.66666666667, ans=0.2
2024-09-22 21:27:01,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=106026.66666666667, ans=10.0
2024-09-22 21:27:13,660 INFO [train.py:1198] (1/4) Epoch 6, batch 3250, loss[loss=0.2856, ctc_loss=0.2027, cr_loss=0.4146, over 17217.00 frames. ], tot_loss[loss=0.2847, ctc_loss=0.2046, cr_loss=0.4007, over 3347851.47 frames. ], batch size: 55, lr: 1.91e-02, grad_scale: 32.0
2024-09-22 21:27:20,476 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.77 vs. limit=6.0
2024-09-22 21:27:38,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=106120.0, ans=0.1
2024-09-22 21:27:38,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=106120.0, ans=0.125
2024-09-22 21:27:45,099 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=106166.66666666667, ans=0.125
2024-09-22 21:27:54,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=106166.66666666667, ans=0.125
2024-09-22 21:27:59,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=106213.33333333333, ans=0.95
2024-09-22 21:28:31,721 INFO [train.py:1198] (1/4) Epoch 6, batch 3300, loss[loss=0.2673, ctc_loss=0.1894, cr_loss=0.3897, over 17354.00 frames. ], tot_loss[loss=0.2859, ctc_loss=0.2055, cr_loss=0.4021, over 3351039.16 frames. ], batch size: 48, lr: 1.91e-02, grad_scale: 64.0
2024-09-22 21:29:04,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=106400.0, ans=0.125
2024-09-22 21:29:06,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=106400.0, ans=0.2
2024-09-22 21:29:18,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=106446.66666666667, ans=10.0
2024-09-22 21:29:26,377 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.242e+02 1.523e+02 1.757e+02 2.023e+02 3.259e+02, threshold=3.514e+02, percent-clipped=0.0
2024-09-22 21:29:39,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=106493.33333333333, ans=0.09899494936611666
2024-09-22 21:29:49,747 INFO [train.py:1198] (1/4) Epoch 6, batch 3350, loss[loss=0.3028, ctc_loss=0.2142, cr_loss=0.4431, over 16865.00 frames. ], tot_loss[loss=0.2853, ctc_loss=0.2048, cr_loss=0.4022, over 3362756.62 frames. ], batch size: 58, lr: 1.90e-02, grad_scale: 32.0
2024-09-22 21:29:54,943 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:30:24,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=106633.33333333333, ans=0.0
2024-09-22 21:30:45,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=106680.0, ans=0.0
2024-09-22 21:30:53,757 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.32 vs. limit=15.0
2024-09-22 21:30:54,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=106726.66666666667, ans=0.1
2024-09-22 21:31:08,299 INFO [train.py:1198] (1/4) Epoch 6, batch 3400, loss[loss=0.3064, ctc_loss=0.2196, cr_loss=0.4339, over 16981.00 frames. ], tot_loss[loss=0.2834, ctc_loss=0.2034, cr_loss=0.4001, over 3369757.23 frames. ], batch size: 53, lr: 1.90e-02, grad_scale: 32.0
2024-09-22 21:31:14,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=106773.33333333333, ans=0.125
2024-09-22 21:31:25,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=106820.0, ans=0.2
2024-09-22 21:32:04,335 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 1.439e+02 1.572e+02 1.824e+02 3.611e+02, threshold=3.144e+02, percent-clipped=1.0
2024-09-22 21:32:09,704 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.81 vs. limit=15.0
2024-09-22 21:32:26,356 INFO [train.py:1198] (1/4) Epoch 6, batch 3450, loss[loss=0.3245, ctc_loss=0.2352, cr_loss=0.4466, over 17218.00 frames. ], tot_loss[loss=0.286, ctc_loss=0.2054, cr_loss=0.4029, over 3361235.14 frames. ], batch size: 55, lr: 1.90e-02, grad_scale: 32.0
2024-09-22 21:32:48,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=107053.33333333333, ans=0.025
2024-09-22 21:33:03,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=107100.0, ans=0.125
2024-09-22 21:33:46,059 INFO [train.py:1198] (1/4) Epoch 6, batch 3500, loss[loss=0.271, ctc_loss=0.1965, cr_loss=0.3726, over 17158.00 frames. ], tot_loss[loss=0.2857, ctc_loss=0.2053, cr_loss=0.4018, over 3354969.07 frames. ], batch size: 41, lr: 1.90e-02, grad_scale: 32.0
2024-09-22 21:33:49,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=107240.0, ans=0.035
2024-09-22 21:34:10,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=107286.66666666667, ans=0.125
2024-09-22 21:34:16,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=107286.66666666667, ans=0.0
2024-09-22 21:34:44,453 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.542e+02 1.676e+02 1.907e+02 3.181e+02, threshold=3.352e+02, percent-clipped=1.0
2024-09-22 21:35:06,138 INFO [train.py:1198] (1/4) Epoch 6, batch 3550, loss[loss=0.2779, ctc_loss=0.2006, cr_loss=0.3867, over 17171.00 frames. ], tot_loss[loss=0.2842, ctc_loss=0.2042, cr_loss=0.4, over 3357463.32 frames. ], batch size: 45, lr: 1.90e-02, grad_scale: 32.0
2024-09-22 21:35:17,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=107473.33333333333, ans=0.125
2024-09-22 21:35:47,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=107566.66666666667, ans=0.125
2024-09-22 21:36:10,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=107660.0, ans=0.035
2024-09-22 21:36:12,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=107660.0, ans=0.125
2024-09-22 21:36:22,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=107660.0, ans=0.0
2024-09-22 21:36:28,360 INFO [train.py:1198] (1/4) Epoch 6, batch 3600, loss[loss=0.3209, ctc_loss=0.2311, cr_loss=0.4494, over 15038.00 frames. ], tot_loss[loss=0.2843, ctc_loss=0.2041, cr_loss=0.4012, over 3357691.22 frames. ], batch size: 89, lr: 1.89e-02, grad_scale: 32.0
2024-09-22 21:36:49,112 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:36:55,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=107753.33333333333, ans=0.025
2024-09-22 21:36:58,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=107800.0, ans=0.0
2024-09-22 21:37:18,847 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=45.35 vs. limit=15.0
2024-09-22 21:37:24,563 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.473e+02 1.668e+02 1.932e+02 3.410e+02, threshold=3.336e+02, percent-clipped=1.0
2024-09-22 21:37:26,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=107846.66666666667, ans=0.2
2024-09-22 21:37:42,599 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.15 vs. limit=15.0
2024-09-22 21:37:46,253 INFO [train.py:1198] (1/4) Epoch 6, batch 3650, loss[loss=0.2462, ctc_loss=0.1755, cr_loss=0.3532, over 17168.00 frames. ], tot_loss[loss=0.2836, ctc_loss=0.2034, cr_loss=0.401, over 3366519.75 frames. ], batch size: 45, lr: 1.89e-02, grad_scale: 32.0
2024-09-22 21:38:19,714 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=10.86 vs. limit=15.0
2024-09-22 21:38:53,071 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.54 vs. limit=6.0
2024-09-22 21:38:54,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer_ff2.min_abs, batch_count=108126.66666666667, ans=0.1
2024-09-22 21:38:56,135 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.32 vs. limit=15.0
2024-09-22 21:39:03,837 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.59 vs. limit=15.0
2024-09-22 21:39:04,804 INFO [train.py:1198] (1/4) Epoch 6, batch 3700, loss[loss=0.2537, ctc_loss=0.1786, cr_loss=0.3753, over 16997.00 frames. ], tot_loss[loss=0.2831, ctc_loss=0.2031, cr_loss=0.4, over 3364588.74 frames. ], batch size: 51, lr: 1.89e-02, grad_scale: 32.0
2024-09-22 21:39:05,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=108173.33333333333, ans=0.125
2024-09-22 21:39:22,907 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.47 vs. limit=6.0
2024-09-22 21:40:01,663 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.455e+02 1.690e+02 2.004e+02 3.040e+02, threshold=3.380e+02, percent-clipped=0.0
2024-09-22 21:40:23,106 INFO [train.py:1198] (1/4) Epoch 6, batch 3750, loss[loss=0.2803, ctc_loss=0.199, cr_loss=0.4064, over 17019.00 frames. ], tot_loss[loss=0.2836, ctc_loss=0.2034, cr_loss=0.4011, over 3367662.10 frames. ], batch size: 53, lr: 1.89e-02, grad_scale: 32.0
2024-09-22 21:40:23,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=108406.66666666667, ans=0.2
2024-09-22 21:40:26,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=108406.66666666667, ans=0.2
2024-09-22 21:40:29,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=108406.66666666667, ans=0.125
2024-09-22 21:40:42,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=108453.33333333333, ans=0.1
2024-09-22 21:41:01,971 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.18 vs. limit=15.0
2024-09-22 21:41:03,231 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.24 vs. limit=15.0
2024-09-22 21:41:28,542 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.01 vs. limit=6.0
2024-09-22 21:41:42,109 INFO [train.py:1198] (1/4) Epoch 6, batch 3800, loss[loss=0.2482, ctc_loss=0.1808, cr_loss=0.337, over 16948.00 frames. ], tot_loss[loss=0.2836, ctc_loss=0.2036, cr_loss=0.4001, over 3346344.18 frames. ], batch size: 42, lr: 1.89e-02, grad_scale: 32.0
2024-09-22 21:41:42,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=108640.0, ans=0.0
2024-09-22 21:41:56,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=108686.66666666667, ans=0.0
2024-09-22 21:42:01,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=108686.66666666667, ans=0.125
2024-09-22 21:42:10,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=108686.66666666667, ans=0.1
2024-09-22 21:42:37,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=108780.0, ans=0.1
2024-09-22 21:42:38,521 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=6.58 vs. limit=15.0
2024-09-22 21:42:39,193 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.543e+02 1.830e+02 2.234e+02 3.927e+02, threshold=3.660e+02, percent-clipped=2.0
2024-09-22 21:42:50,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=108826.66666666667, ans=0.125
2024-09-22 21:42:58,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=108826.66666666667, ans=0.125
2024-09-22 21:43:00,879 INFO [train.py:1198] (1/4) Epoch 6, batch 3850, loss[loss=0.3404, ctc_loss=0.255, cr_loss=0.4271, over 11699.00 frames. ], tot_loss[loss=0.2865, ctc_loss=0.2062, cr_loss=0.4012, over 3295246.30 frames. ], batch size: 123, lr: 1.89e-02, grad_scale: 32.0
2024-09-22 21:43:38,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=108966.66666666667, ans=0.2
2024-09-22 21:43:39,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=108966.66666666667, ans=0.025
2024-09-22 21:44:00,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=109060.0, ans=0.2
2024-09-22 21:44:08,876 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.74 vs. limit=15.0
2024-09-22 21:45:03,034 INFO [train.py:1198] (1/4) Epoch 7, batch 0, loss[loss=0.2659, ctc_loss=0.1909, cr_loss=0.3749, over 17061.00 frames. ], tot_loss[loss=0.2659, ctc_loss=0.1909, cr_loss=0.3749, over 17061.00 frames. ], batch size: 46, lr: 1.77e-02, grad_scale: 32.0
2024-09-22 21:45:03,035 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 21:45:18,429 INFO [train.py:1230] (1/4) Epoch 7, validation: loss=0.06283, ctc_loss=0.06283, cr_loss=7.028e-15, over 944034.00 frames. 
2024-09-22 21:45:18,429 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 21:45:22,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=109088.0, ans=0.1
2024-09-22 21:46:24,181 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.560e+02 1.911e+02 2.609e+02 3.890e+02, threshold=3.822e+02, percent-clipped=2.0
2024-09-22 21:46:39,894 INFO [train.py:1198] (1/4) Epoch 7, batch 50, loss[loss=0.2636, ctc_loss=0.1876, cr_loss=0.3802, over 17307.00 frames. ], tot_loss[loss=0.2849, ctc_loss=0.2046, cr_loss=0.4016, over 760460.96 frames. ], batch size: 46, lr: 1.76e-02, grad_scale: 32.0
2024-09-22 21:47:11,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=109368.0, ans=0.0
2024-09-22 21:47:13,410 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=22.71 vs. limit=22.5
2024-09-22 21:47:46,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=109508.0, ans=0.2
2024-09-22 21:47:55,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=109508.0, ans=0.0
2024-09-22 21:48:00,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=109508.0, ans=0.0
2024-09-22 21:48:04,578 INFO [train.py:1198] (1/4) Epoch 7, batch 100, loss[loss=0.3151, ctc_loss=0.2243, cr_loss=0.4543, over 16705.00 frames. ], tot_loss[loss=0.2844, ctc_loss=0.2039, cr_loss=0.4026, over 1332340.49 frames. ], batch size: 61, lr: 1.76e-02, grad_scale: 32.0
2024-09-22 21:48:16,191 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.43 vs. limit=15.0
2024-09-22 21:48:36,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=109648.0, ans=0.125
2024-09-22 21:49:07,937 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.339e+02 1.498e+02 1.759e+02 2.443e+02, threshold=2.996e+02, percent-clipped=0.0
2024-09-22 21:49:26,881 INFO [train.py:1198] (1/4) Epoch 7, batch 150, loss[loss=0.2711, ctc_loss=0.1904, cr_loss=0.4039, over 17264.00 frames. ], tot_loss[loss=0.2842, ctc_loss=0.2038, cr_loss=0.4018, over 1771014.11 frames. ], batch size: 42, lr: 1.76e-02, grad_scale: 32.0
2024-09-22 21:49:43,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=109834.66666666667, ans=0.125
2024-09-22 21:49:46,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=109834.66666666667, ans=0.1
2024-09-22 21:49:54,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=109834.66666666667, ans=0.025
2024-09-22 21:49:59,051 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:49:59,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=109881.33333333333, ans=0.0
2024-09-22 21:50:08,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=109881.33333333333, ans=0.125
2024-09-22 21:50:30,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=109974.66666666667, ans=0.125
2024-09-22 21:50:31,773 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.75 vs. limit=15.0
2024-09-22 21:50:46,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=109974.66666666667, ans=0.125
2024-09-22 21:50:49,452 INFO [train.py:1198] (1/4) Epoch 7, batch 200, loss[loss=0.2949, ctc_loss=0.2084, cr_loss=0.4324, over 17153.00 frames. ], tot_loss[loss=0.2819, ctc_loss=0.202, cr_loss=0.3993, over 2117599.92 frames. ], batch size: 48, lr: 1.76e-02, grad_scale: 32.0
2024-09-22 21:50:49,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=110021.33333333333, ans=0.125
2024-09-22 21:51:12,717 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.54 vs. limit=15.0
2024-09-22 21:51:27,003 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.58 vs. limit=15.0
2024-09-22 21:51:48,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=110161.33333333333, ans=0.0
2024-09-22 21:51:53,494 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.411e+02 1.617e+02 1.825e+02 4.000e+02, threshold=3.234e+02, percent-clipped=2.0
2024-09-22 21:52:12,070 INFO [train.py:1198] (1/4) Epoch 7, batch 250, loss[loss=0.2699, ctc_loss=0.1887, cr_loss=0.4062, over 17153.00 frames. ], tot_loss[loss=0.2786, ctc_loss=0.1994, cr_loss=0.3963, over 2401693.89 frames. ], batch size: 48, lr: 1.76e-02, grad_scale: 32.0
2024-09-22 21:52:22,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=110254.66666666667, ans=0.125
2024-09-22 21:52:58,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=110348.0, ans=0.125
2024-09-22 21:53:01,945 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.99 vs. limit=22.5
2024-09-22 21:53:17,718 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:53:30,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=110441.33333333333, ans=0.1
2024-09-22 21:53:33,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=110488.0, ans=0.1
2024-09-22 21:53:34,853 INFO [train.py:1198] (1/4) Epoch 7, batch 300, loss[loss=0.298, ctc_loss=0.2157, cr_loss=0.4113, over 17262.00 frames. ], tot_loss[loss=0.2817, ctc_loss=0.2018, cr_loss=0.3995, over 2607611.66 frames. ], batch size: 55, lr: 1.76e-02, grad_scale: 32.0
2024-09-22 21:54:41,018 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.446e+02 1.620e+02 1.814e+02 2.683e+02, threshold=3.241e+02, percent-clipped=0.0
2024-09-22 21:54:51,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=110674.66666666667, ans=0.125
2024-09-22 21:54:54,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=110674.66666666667, ans=0.125
2024-09-22 21:54:56,993 INFO [train.py:1198] (1/4) Epoch 7, batch 350, loss[loss=0.2291, ctc_loss=0.1598, cr_loss=0.3463, over 16326.00 frames. ], tot_loss[loss=0.2808, ctc_loss=0.2009, cr_loss=0.3994, over 2777118.42 frames. ], batch size: 36, lr: 1.75e-02, grad_scale: 32.0
2024-09-22 21:55:03,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=110721.33333333333, ans=0.0
2024-09-22 21:55:21,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=110768.0, ans=0.125
2024-09-22 21:55:54,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=110861.33333333333, ans=0.05
2024-09-22 21:56:00,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2.whitening_limit, batch_count=110861.33333333333, ans=15.0
2024-09-22 21:56:10,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=110908.0, ans=0.05
2024-09-22 21:56:11,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=110908.0, ans=0.0
2024-09-22 21:56:12,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=110908.0, ans=0.2
2024-09-22 21:56:19,158 INFO [train.py:1198] (1/4) Epoch 7, batch 400, loss[loss=0.2802, ctc_loss=0.2015, cr_loss=0.3936, over 17022.00 frames. ], tot_loss[loss=0.2814, ctc_loss=0.2014, cr_loss=0.4001, over 2904529.31 frames. ], batch size: 56, lr: 1.75e-02, grad_scale: 32.0
2024-09-22 21:56:29,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=110954.66666666667, ans=0.1
2024-09-22 21:56:41,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=111001.33333333333, ans=0.2
2024-09-22 21:56:44,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=111001.33333333333, ans=0.2
2024-09-22 21:57:13,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=111094.66666666667, ans=0.125
2024-09-22 21:57:25,760 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.408e+02 1.555e+02 1.806e+02 2.695e+02, threshold=3.109e+02, percent-clipped=0.0
2024-09-22 21:57:41,728 INFO [train.py:1198] (1/4) Epoch 7, batch 450, loss[loss=0.3051, ctc_loss=0.2181, cr_loss=0.435, over 17307.00 frames. ], tot_loss[loss=0.2814, ctc_loss=0.2014, cr_loss=0.4002, over 2992998.35 frames. ], batch size: 49, lr: 1.75e-02, grad_scale: 32.0
2024-09-22 21:58:03,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=111234.66666666667, ans=0.125
2024-09-22 21:58:11,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=111234.66666666667, ans=0.125
2024-09-22 21:58:21,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=111281.33333333333, ans=0.0
2024-09-22 21:58:46,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=111374.66666666667, ans=0.125
2024-09-22 21:58:59,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=111374.66666666667, ans=0.0
2024-09-22 21:59:03,788 INFO [train.py:1198] (1/4) Epoch 7, batch 500, loss[loss=0.2497, ctc_loss=0.1739, cr_loss=0.3787, over 17262.00 frames. ], tot_loss[loss=0.2809, ctc_loss=0.201, cr_loss=0.3996, over 3081925.55 frames. ], batch size: 42, lr: 1.75e-02, grad_scale: 32.0
2024-09-22 21:59:06,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.47 vs. limit=15.0
2024-09-22 21:59:16,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=111421.33333333333, ans=0.125
2024-09-22 21:59:29,163 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 21:59:56,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=111561.33333333333, ans=0.125
2024-09-22 22:00:09,622 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.405e+02 1.626e+02 1.844e+02 3.754e+02, threshold=3.253e+02, percent-clipped=1.0
2024-09-22 22:00:25,406 INFO [train.py:1198] (1/4) Epoch 7, batch 550, loss[loss=0.2821, ctc_loss=0.2012, cr_loss=0.4048, over 17058.00 frames. ], tot_loss[loss=0.2809, ctc_loss=0.2009, cr_loss=0.4, over 3147993.80 frames. ], batch size: 46, lr: 1.75e-02, grad_scale: 32.0
2024-09-22 22:01:06,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=111748.0, ans=0.1
2024-09-22 22:01:47,901 INFO [train.py:1198] (1/4) Epoch 7, batch 600, loss[loss=0.264, ctc_loss=0.1871, cr_loss=0.3846, over 17018.00 frames. ], tot_loss[loss=0.2808, ctc_loss=0.2008, cr_loss=0.3998, over 3188483.80 frames. ], batch size: 51, lr: 1.75e-02, grad_scale: 32.0
2024-09-22 22:01:51,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=111888.0, ans=0.0
2024-09-22 22:02:27,524 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.69 vs. limit=15.0
2024-09-22 22:02:36,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=111981.33333333333, ans=0.2
2024-09-22 22:02:45,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=112028.0, ans=0.1
2024-09-22 22:02:51,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=112028.0, ans=0.125
2024-09-22 22:02:58,792 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.416e+02 1.598e+02 2.018e+02 3.504e+02, threshold=3.196e+02, percent-clipped=2.0
2024-09-22 22:03:14,672 INFO [train.py:1198] (1/4) Epoch 7, batch 650, loss[loss=0.2657, ctc_loss=0.1872, cr_loss=0.3921, over 16313.00 frames. ], tot_loss[loss=0.2804, ctc_loss=0.2006, cr_loss=0.3989, over 3220726.46 frames. ], batch size: 36, lr: 1.74e-02, grad_scale: 32.0
2024-09-22 22:03:29,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=112168.0, ans=0.0
2024-09-22 22:04:21,584 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.43 vs. limit=15.0
2024-09-22 22:04:35,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=112354.66666666667, ans=0.125
2024-09-22 22:04:36,942 INFO [train.py:1198] (1/4) Epoch 7, batch 700, loss[loss=0.2581, ctc_loss=0.1813, cr_loss=0.384, over 17118.00 frames. ], tot_loss[loss=0.2803, ctc_loss=0.2006, cr_loss=0.3988, over 3245021.77 frames. ], batch size: 40, lr: 1.74e-02, grad_scale: 32.0
2024-09-22 22:04:38,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=112354.66666666667, ans=0.1
2024-09-22 22:04:38,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=112354.66666666667, ans=0.125
2024-09-22 22:04:44,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=112354.66666666667, ans=0.0
2024-09-22 22:04:56,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=112401.33333333333, ans=0.0
2024-09-22 22:05:07,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=112448.0, ans=0.0
2024-09-22 22:05:18,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=112448.0, ans=0.0
2024-09-22 22:05:28,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=112494.66666666667, ans=0.1
2024-09-22 22:05:31,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=112494.66666666667, ans=0.025
2024-09-22 22:05:32,147 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.36 vs. limit=10.0
2024-09-22 22:05:39,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=112494.66666666667, ans=0.125
2024-09-22 22:05:42,318 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.430e+02 1.586e+02 1.856e+02 3.477e+02, threshold=3.173e+02, percent-clipped=1.0
2024-09-22 22:05:58,155 INFO [train.py:1198] (1/4) Epoch 7, batch 750, loss[loss=0.2764, ctc_loss=0.1976, cr_loss=0.3937, over 17021.00 frames. ], tot_loss[loss=0.2797, ctc_loss=0.1998, cr_loss=0.3993, over 3272677.14 frames. ], batch size: 44, lr: 1.74e-02, grad_scale: 32.0
2024-09-22 22:06:16,712 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.59 vs. limit=15.0
2024-09-22 22:06:25,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=112634.66666666667, ans=0.0
2024-09-22 22:06:45,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=112728.0, ans=0.125
2024-09-22 22:06:46,153 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.85 vs. limit=10.0
2024-09-22 22:07:02,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=112774.66666666667, ans=0.125
2024-09-22 22:07:19,627 INFO [train.py:1198] (1/4) Epoch 7, batch 800, loss[loss=0.2928, ctc_loss=0.2151, cr_loss=0.3887, over 14760.00 frames. ], tot_loss[loss=0.2805, ctc_loss=0.2006, cr_loss=0.3996, over 3286148.38 frames. ], batch size: 89, lr: 1.74e-02, grad_scale: 32.0
2024-09-22 22:07:21,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=112821.33333333333, ans=0.125
2024-09-22 22:07:44,645 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:07:57,560 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.73 vs. limit=15.0
2024-09-22 22:08:26,195 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.385e+02 1.523e+02 1.724e+02 2.705e+02, threshold=3.046e+02, percent-clipped=0.0
2024-09-22 22:08:26,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=113008.0, ans=0.0
2024-09-22 22:08:41,932 INFO [train.py:1198] (1/4) Epoch 7, batch 850, loss[loss=0.2711, ctc_loss=0.1932, cr_loss=0.3898, over 17144.00 frames. ], tot_loss[loss=0.2808, ctc_loss=0.201, cr_loss=0.3994, over 3301744.41 frames. ], batch size: 48, lr: 1.74e-02, grad_scale: 32.0
2024-09-22 22:08:50,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=113054.66666666667, ans=0.125
2024-09-22 22:09:14,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=113148.0, ans=0.0
2024-09-22 22:09:20,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=113148.0, ans=0.07
2024-09-22 22:09:42,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=113194.66666666667, ans=0.125
2024-09-22 22:10:03,662 INFO [train.py:1198] (1/4) Epoch 7, batch 900, loss[loss=0.2371, ctc_loss=0.1676, cr_loss=0.3474, over 17252.00 frames. ], tot_loss[loss=0.2799, ctc_loss=0.2002, cr_loss=0.3984, over 3309842.08 frames. ], batch size: 44, lr: 1.74e-02, grad_scale: 32.0
2024-09-22 22:10:05,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=113288.0, ans=0.125
2024-09-22 22:10:11,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=113288.0, ans=0.0
2024-09-22 22:10:14,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=113288.0, ans=0.125
2024-09-22 22:10:19,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=113288.0, ans=0.125
2024-09-22 22:10:28,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=113334.66666666667, ans=0.125
2024-09-22 22:10:28,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=113334.66666666667, ans=0.1
2024-09-22 22:10:41,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=113381.33333333333, ans=0.1
2024-09-22 22:10:52,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=113428.0, ans=0.125
2024-09-22 22:10:57,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=113428.0, ans=0.0
2024-09-22 22:10:58,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=113428.0, ans=0.025
2024-09-22 22:11:00,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=113428.0, ans=0.125
2024-09-22 22:11:03,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=113428.0, ans=0.125
2024-09-22 22:11:09,265 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.440e+02 1.570e+02 1.832e+02 2.236e+02, threshold=3.140e+02, percent-clipped=0.0
2024-09-22 22:11:11,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=113474.66666666667, ans=0.125
2024-09-22 22:11:25,181 INFO [train.py:1198] (1/4) Epoch 7, batch 950, loss[loss=0.282, ctc_loss=0.2005, cr_loss=0.4072, over 17219.00 frames. ], tot_loss[loss=0.2793, ctc_loss=0.1998, cr_loss=0.3975, over 3320893.25 frames. ], batch size: 47, lr: 1.73e-02, grad_scale: 32.0
2024-09-22 22:11:49,233 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.75 vs. limit=15.0
2024-09-22 22:12:01,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=113614.66666666667, ans=0.125
2024-09-22 22:12:32,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=113708.0, ans=0.125
2024-09-22 22:12:35,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.15 vs. limit=15.0
2024-09-22 22:12:36,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=113708.0, ans=0.125
2024-09-22 22:12:50,311 INFO [train.py:1198] (1/4) Epoch 7, batch 1000, loss[loss=0.2931, ctc_loss=0.2107, cr_loss=0.4117, over 17317.00 frames. ], tot_loss[loss=0.2782, ctc_loss=0.1989, cr_loss=0.3964, over 3332964.05 frames. ], batch size: 51, lr: 1.73e-02, grad_scale: 32.0
2024-09-22 22:13:07,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=113801.33333333333, ans=0.0
2024-09-22 22:13:11,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=113801.33333333333, ans=0.0
2024-09-22 22:13:53,767 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.387e+02 1.549e+02 1.823e+02 4.640e+02, threshold=3.099e+02, percent-clipped=1.0
2024-09-22 22:14:12,272 INFO [train.py:1198] (1/4) Epoch 7, batch 1050, loss[loss=0.2648, ctc_loss=0.1893, cr_loss=0.3774, over 17018.00 frames. ], tot_loss[loss=0.2771, ctc_loss=0.198, cr_loss=0.3956, over 3347680.61 frames. ], batch size: 53, lr: 1.73e-02, grad_scale: 32.0
2024-09-22 22:14:14,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=113988.0, ans=0.125
2024-09-22 22:14:17,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=113988.0, ans=0.125
2024-09-22 22:14:47,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=114081.33333333333, ans=0.09899494936611666
2024-09-22 22:14:52,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=114081.33333333333, ans=0.035
2024-09-22 22:15:33,221 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:15:34,439 INFO [train.py:1198] (1/4) Epoch 7, batch 1100, loss[loss=0.2576, ctc_loss=0.1845, cr_loss=0.3655, over 17224.00 frames. ], tot_loss[loss=0.2772, ctc_loss=0.1981, cr_loss=0.3956, over 3345637.33 frames. ], batch size: 50, lr: 1.73e-02, grad_scale: 32.0
2024-09-22 22:15:38,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=114221.33333333333, ans=0.125
2024-09-22 22:15:57,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=114268.0, ans=0.125
2024-09-22 22:16:14,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=114314.66666666667, ans=0.015
2024-09-22 22:16:34,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=114361.33333333333, ans=0.0
2024-09-22 22:16:37,933 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.457e+02 1.783e+02 2.141e+02 3.294e+02, threshold=3.566e+02, percent-clipped=3.0
2024-09-22 22:16:47,802 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=16.85 vs. limit=22.5
2024-09-22 22:16:50,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=114408.0, ans=0.125
2024-09-22 22:16:56,275 INFO [train.py:1198] (1/4) Epoch 7, batch 1150, loss[loss=0.231, ctc_loss=0.1599, cr_loss=0.3555, over 17033.00 frames. ], tot_loss[loss=0.275, ctc_loss=0.1963, cr_loss=0.3936, over 3349930.56 frames. ], batch size: 39, lr: 1.73e-02, grad_scale: 32.0
2024-09-22 22:17:02,075 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.10 vs. limit=15.0
2024-09-22 22:17:09,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=114454.66666666667, ans=0.125
2024-09-22 22:17:45,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=114594.66666666667, ans=0.0
2024-09-22 22:17:46,480 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:17:51,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=114594.66666666667, ans=0.1
2024-09-22 22:18:04,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=114641.33333333333, ans=0.05
2024-09-22 22:18:12,270 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.74 vs. limit=22.5
2024-09-22 22:18:17,892 INFO [train.py:1198] (1/4) Epoch 7, batch 1200, loss[loss=0.2541, ctc_loss=0.182, cr_loss=0.3603, over 17112.00 frames. ], tot_loss[loss=0.2765, ctc_loss=0.1974, cr_loss=0.3954, over 3345366.75 frames. ], batch size: 40, lr: 1.73e-02, grad_scale: 32.0
2024-09-22 22:18:22,022 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.78 vs. limit=10.0
2024-09-22 22:18:34,689 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.65 vs. limit=6.0
2024-09-22 22:18:40,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=114734.66666666667, ans=0.125
2024-09-22 22:18:51,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=114781.33333333333, ans=0.125
2024-09-22 22:19:07,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=114828.0, ans=0.125
2024-09-22 22:19:08,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=114828.0, ans=0.025
2024-09-22 22:19:12,738 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.11 vs. limit=6.0
2024-09-22 22:19:24,537 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.404e+02 1.583e+02 1.854e+02 5.575e+02, threshold=3.166e+02, percent-clipped=2.0
2024-09-22 22:19:40,343 INFO [train.py:1198] (1/4) Epoch 7, batch 1250, loss[loss=0.2625, ctc_loss=0.1856, cr_loss=0.3845, over 17162.00 frames. ], tot_loss[loss=0.2753, ctc_loss=0.1963, cr_loss=0.3945, over 3349630.86 frames. ], batch size: 41, lr: 1.72e-02, grad_scale: 32.0
2024-09-22 22:20:18,357 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.15 vs. limit=6.0
2024-09-22 22:20:43,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=115061.33333333333, ans=0.1
2024-09-22 22:20:52,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=115108.0, ans=0.125
2024-09-22 22:21:01,583 INFO [train.py:1198] (1/4) Epoch 7, batch 1300, loss[loss=0.2712, ctc_loss=0.192, cr_loss=0.396, over 17154.00 frames. ], tot_loss[loss=0.2753, ctc_loss=0.1964, cr_loss=0.3944, over 3350460.30 frames. ], batch size: 48, lr: 1.72e-02, grad_scale: 16.0
2024-09-22 22:21:23,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=115201.33333333333, ans=0.0
2024-09-22 22:21:45,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=115248.0, ans=0.1
2024-09-22 22:21:47,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=115248.0, ans=0.125
2024-09-22 22:22:09,482 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.485e+02 1.678e+02 2.013e+02 3.139e+02, threshold=3.356e+02, percent-clipped=0.0
2024-09-22 22:22:13,955 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.97 vs. limit=22.5
2024-09-22 22:22:26,394 INFO [train.py:1198] (1/4) Epoch 7, batch 1350, loss[loss=0.2648, ctc_loss=0.1856, cr_loss=0.3961, over 17094.00 frames. ], tot_loss[loss=0.275, ctc_loss=0.1962, cr_loss=0.3936, over 3346725.91 frames. ], batch size: 43, lr: 1.72e-02, grad_scale: 16.0
2024-09-22 22:22:39,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=115388.0, ans=0.1
2024-09-22 22:22:39,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=115388.0, ans=0.0
2024-09-22 22:22:43,237 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.89 vs. limit=22.5
2024-09-22 22:22:47,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=115434.66666666667, ans=0.09899494936611666
2024-09-22 22:22:49,093 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:23:33,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=115574.66666666667, ans=0.2
2024-09-22 22:23:39,239 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.00 vs. limit=15.0
2024-09-22 22:23:46,222 INFO [train.py:1198] (1/4) Epoch 7, batch 1400, loss[loss=0.2702, ctc_loss=0.1998, cr_loss=0.3519, over 16771.00 frames. ], tot_loss[loss=0.2747, ctc_loss=0.1962, cr_loss=0.3927, over 3340654.98 frames. ], batch size: 61, lr: 1.72e-02, grad_scale: 16.0
2024-09-22 22:23:55,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=115621.33333333333, ans=0.125
2024-09-22 22:23:55,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=115621.33333333333, ans=0.0
2024-09-22 22:24:11,808 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.26 vs. limit=15.0
2024-09-22 22:24:12,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=115668.0, ans=0.125
2024-09-22 22:24:22,996 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.54 vs. limit=10.0
2024-09-22 22:24:40,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=115761.33333333333, ans=0.0
2024-09-22 22:24:46,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=115761.33333333333, ans=0.0
2024-09-22 22:24:54,139 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.408e+02 1.629e+02 2.097e+02 4.051e+02, threshold=3.259e+02, percent-clipped=2.0
2024-09-22 22:24:59,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=115808.0, ans=0.1
2024-09-22 22:25:02,454 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.91 vs. limit=12.0
2024-09-22 22:25:04,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=115808.0, ans=0.125
2024-09-22 22:25:10,923 INFO [train.py:1198] (1/4) Epoch 7, batch 1450, loss[loss=0.3201, ctc_loss=0.23, cr_loss=0.4509, over 16606.00 frames. ], tot_loss[loss=0.2759, ctc_loss=0.197, cr_loss=0.3945, over 3347792.85 frames. ], batch size: 66, lr: 1.72e-02, grad_scale: 16.0
2024-09-22 22:25:44,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=115948.0, ans=0.125
2024-09-22 22:25:50,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=115948.0, ans=0.1
2024-09-22 22:26:02,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=115994.66666666667, ans=0.035
2024-09-22 22:26:06,701 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.32 vs. limit=15.0
2024-09-22 22:26:31,057 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:26:32,310 INFO [train.py:1198] (1/4) Epoch 7, batch 1500, loss[loss=0.2893, ctc_loss=0.2046, cr_loss=0.4231, over 17014.00 frames. ], tot_loss[loss=0.2748, ctc_loss=0.1961, cr_loss=0.3934, over 3342700.91 frames. ], batch size: 56, lr: 1.72e-02, grad_scale: 16.0
2024-09-22 22:26:33,399 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=16.18 vs. limit=22.5
2024-09-22 22:26:34,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=116088.0, ans=0.0
2024-09-22 22:26:42,582 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=22.33 vs. limit=22.5
2024-09-22 22:27:11,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=116181.33333333333, ans=0.125
2024-09-22 22:27:15,260 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.06 vs. limit=15.0
2024-09-22 22:27:40,531 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 1.482e+02 1.664e+02 1.936e+02 3.285e+02, threshold=3.328e+02, percent-clipped=1.0
2024-09-22 22:27:54,707 INFO [train.py:1198] (1/4) Epoch 7, batch 1550, loss[loss=0.2846, ctc_loss=0.2044, cr_loss=0.4006, over 16497.00 frames. ], tot_loss[loss=0.2732, ctc_loss=0.1947, cr_loss=0.3922, over 3353024.82 frames. ], batch size: 66, lr: 1.71e-02, grad_scale: 16.0
2024-09-22 22:28:04,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=116321.33333333333, ans=0.0
2024-09-22 22:28:15,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=116368.0, ans=0.125
2024-09-22 22:28:56,874 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.81 vs. limit=15.0
2024-09-22 22:29:06,168 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.82 vs. limit=15.0
2024-09-22 22:29:16,546 INFO [train.py:1198] (1/4) Epoch 7, batch 1600, loss[loss=0.2578, ctc_loss=0.1838, cr_loss=0.3701, over 16957.00 frames. ], tot_loss[loss=0.274, ctc_loss=0.1953, cr_loss=0.3935, over 3352394.85 frames. ], batch size: 42, lr: 1.71e-02, grad_scale: 32.0
2024-09-22 22:29:40,508 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:29:43,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=116601.33333333333, ans=0.0
2024-09-22 22:29:55,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=116648.0, ans=0.1
2024-09-22 22:30:03,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=116648.0, ans=0.035
2024-09-22 22:30:17,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.35 vs. limit=15.0
2024-09-22 22:30:18,431 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=5.24 vs. limit=12.0
2024-09-22 22:30:24,104 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.410e+02 1.579e+02 1.948e+02 3.056e+02, threshold=3.158e+02, percent-clipped=0.0
2024-09-22 22:30:38,422 INFO [train.py:1198] (1/4) Epoch 7, batch 1650, loss[loss=0.292, ctc_loss=0.2087, cr_loss=0.4168, over 17118.00 frames. ], tot_loss[loss=0.2754, ctc_loss=0.1963, cr_loss=0.3952, over 3347680.62 frames. ], batch size: 40, lr: 1.71e-02, grad_scale: 32.0
2024-09-22 22:30:43,984 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.68 vs. limit=10.0
2024-09-22 22:30:49,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=116788.0, ans=0.125
2024-09-22 22:30:59,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=116834.66666666667, ans=0.0
2024-09-22 22:31:37,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=116928.0, ans=0.1
2024-09-22 22:31:56,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=116974.66666666667, ans=0.125
2024-09-22 22:31:59,727 INFO [train.py:1198] (1/4) Epoch 7, batch 1700, loss[loss=0.284, ctc_loss=0.1998, cr_loss=0.4211, over 17293.00 frames. ], tot_loss[loss=0.2751, ctc_loss=0.196, cr_loss=0.3954, over 3353968.19 frames. ], batch size: 51, lr: 1.71e-02, grad_scale: 32.0
2024-09-22 22:32:13,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=117021.33333333333, ans=0.2
2024-09-22 22:32:28,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=117068.0, ans=0.125
2024-09-22 22:32:53,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=117161.33333333333, ans=0.0
2024-09-22 22:33:04,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=117208.0, ans=0.0
2024-09-22 22:33:07,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.332e+02 1.436e+02 1.651e+02 2.321e+02, threshold=2.871e+02, percent-clipped=0.0
2024-09-22 22:33:07,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=117208.0, ans=0.125
2024-09-22 22:33:11,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=117208.0, ans=0.125
2024-09-22 22:33:21,094 INFO [train.py:1198] (1/4) Epoch 7, batch 1750, loss[loss=0.3104, ctc_loss=0.226, cr_loss=0.4221, over 15076.00 frames. ], tot_loss[loss=0.2742, ctc_loss=0.1953, cr_loss=0.3947, over 3352927.79 frames. ], batch size: 89, lr: 1.71e-02, grad_scale: 32.0
2024-09-22 22:33:21,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer_ff3.min_abs, batch_count=117254.66666666667, ans=0.2
2024-09-22 22:33:35,890 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:34:02,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=117348.0, ans=0.125
2024-09-22 22:34:18,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=117394.66666666667, ans=0.1
2024-09-22 22:34:22,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=117394.66666666667, ans=0.5
2024-09-22 22:34:26,399 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.59 vs. limit=15.0
2024-09-22 22:34:35,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=117441.33333333333, ans=0.125
2024-09-22 22:34:45,735 INFO [train.py:1198] (1/4) Epoch 7, batch 1800, loss[loss=0.3128, ctc_loss=0.2242, cr_loss=0.443, over 17007.00 frames. ], tot_loss[loss=0.2737, ctc_loss=0.1949, cr_loss=0.3938, over 3347848.67 frames. ], batch size: 51, lr: 1.71e-02, grad_scale: 32.0
2024-09-22 22:35:20,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=117581.33333333333, ans=0.1
2024-09-22 22:35:33,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=117628.0, ans=0.125
2024-09-22 22:35:39,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=117628.0, ans=0.0
2024-09-22 22:35:44,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=117628.0, ans=0.125
2024-09-22 22:35:50,922 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.391e+02 1.562e+02 1.960e+02 3.440e+02, threshold=3.125e+02, percent-clipped=2.0
2024-09-22 22:35:55,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=117674.66666666667, ans=0.1
2024-09-22 22:36:02,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=117674.66666666667, ans=0.125
2024-09-22 22:36:05,278 INFO [train.py:1198] (1/4) Epoch 7, batch 1850, loss[loss=0.2334, ctc_loss=0.167, cr_loss=0.332, over 17262.00 frames. ], tot_loss[loss=0.2729, ctc_loss=0.1943, cr_loss=0.3928, over 3363198.87 frames. ], batch size: 44, lr: 1.71e-02, grad_scale: 32.0
2024-09-22 22:36:08,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=117721.33333333333, ans=0.125
2024-09-22 22:36:29,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2.whitening_limit, batch_count=117768.0, ans=15.0
2024-09-22 22:36:42,287 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.11 vs. limit=6.0
2024-09-22 22:36:55,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=117814.66666666667, ans=0.2
2024-09-22 22:37:04,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=117861.33333333333, ans=0.125
2024-09-22 22:37:29,663 INFO [train.py:1198] (1/4) Epoch 7, batch 1900, loss[loss=0.2151, ctc_loss=0.1521, cr_loss=0.3154, over 17256.00 frames. ], tot_loss[loss=0.2743, ctc_loss=0.1954, cr_loss=0.3943, over 3356419.97 frames. ], batch size: 42, lr: 1.70e-02, grad_scale: 32.0
2024-09-22 22:37:46,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.54 vs. limit=15.0
2024-09-22 22:37:58,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=118001.33333333333, ans=0.125
2024-09-22 22:38:22,565 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=18.10 vs. limit=22.5
2024-09-22 22:38:37,122 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.462e+02 1.627e+02 1.880e+02 2.641e+02, threshold=3.255e+02, percent-clipped=0.0
2024-09-22 22:38:51,397 INFO [train.py:1198] (1/4) Epoch 7, batch 1950, loss[loss=0.3387, ctc_loss=0.2443, cr_loss=0.4722, over 15004.00 frames. ], tot_loss[loss=0.2748, ctc_loss=0.1958, cr_loss=0.3953, over 3360136.70 frames. ], batch size: 89, lr: 1.70e-02, grad_scale: 32.0
2024-09-22 22:39:10,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=118234.66666666667, ans=0.125
2024-09-22 22:39:12,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=118234.66666666667, ans=10.0
2024-09-22 22:39:15,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=118234.66666666667, ans=0.1
2024-09-22 22:39:23,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=118281.33333333333, ans=0.1
2024-09-22 22:39:26,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer_na.min_abs, batch_count=118281.33333333333, ans=0.02
2024-09-22 22:39:29,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=118281.33333333333, ans=0.0
2024-09-22 22:39:37,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=118281.33333333333, ans=0.0
2024-09-22 22:39:37,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.01 vs. limit=15.0
2024-09-22 22:39:37,796 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.42 vs. limit=15.0
2024-09-22 22:39:44,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.27 vs. limit=22.5
2024-09-22 22:40:11,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=118421.33333333333, ans=0.125
2024-09-22 22:40:13,182 INFO [train.py:1198] (1/4) Epoch 7, batch 2000, loss[loss=0.2476, ctc_loss=0.1736, cr_loss=0.37, over 17093.00 frames. ], tot_loss[loss=0.2749, ctc_loss=0.1958, cr_loss=0.3954, over 3352468.58 frames. ], batch size: 43, lr: 1.70e-02, grad_scale: 32.0
2024-09-22 22:40:13,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=118421.33333333333, ans=0.125
2024-09-22 22:40:56,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=118514.66666666667, ans=0.0
2024-09-22 22:40:56,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=118514.66666666667, ans=0.125
2024-09-22 22:41:21,329 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.357e+02 1.496e+02 1.687e+02 2.654e+02, threshold=2.993e+02, percent-clipped=0.0
2024-09-22 22:41:35,547 INFO [train.py:1198] (1/4) Epoch 7, batch 2050, loss[loss=0.2976, ctc_loss=0.2153, cr_loss=0.4115, over 16871.00 frames. ], tot_loss[loss=0.275, ctc_loss=0.1959, cr_loss=0.3954, over 3350358.58 frames. ], batch size: 58, lr: 1.70e-02, grad_scale: 32.0
2024-09-22 22:41:44,254 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.64 vs. limit=10.0
2024-09-22 22:42:28,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=118794.66666666667, ans=0.125
2024-09-22 22:42:28,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.89 vs. limit=22.5
2024-09-22 22:42:42,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=118841.33333333333, ans=0.05
2024-09-22 22:42:42,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=118841.33333333333, ans=0.125
2024-09-22 22:42:58,045 INFO [train.py:1198] (1/4) Epoch 7, batch 2100, loss[loss=0.2379, ctc_loss=0.1698, cr_loss=0.3403, over 16937.00 frames. ], tot_loss[loss=0.2752, ctc_loss=0.196, cr_loss=0.396, over 3356143.86 frames. ], batch size: 42, lr: 1.70e-02, grad_scale: 32.0
2024-09-22 22:43:36,433 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.45 vs. limit=15.0
2024-09-22 22:43:42,580 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.56 vs. limit=15.0
2024-09-22 22:44:06,953 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.431e+02 1.548e+02 1.785e+02 2.946e+02, threshold=3.097e+02, percent-clipped=0.0
2024-09-22 22:44:18,976 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.00 vs. limit=6.0
2024-09-22 22:44:19,896 INFO [train.py:1198] (1/4) Epoch 7, batch 2150, loss[loss=0.2969, ctc_loss=0.2174, cr_loss=0.3976, over 17235.00 frames. ], tot_loss[loss=0.2774, ctc_loss=0.1978, cr_loss=0.3978, over 3360691.96 frames. ], batch size: 55, lr: 1.70e-02, grad_scale: 16.0
2024-09-22 22:44:29,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=119121.33333333333, ans=0.125
2024-09-22 22:44:33,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=119121.33333333333, ans=0.025
2024-09-22 22:44:40,049 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:44:59,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=119214.66666666667, ans=0.04949747468305833
2024-09-22 22:45:16,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=119261.33333333333, ans=0.0
2024-09-22 22:45:22,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=119261.33333333333, ans=0.09899494936611666
2024-09-22 22:45:29,181 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.84 vs. limit=15.0
2024-09-22 22:45:41,316 INFO [train.py:1198] (1/4) Epoch 7, batch 2200, loss[loss=0.2605, ctc_loss=0.1815, cr_loss=0.395, over 17159.00 frames. ], tot_loss[loss=0.2769, ctc_loss=0.1974, cr_loss=0.3972, over 3359516.63 frames. ], batch size: 41, lr: 1.69e-02, grad_scale: 16.0
2024-09-22 22:46:01,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=119401.33333333333, ans=0.125
2024-09-22 22:46:09,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=119401.33333333333, ans=10.0
2024-09-22 22:46:13,503 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.37 vs. limit=15.0
2024-09-22 22:46:29,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=119448.0, ans=0.09899494936611666
2024-09-22 22:46:40,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=119494.66666666667, ans=0.0
2024-09-22 22:46:53,571 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.357e+02 1.455e+02 1.682e+02 2.486e+02, threshold=2.909e+02, percent-clipped=0.0
2024-09-22 22:46:55,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=119541.33333333333, ans=0.125
2024-09-22 22:47:06,285 INFO [train.py:1198] (1/4) Epoch 7, batch 2250, loss[loss=0.2516, ctc_loss=0.1801, cr_loss=0.3577, over 17009.00 frames. ], tot_loss[loss=0.2761, ctc_loss=0.1968, cr_loss=0.3968, over 3359382.49 frames. ], batch size: 44, lr: 1.69e-02, grad_scale: 16.0
2024-09-22 22:47:19,363 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.90 vs. limit=22.5
2024-09-22 22:47:19,432 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.83 vs. limit=15.0
2024-09-22 22:47:28,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=119634.66666666667, ans=0.125
2024-09-22 22:47:51,628 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.10 vs. limit=15.0
2024-09-22 22:48:03,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=119728.0, ans=0.1
2024-09-22 22:48:16,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=119774.66666666667, ans=0.0
2024-09-22 22:48:25,672 INFO [train.py:1198] (1/4) Epoch 7, batch 2300, loss[loss=0.2694, ctc_loss=0.1882, cr_loss=0.4061, over 17309.00 frames. ], tot_loss[loss=0.2752, ctc_loss=0.1961, cr_loss=0.3958, over 3366122.92 frames. ], batch size: 51, lr: 1.69e-02, grad_scale: 16.0
2024-09-22 22:48:39,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=119821.33333333333, ans=10.0
2024-09-22 22:48:44,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=119868.0, ans=0.025
2024-09-22 22:49:00,826 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.65 vs. limit=22.5
2024-09-22 22:49:37,565 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 1.381e+02 1.499e+02 1.768e+02 3.628e+02, threshold=2.997e+02, percent-clipped=2.0
2024-09-22 22:49:48,303 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.85 vs. limit=15.0
2024-09-22 22:49:50,250 INFO [train.py:1198] (1/4) Epoch 7, batch 2350, loss[loss=0.2891, ctc_loss=0.2093, cr_loss=0.3986, over 17216.00 frames. ], tot_loss[loss=0.275, ctc_loss=0.196, cr_loss=0.3953, over 3359289.56 frames. ], batch size: 55, lr: 1.69e-02, grad_scale: 16.0
2024-09-22 22:50:01,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=120054.66666666667, ans=0.0
2024-09-22 22:50:01,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=120054.66666666667, ans=0.125
2024-09-22 22:50:01,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=120054.66666666667, ans=0.125
2024-09-22 22:50:25,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=120148.0, ans=0.04949747468305833
2024-09-22 22:50:53,660 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=6.70 vs. limit=15.0
2024-09-22 22:51:11,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=120288.0, ans=0.125
2024-09-22 22:51:12,229 INFO [train.py:1198] (1/4) Epoch 7, batch 2400, loss[loss=0.2795, ctc_loss=0.1936, cr_loss=0.4295, over 17146.00 frames. ], tot_loss[loss=0.2749, ctc_loss=0.1958, cr_loss=0.3952, over 3354955.54 frames. ], batch size: 48, lr: 1.69e-02, grad_scale: 32.0
2024-09-22 22:51:24,660 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=7.95 vs. limit=15.0
2024-09-22 22:51:28,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=120334.66666666667, ans=0.025
2024-09-22 22:51:38,605 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.07 vs. limit=22.5
2024-09-22 22:52:16,134 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.37 vs. limit=12.0
2024-09-22 22:52:21,490 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.376e+02 1.485e+02 1.691e+02 2.822e+02, threshold=2.971e+02, percent-clipped=0.0
2024-09-22 22:52:21,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=120474.66666666667, ans=0.0
2024-09-22 22:52:25,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.52 vs. limit=15.0
2024-09-22 22:52:34,407 INFO [train.py:1198] (1/4) Epoch 7, batch 2450, loss[loss=0.2795, ctc_loss=0.1998, cr_loss=0.3984, over 17007.00 frames. ], tot_loss[loss=0.2763, ctc_loss=0.1968, cr_loss=0.3973, over 3345629.72 frames. ], batch size: 53, lr: 1.69e-02, grad_scale: 32.0
2024-09-22 22:52:39,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=120521.33333333333, ans=0.2
2024-09-22 22:52:45,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=120521.33333333333, ans=0.07
2024-09-22 22:53:01,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=120568.0, ans=0.04949747468305833
2024-09-22 22:53:03,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=120568.0, ans=10.0
2024-09-22 22:53:08,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=120614.66666666667, ans=0.0
2024-09-22 22:53:12,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=120614.66666666667, ans=0.125
2024-09-22 22:53:42,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=120708.0, ans=0.125
2024-09-22 22:53:56,719 INFO [train.py:1198] (1/4) Epoch 7, batch 2500, loss[loss=0.3063, ctc_loss=0.2164, cr_loss=0.4496, over 17363.00 frames. ], tot_loss[loss=0.2771, ctc_loss=0.1974, cr_loss=0.3989, over 3356370.05 frames. ], batch size: 48, lr: 1.69e-02, grad_scale: 32.0
2024-09-22 22:54:01,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=120754.66666666667, ans=0.0
2024-09-22 22:54:09,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=120754.66666666667, ans=0.0
2024-09-22 22:54:13,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=120801.33333333333, ans=0.1
2024-09-22 22:54:13,669 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.73 vs. limit=15.0
2024-09-22 22:54:25,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=120801.33333333333, ans=0.125
2024-09-22 22:55:06,269 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.238e+02 1.457e+02 1.729e+02 2.020e+02 3.233e+02, threshold=3.458e+02, percent-clipped=3.0
2024-09-22 22:55:08,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=120941.33333333333, ans=0.0
2024-09-22 22:55:18,951 INFO [train.py:1198] (1/4) Epoch 7, batch 2550, loss[loss=0.2486, ctc_loss=0.1736, cr_loss=0.3747, over 17119.00 frames. ], tot_loss[loss=0.275, ctc_loss=0.1958, cr_loss=0.396, over 3354878.88 frames. ], batch size: 40, lr: 1.68e-02, grad_scale: 32.0
2024-09-22 22:55:42,148 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.34 vs. limit=15.0
2024-09-22 22:55:47,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=121034.66666666667, ans=0.125
2024-09-22 22:55:58,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=121081.33333333333, ans=0.0
2024-09-22 22:56:00,000 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.11 vs. limit=22.5
2024-09-22 22:56:40,342 INFO [train.py:1198] (1/4) Epoch 7, batch 2600, loss[loss=0.2995, ctc_loss=0.2146, cr_loss=0.4244, over 17221.00 frames. ], tot_loss[loss=0.2752, ctc_loss=0.196, cr_loss=0.396, over 3358765.27 frames. ], batch size: 55, lr: 1.68e-02, grad_scale: 32.0
2024-09-22 22:56:49,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=121221.33333333333, ans=0.2
2024-09-22 22:57:03,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=121268.0, ans=0.0
2024-09-22 22:57:14,989 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.30 vs. limit=10.0
2024-09-22 22:57:30,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=121361.33333333333, ans=0.2
2024-09-22 22:57:40,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=121361.33333333333, ans=0.125
2024-09-22 22:57:49,627 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.441e+02 1.588e+02 1.825e+02 2.905e+02, threshold=3.176e+02, percent-clipped=0.0
2024-09-22 22:57:50,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=121408.0, ans=0.125
2024-09-22 22:58:02,494 INFO [train.py:1198] (1/4) Epoch 7, batch 2650, loss[loss=0.2587, ctc_loss=0.1818, cr_loss=0.3842, over 17255.00 frames. ], tot_loss[loss=0.2728, ctc_loss=0.194, cr_loss=0.3937, over 3364505.67 frames. ], batch size: 44, lr: 1.68e-02, grad_scale: 32.0
2024-09-22 22:58:07,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=121454.66666666667, ans=0.125
2024-09-22 22:58:21,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=121501.33333333333, ans=0.0
2024-09-22 22:58:29,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=17.03 vs. limit=22.5
2024-09-22 22:58:38,775 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-22 22:58:42,064 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.22 vs. limit=6.0
2024-09-22 22:59:10,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=121641.33333333333, ans=0.0
2024-09-22 22:59:18,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=121641.33333333333, ans=0.0
2024-09-22 22:59:20,005 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=13.39 vs. limit=22.5
2024-09-22 22:59:27,152 INFO [train.py:1198] (1/4) Epoch 7, batch 2700, loss[loss=0.281, ctc_loss=0.2057, cr_loss=0.3761, over 17162.00 frames. ], tot_loss[loss=0.2736, ctc_loss=0.1947, cr_loss=0.3948, over 3363759.80 frames. ], batch size: 48, lr: 1.68e-02, grad_scale: 32.0
2024-09-22 22:59:29,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=121688.0, ans=0.125
2024-09-22 22:59:39,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.54 vs. limit=6.0
2024-09-22 23:00:32,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=121874.66666666667, ans=0.2
2024-09-22 23:00:33,554 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.410e+02 1.530e+02 1.699e+02 3.124e+02, threshold=3.060e+02, percent-clipped=0.0
2024-09-22 23:00:36,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=121874.66666666667, ans=0.1
2024-09-22 23:00:38,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=121874.66666666667, ans=0.0
2024-09-22 23:00:40,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=121874.66666666667, ans=0.0
2024-09-22 23:00:48,618 INFO [train.py:1198] (1/4) Epoch 7, batch 2750, loss[loss=0.2273, ctc_loss=0.1592, cr_loss=0.3401, over 16688.00 frames. ], tot_loss[loss=0.2747, ctc_loss=0.1955, cr_loss=0.3958, over 3364939.02 frames. ], batch size: 37, lr: 1.68e-02, grad_scale: 32.0
2024-09-22 23:00:49,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.74 vs. limit=6.0
2024-09-22 23:01:07,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=121968.0, ans=0.125
2024-09-22 23:01:08,797 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.07 vs. limit=8.0
2024-09-22 23:01:14,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=121968.0, ans=0.125
2024-09-22 23:01:17,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=121968.0, ans=0.0
2024-09-22 23:01:18,270 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.59 vs. limit=15.0
2024-09-22 23:01:26,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=122014.66666666667, ans=0.125
2024-09-22 23:01:27,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.58 vs. limit=15.0
2024-09-22 23:01:30,829 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.77 vs. limit=12.0
2024-09-22 23:02:03,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=122108.0, ans=0.125
2024-09-22 23:02:10,731 INFO [train.py:1198] (1/4) Epoch 7, batch 2800, loss[loss=0.3008, ctc_loss=0.215, cr_loss=0.4288, over 17306.00 frames. ], tot_loss[loss=0.2743, ctc_loss=0.1952, cr_loss=0.3957, over 3371134.82 frames. ], batch size: 51, lr: 1.68e-02, grad_scale: 32.0
2024-09-22 23:03:08,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=122294.66666666667, ans=0.05
2024-09-22 23:03:18,221 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.349e+02 1.457e+02 1.581e+02 2.828e+02, threshold=2.913e+02, percent-clipped=0.0
2024-09-22 23:03:33,472 INFO [train.py:1198] (1/4) Epoch 7, batch 2850, loss[loss=0.2673, ctc_loss=0.1908, cr_loss=0.3828, over 17010.00 frames. ], tot_loss[loss=0.274, ctc_loss=0.1949, cr_loss=0.3954, over 3359876.49 frames. ], batch size: 44, lr: 1.67e-02, grad_scale: 32.0
2024-09-22 23:03:46,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=122388.0, ans=0.95
2024-09-22 23:04:03,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=122481.33333333333, ans=0.125
2024-09-22 23:04:03,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=122481.33333333333, ans=0.0
2024-09-22 23:04:55,303 INFO [train.py:1198] (1/4) Epoch 7, batch 2900, loss[loss=0.2735, ctc_loss=0.19, cr_loss=0.4176, over 17179.00 frames. ], tot_loss[loss=0.2743, ctc_loss=0.1952, cr_loss=0.3957, over 3357099.24 frames. ], batch size: 41, lr: 1.67e-02, grad_scale: 16.0
2024-09-22 23:04:57,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.57 vs. limit=22.5
2024-09-22 23:05:25,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=122714.66666666667, ans=0.0
2024-09-22 23:05:33,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=122714.66666666667, ans=0.125
2024-09-22 23:05:36,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=122714.66666666667, ans=0.0
2024-09-22 23:06:05,781 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.397e+02 1.534e+02 1.877e+02 3.701e+02, threshold=3.067e+02, percent-clipped=2.0
2024-09-22 23:06:07,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=122808.0, ans=0.125
2024-09-22 23:06:15,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=122854.66666666667, ans=0.1
2024-09-22 23:06:16,923 INFO [train.py:1198] (1/4) Epoch 7, batch 2950, loss[loss=0.2904, ctc_loss=0.2115, cr_loss=0.3946, over 15820.00 frames. ], tot_loss[loss=0.2743, ctc_loss=0.1953, cr_loss=0.3953, over 3355702.47 frames. ], batch size: 74, lr: 1.67e-02, grad_scale: 16.0
2024-09-22 23:06:25,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.10 vs. limit=15.0
2024-09-22 23:06:28,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=122854.66666666667, ans=0.125
2024-09-22 23:06:29,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=122854.66666666667, ans=0.125
2024-09-22 23:06:52,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2.whitening_limit, batch_count=122948.0, ans=15.0
2024-09-22 23:07:06,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.12 vs. limit=15.0
2024-09-22 23:07:17,472 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.54 vs. limit=15.0
2024-09-22 23:07:25,187 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.91 vs. limit=15.0
2024-09-22 23:07:38,710 INFO [train.py:1198] (1/4) Epoch 7, batch 3000, loss[loss=0.2593, ctc_loss=0.1833, cr_loss=0.38, over 17211.00 frames. ], tot_loss[loss=0.2743, ctc_loss=0.1952, cr_loss=0.3954, over 3356304.77 frames. ], batch size: 47, lr: 1.67e-02, grad_scale: 16.0
2024-09-22 23:07:38,711 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 23:07:54,144 INFO [train.py:1230] (1/4) Epoch 7, validation: loss=0.05688, ctc_loss=0.05688, cr_loss=7.669e-15, over 944034.00 frames. 
2024-09-22 23:07:54,145 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 23:07:59,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=123088.0, ans=0.2
2024-09-22 23:08:08,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=123134.66666666667, ans=0.125
2024-09-22 23:08:19,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=123134.66666666667, ans=0.2
2024-09-22 23:08:49,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=123228.0, ans=0.125
2024-09-22 23:09:01,890 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.381e+02 1.504e+02 1.827e+02 5.428e+02, threshold=3.008e+02, percent-clipped=9.0
2024-09-22 23:09:12,840 INFO [train.py:1198] (1/4) Epoch 7, batch 3050, loss[loss=0.2422, ctc_loss=0.1701, cr_loss=0.3606, over 17202.00 frames. ], tot_loss[loss=0.2716, ctc_loss=0.1931, cr_loss=0.3927, over 3366520.19 frames. ], batch size: 41, lr: 1.67e-02, grad_scale: 16.0
2024-09-22 23:10:08,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=123461.33333333333, ans=0.2
2024-09-22 23:10:16,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=123508.0, ans=0.125
2024-09-22 23:10:33,468 INFO [train.py:1198] (1/4) Epoch 7, batch 3100, loss[loss=0.2791, ctc_loss=0.1988, cr_loss=0.4017, over 16989.00 frames. ], tot_loss[loss=0.2724, ctc_loss=0.1936, cr_loss=0.3939, over 3367518.42 frames. ], batch size: 53, lr: 1.67e-02, grad_scale: 16.0
2024-09-22 23:10:36,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=123554.66666666667, ans=0.2
2024-09-22 23:10:49,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=123601.33333333333, ans=0.0
2024-09-22 23:10:52,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=123601.33333333333, ans=0.0
2024-09-22 23:10:57,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=123601.33333333333, ans=0.0
2024-09-22 23:11:41,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=123741.33333333333, ans=0.2
2024-09-22 23:11:42,860 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.455e+02 1.593e+02 1.857e+02 3.038e+02, threshold=3.186e+02, percent-clipped=1.0
2024-09-22 23:11:46,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=123741.33333333333, ans=0.0
2024-09-22 23:11:53,642 INFO [train.py:1198] (1/4) Epoch 7, batch 3150, loss[loss=0.2803, ctc_loss=0.2005, cr_loss=0.3989, over 17227.00 frames. ], tot_loss[loss=0.2719, ctc_loss=0.1932, cr_loss=0.3935, over 3365750.00 frames. ], batch size: 47, lr: 1.67e-02, grad_scale: 16.0
2024-09-22 23:13:12,037 INFO [train.py:1198] (1/4) Epoch 7, batch 3200, loss[loss=0.2896, ctc_loss=0.2093, cr_loss=0.4015, over 14842.00 frames. ], tot_loss[loss=0.272, ctc_loss=0.1932, cr_loss=0.394, over 3368708.41 frames. ], batch size: 90, lr: 1.66e-02, grad_scale: 32.0
2024-09-22 23:13:33,053 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.44 vs. limit=15.0
2024-09-22 23:13:40,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=124068.0, ans=0.125
2024-09-22 23:13:57,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=124161.33333333333, ans=0.125
2024-09-22 23:13:59,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=124161.33333333333, ans=0.125
2024-09-22 23:14:19,024 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.425e+02 1.561e+02 1.765e+02 3.696e+02, threshold=3.122e+02, percent-clipped=1.0
2024-09-22 23:14:29,938 INFO [train.py:1198] (1/4) Epoch 7, batch 3250, loss[loss=0.2598, ctc_loss=0.1809, cr_loss=0.3946, over 17177.00 frames. ], tot_loss[loss=0.2721, ctc_loss=0.1933, cr_loss=0.3941, over 3360598.37 frames. ], batch size: 41, lr: 1.66e-02, grad_scale: 32.0
2024-09-22 23:14:44,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=124254.66666666667, ans=0.05
2024-09-22 23:15:03,938 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.42 vs. limit=15.0
2024-09-22 23:15:06,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=124348.0, ans=0.125
2024-09-22 23:15:16,429 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.27 vs. limit=15.0
2024-09-22 23:15:39,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=124441.33333333333, ans=0.0
2024-09-22 23:15:49,579 INFO [train.py:1198] (1/4) Epoch 7, batch 3300, loss[loss=0.2856, ctc_loss=0.2011, cr_loss=0.4226, over 17049.00 frames. ], tot_loss[loss=0.2707, ctc_loss=0.1922, cr_loss=0.3928, over 3354931.44 frames. ], batch size: 46, lr: 1.66e-02, grad_scale: 32.0
2024-09-22 23:15:56,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=124488.0, ans=0.125
2024-09-22 23:16:11,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=124534.66666666667, ans=0.1
2024-09-22 23:16:15,415 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.33 vs. limit=15.0
2024-09-22 23:16:22,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=124581.33333333333, ans=0.025
2024-09-22 23:16:42,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=124628.0, ans=0.125
2024-09-22 23:16:58,143 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.418e+02 1.603e+02 1.871e+02 3.678e+02, threshold=3.206e+02, percent-clipped=3.0
2024-09-22 23:17:09,115 INFO [train.py:1198] (1/4) Epoch 7, batch 3350, loss[loss=0.2488, ctc_loss=0.1797, cr_loss=0.3453, over 17257.00 frames. ], tot_loss[loss=0.2714, ctc_loss=0.1928, cr_loss=0.393, over 3352856.71 frames. ], batch size: 44, lr: 1.66e-02, grad_scale: 32.0
2024-09-22 23:17:18,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=124721.33333333333, ans=0.2
2024-09-22 23:17:20,842 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.30 vs. limit=15.0
2024-09-22 23:17:43,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=124814.66666666667, ans=0.09899494936611666
2024-09-22 23:17:58,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=124861.33333333333, ans=0.125
2024-09-22 23:18:16,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=124908.0, ans=0.125
2024-09-22 23:18:16,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=124908.0, ans=0.125
2024-09-22 23:18:25,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=124954.66666666667, ans=0.0
2024-09-22 23:18:26,924 INFO [train.py:1198] (1/4) Epoch 7, batch 3400, loss[loss=0.2685, ctc_loss=0.1907, cr_loss=0.3891, over 16962.00 frames. ], tot_loss[loss=0.2722, ctc_loss=0.1933, cr_loss=0.3943, over 3351574.49 frames. ], batch size: 53, lr: 1.66e-02, grad_scale: 32.0
2024-09-22 23:18:33,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=124954.66666666667, ans=0.0
2024-09-22 23:18:33,679 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 23:18:45,077 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=15.53 vs. limit=22.5
2024-09-22 23:18:49,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=125001.33333333333, ans=0.125
2024-09-22 23:18:55,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=125001.33333333333, ans=0.125
2024-09-22 23:19:33,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=125141.33333333333, ans=0.1
2024-09-22 23:19:34,686 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.374e+02 1.483e+02 1.750e+02 2.564e+02, threshold=2.966e+02, percent-clipped=0.0
2024-09-22 23:19:36,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=125141.33333333333, ans=0.025
2024-09-22 23:19:45,700 INFO [train.py:1198] (1/4) Epoch 7, batch 3450, loss[loss=0.2404, ctc_loss=0.17, cr_loss=0.3518, over 17188.00 frames. ], tot_loss[loss=0.2741, ctc_loss=0.195, cr_loss=0.3957, over 3347974.68 frames. ], batch size: 41, lr: 1.66e-02, grad_scale: 32.0
2024-09-22 23:20:30,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=125281.33333333333, ans=0.125
2024-09-22 23:20:53,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=125374.66666666667, ans=0.125
2024-09-22 23:21:05,780 INFO [train.py:1198] (1/4) Epoch 7, batch 3500, loss[loss=0.2747, ctc_loss=0.1963, cr_loss=0.3923, over 17024.00 frames. ], tot_loss[loss=0.276, ctc_loss=0.1963, cr_loss=0.3985, over 3353900.88 frames. ], batch size: 51, lr: 1.66e-02, grad_scale: 32.0
2024-09-22 23:21:13,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=125421.33333333333, ans=0.125
2024-09-22 23:21:18,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=125421.33333333333, ans=0.1
2024-09-22 23:21:45,190 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 23:21:46,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=125514.66666666667, ans=0.125
2024-09-22 23:21:51,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=125514.66666666667, ans=0.0
2024-09-22 23:21:57,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=125561.33333333333, ans=0.125
2024-09-22 23:21:59,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=125561.33333333333, ans=0.125
2024-09-22 23:22:05,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=125561.33333333333, ans=0.0
2024-09-22 23:22:06,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=125561.33333333333, ans=0.125
2024-09-22 23:22:07,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=125561.33333333333, ans=0.125
2024-09-22 23:22:13,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=125608.0, ans=0.125
2024-09-22 23:22:14,393 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.377e+02 1.536e+02 1.902e+02 3.624e+02, threshold=3.071e+02, percent-clipped=2.0
2024-09-22 23:22:25,092 INFO [train.py:1198] (1/4) Epoch 7, batch 3550, loss[loss=0.2712, ctc_loss=0.1923, cr_loss=0.3948, over 17210.00 frames. ], tot_loss[loss=0.2756, ctc_loss=0.1961, cr_loss=0.3977, over 3347898.44 frames. ], batch size: 50, lr: 1.65e-02, grad_scale: 32.0
2024-09-22 23:22:49,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=125701.33333333333, ans=0.125
2024-09-22 23:22:51,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=125701.33333333333, ans=0.0
2024-09-22 23:22:57,621 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=125748.0, ans=0.125
2024-09-22 23:22:57,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=125748.0, ans=0.05
2024-09-22 23:23:23,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=125794.66666666667, ans=0.125
2024-09-22 23:23:39,969 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.03 vs. limit=15.0
2024-09-22 23:23:42,344 INFO [train.py:1198] (1/4) Epoch 7, batch 3600, loss[loss=0.2863, ctc_loss=0.2015, cr_loss=0.4241, over 16457.00 frames. ], tot_loss[loss=0.2744, ctc_loss=0.1951, cr_loss=0.3962, over 3344446.29 frames. ], batch size: 66, lr: 1.65e-02, grad_scale: 32.0
2024-09-22 23:23:55,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=125888.0, ans=0.0
2024-09-22 23:24:07,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=125934.66666666667, ans=0.125
2024-09-22 23:24:32,620 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-22 23:24:34,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=126028.0, ans=0.07
2024-09-22 23:24:46,926 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.05 vs. limit=15.0
2024-09-22 23:24:50,665 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.445e+02 1.618e+02 1.998e+02 3.129e+02, threshold=3.236e+02, percent-clipped=1.0
2024-09-22 23:24:50,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=126074.66666666667, ans=0.125
2024-09-22 23:24:50,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=126074.66666666667, ans=0.1
2024-09-22 23:25:01,498 INFO [train.py:1198] (1/4) Epoch 7, batch 3650, loss[loss=0.2941, ctc_loss=0.2104, cr_loss=0.4182, over 15922.00 frames. ], tot_loss[loss=0.2744, ctc_loss=0.1951, cr_loss=0.3964, over 3346921.83 frames. ], batch size: 74, lr: 1.65e-02, grad_scale: 32.0
2024-09-22 23:25:31,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=126168.0, ans=0.125
2024-09-22 23:25:58,344 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.09 vs. limit=15.0
2024-09-22 23:26:08,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=126308.0, ans=0.1
2024-09-22 23:26:19,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=126308.0, ans=0.125
2024-09-22 23:26:22,288 INFO [train.py:1198] (1/4) Epoch 7, batch 3700, loss[loss=0.2798, ctc_loss=0.196, cr_loss=0.4191, over 17218.00 frames. ], tot_loss[loss=0.2736, ctc_loss=0.1945, cr_loss=0.3956, over 3353341.48 frames. ], batch size: 47, lr: 1.65e-02, grad_scale: 32.0
2024-09-22 23:26:27,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=126354.66666666667, ans=0.125
2024-09-22 23:27:19,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=126494.66666666667, ans=0.2
2024-09-22 23:27:25,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=126541.33333333333, ans=0.125
2024-09-22 23:27:30,062 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.388e+02 1.512e+02 1.895e+02 2.751e+02, threshold=3.024e+02, percent-clipped=0.0
2024-09-22 23:27:30,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=126541.33333333333, ans=0.125
2024-09-22 23:27:36,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=126541.33333333333, ans=0.0
2024-09-22 23:27:41,022 INFO [train.py:1198] (1/4) Epoch 7, batch 3750, loss[loss=0.3007, ctc_loss=0.2154, cr_loss=0.4266, over 17216.00 frames. ], tot_loss[loss=0.2745, ctc_loss=0.1953, cr_loss=0.3963, over 3343983.02 frames. ], batch size: 50, lr: 1.65e-02, grad_scale: 32.0
2024-09-22 23:27:46,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=126588.0, ans=0.125
2024-09-22 23:27:50,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=126588.0, ans=0.125
2024-09-22 23:28:06,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=126634.66666666667, ans=0.07
2024-09-22 23:28:33,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.99 vs. limit=15.0
2024-09-22 23:28:56,778 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.71 vs. limit=10.0
2024-09-22 23:28:59,018 INFO [train.py:1198] (1/4) Epoch 7, batch 3800, loss[loss=0.2804, ctc_loss=0.1999, cr_loss=0.4023, over 15934.00 frames. ], tot_loss[loss=0.2745, ctc_loss=0.1956, cr_loss=0.3949, over 3321925.16 frames. ], batch size: 74, lr: 1.65e-02, grad_scale: 32.0
2024-09-22 23:29:07,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=126821.33333333333, ans=0.2
2024-09-22 23:29:12,398 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.29 vs. limit=15.0
2024-09-22 23:29:22,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=126868.0, ans=0.0
2024-09-22 23:29:25,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=126868.0, ans=0.04949747468305833
2024-09-22 23:29:33,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=126914.66666666667, ans=0.1
2024-09-22 23:29:33,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=126914.66666666667, ans=0.1
2024-09-22 23:29:35,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=126914.66666666667, ans=0.0
2024-09-22 23:29:35,913 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 23:29:41,381 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.83 vs. limit=15.0
2024-09-22 23:29:47,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=126961.33333333333, ans=0.125
2024-09-22 23:29:48,967 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.01 vs. limit=15.0
2024-09-22 23:30:07,020 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.460e+02 1.621e+02 1.904e+02 2.959e+02, threshold=3.241e+02, percent-clipped=0.0
2024-09-22 23:30:17,775 INFO [train.py:1198] (1/4) Epoch 7, batch 3850, loss[loss=0.3008, ctc_loss=0.2156, cr_loss=0.4258, over 16942.00 frames. ], tot_loss[loss=0.2777, ctc_loss=0.1983, cr_loss=0.3972, over 3288731.50 frames. ], batch size: 58, lr: 1.65e-02, grad_scale: 32.0
2024-09-22 23:31:05,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=9.49 vs. limit=15.0
2024-09-22 23:31:18,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=127241.33333333333, ans=0.125
2024-09-22 23:32:20,222 INFO [train.py:1198] (1/4) Epoch 8, batch 0, loss[loss=0.2738, ctc_loss=0.1965, cr_loss=0.3864, over 16790.00 frames. ], tot_loss[loss=0.2738, ctc_loss=0.1965, cr_loss=0.3864, over 16790.00 frames. ], batch size: 61, lr: 1.55e-02, grad_scale: 32.0
2024-09-22 23:32:20,223 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-22 23:32:35,554 INFO [train.py:1230] (1/4) Epoch 8, validation: loss=0.05692, ctc_loss=0.05692, cr_loss=7.316e-15, over 944034.00 frames. 
2024-09-22 23:32:35,555 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-22 23:33:12,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=127362.66666666667, ans=0.1
2024-09-22 23:33:25,909 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.67 vs. limit=22.5
2024-09-22 23:33:38,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=127456.0, ans=0.0
2024-09-22 23:33:48,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=127456.0, ans=0.0
2024-09-22 23:33:52,393 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.394e+02 1.752e+02 2.098e+02 6.301e+02, threshold=3.504e+02, percent-clipped=3.0
2024-09-22 23:33:55,628 INFO [train.py:1198] (1/4) Epoch 8, batch 50, loss[loss=0.3705, ctc_loss=0.2694, cr_loss=0.5054, over 11538.00 frames. ], tot_loss[loss=0.275, ctc_loss=0.1956, cr_loss=0.3972, over 759277.33 frames. ], batch size: 123, lr: 1.55e-02, grad_scale: 32.0
2024-09-22 23:35:11,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.13 vs. limit=22.5
2024-09-22 23:35:19,391 INFO [train.py:1198] (1/4) Epoch 8, batch 100, loss[loss=0.292, ctc_loss=0.209, cr_loss=0.4151, over 16781.00 frames. ], tot_loss[loss=0.2735, ctc_loss=0.1945, cr_loss=0.3951, over 1332088.48 frames. ], batch size: 61, lr: 1.55e-02, grad_scale: 32.0
2024-09-22 23:35:19,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=127736.0, ans=0.0
2024-09-22 23:35:46,029 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.18 vs. limit=5.0
2024-09-22 23:36:17,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=127876.0, ans=0.2
2024-09-22 23:36:28,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=127922.66666666667, ans=0.125
2024-09-22 23:36:33,009 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 23:36:37,479 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.179e+02 1.356e+02 1.486e+02 1.737e+02 3.200e+02, threshold=2.973e+02, percent-clipped=0.0
2024-09-22 23:36:40,697 INFO [train.py:1198] (1/4) Epoch 8, batch 150, loss[loss=0.2643, ctc_loss=0.1899, cr_loss=0.3718, over 16989.00 frames. ], tot_loss[loss=0.2714, ctc_loss=0.1925, cr_loss=0.3941, over 1780015.36 frames. ], batch size: 51, lr: 1.55e-02, grad_scale: 32.0
2024-09-22 23:36:54,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=127969.33333333333, ans=0.0
2024-09-22 23:37:19,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=128062.66666666667, ans=0.1
2024-09-22 23:37:25,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=128062.66666666667, ans=0.125
2024-09-22 23:37:36,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=128109.33333333333, ans=0.125
2024-09-22 23:37:46,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=128156.0, ans=0.1
2024-09-22 23:38:02,290 INFO [train.py:1198] (1/4) Epoch 8, batch 200, loss[loss=0.2614, ctc_loss=0.1799, cr_loss=0.4076, over 17211.00 frames. ], tot_loss[loss=0.2743, ctc_loss=0.1949, cr_loss=0.3968, over 2117823.43 frames. ], batch size: 50, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:38:37,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=128296.0, ans=0.0
2024-09-22 23:38:53,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=128342.66666666667, ans=0.125
2024-09-22 23:39:04,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=128389.33333333333, ans=0.2
2024-09-22 23:39:20,519 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.362e+02 1.519e+02 1.695e+02 2.654e+02, threshold=3.037e+02, percent-clipped=0.0
2024-09-22 23:39:23,709 INFO [train.py:1198] (1/4) Epoch 8, batch 250, loss[loss=0.3496, ctc_loss=0.2667, cr_loss=0.4143, over 11803.00 frames. ], tot_loss[loss=0.273, ctc_loss=0.1937, cr_loss=0.3965, over 2390488.13 frames. ], batch size: 123, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:39:24,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=128436.0, ans=0.125
2024-09-22 23:39:43,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=128482.66666666667, ans=0.1
2024-09-22 23:39:46,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=128482.66666666667, ans=0.2
2024-09-22 23:39:48,723 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.61 vs. limit=6.0
2024-09-22 23:40:10,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=128529.33333333333, ans=0.125
2024-09-22 23:40:27,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=128576.0, ans=0.0
2024-09-22 23:40:46,250 INFO [train.py:1198] (1/4) Epoch 8, batch 300, loss[loss=0.3239, ctc_loss=0.2449, cr_loss=0.3949, over 12271.00 frames. ], tot_loss[loss=0.2694, ctc_loss=0.1908, cr_loss=0.3932, over 2608086.36 frames. ], batch size: 123, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:41:08,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=128716.0, ans=0.0
2024-09-22 23:41:09,971 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-22 23:41:37,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=128809.33333333333, ans=0.0
2024-09-22 23:41:37,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=128809.33333333333, ans=0.0
2024-09-22 23:41:48,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=128809.33333333333, ans=0.1
2024-09-22 23:41:58,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=128856.0, ans=0.2
2024-09-22 23:42:03,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=128856.0, ans=0.125
2024-09-22 23:42:07,483 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.437e+02 1.612e+02 1.809e+02 3.626e+02, threshold=3.224e+02, percent-clipped=3.0
2024-09-22 23:42:10,765 INFO [train.py:1198] (1/4) Epoch 8, batch 350, loss[loss=0.2129, ctc_loss=0.1453, cr_loss=0.3384, over 17194.00 frames. ], tot_loss[loss=0.2701, ctc_loss=0.1913, cr_loss=0.3936, over 2767557.28 frames. ], batch size: 41, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:42:15,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=128902.66666666667, ans=0.125
2024-09-22 23:42:22,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=128902.66666666667, ans=0.125
2024-09-22 23:42:30,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=128949.33333333333, ans=0.125
2024-09-22 23:43:05,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=129042.66666666667, ans=10.0
2024-09-22 23:43:05,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=129042.66666666667, ans=0.125
2024-09-22 23:43:24,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=129089.33333333333, ans=0.125
2024-09-22 23:43:30,393 INFO [train.py:1198] (1/4) Epoch 8, batch 400, loss[loss=0.2611, ctc_loss=0.1827, cr_loss=0.3918, over 16993.00 frames. ], tot_loss[loss=0.2694, ctc_loss=0.1908, cr_loss=0.3929, over 2890915.00 frames. ], batch size: 56, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:43:56,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=129182.66666666667, ans=0.125
2024-09-22 23:44:32,691 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.75 vs. limit=10.0
2024-09-22 23:44:48,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=129322.66666666667, ans=0.0
2024-09-22 23:44:49,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 1.364e+02 1.519e+02 1.603e+02 2.870e+02, threshold=3.038e+02, percent-clipped=0.0
2024-09-22 23:44:52,858 INFO [train.py:1198] (1/4) Epoch 8, batch 450, loss[loss=0.2667, ctc_loss=0.1915, cr_loss=0.3763, over 16447.00 frames. ], tot_loss[loss=0.2696, ctc_loss=0.1909, cr_loss=0.3931, over 2993934.28 frames. ], batch size: 66, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:44:56,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=129369.33333333333, ans=0.125
2024-09-22 23:45:22,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=129416.0, ans=0.2
2024-09-22 23:45:23,392 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.78 vs. limit=15.0
2024-09-22 23:46:07,428 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=129556.0, ans=0.1
2024-09-22 23:46:18,033 INFO [train.py:1198] (1/4) Epoch 8, batch 500, loss[loss=0.2485, ctc_loss=0.1735, cr_loss=0.375, over 17033.00 frames. ], tot_loss[loss=0.2692, ctc_loss=0.1906, cr_loss=0.3932, over 3078147.20 frames. ], batch size: 39, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:46:49,375 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.47 vs. limit=15.0
2024-09-22 23:47:01,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=129696.0, ans=0.0
2024-09-22 23:47:30,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=129789.33333333333, ans=0.0
2024-09-22 23:47:36,030 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.377e+02 1.524e+02 1.726e+02 2.894e+02, threshold=3.047e+02, percent-clipped=0.0
2024-09-22 23:47:36,653 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.20 vs. limit=6.0
2024-09-22 23:47:39,314 INFO [train.py:1198] (1/4) Epoch 8, batch 550, loss[loss=0.2709, ctc_loss=0.1963, cr_loss=0.3734, over 17235.00 frames. ], tot_loss[loss=0.269, ctc_loss=0.1904, cr_loss=0.393, over 3135619.23 frames. ], batch size: 50, lr: 1.54e-02, grad_scale: 32.0
2024-09-22 23:48:00,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=129882.66666666667, ans=0.125
2024-09-22 23:48:01,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=129882.66666666667, ans=0.125
2024-09-22 23:48:07,383 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=6.33 vs. limit=15.0
2024-09-22 23:48:08,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=129882.66666666667, ans=0.125
2024-09-22 23:48:13,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=129929.33333333333, ans=0.025
2024-09-22 23:48:36,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=129976.0, ans=0.025
2024-09-22 23:48:44,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=130022.66666666667, ans=0.0
2024-09-22 23:48:49,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=130022.66666666667, ans=0.125
2024-09-22 23:48:58,803 INFO [train.py:1198] (1/4) Epoch 8, batch 600, loss[loss=0.3178, ctc_loss=0.2348, cr_loss=0.4151, over 15130.00 frames. ], tot_loss[loss=0.2696, ctc_loss=0.1907, cr_loss=0.3946, over 3184420.00 frames. ], batch size: 89, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:49:27,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=130116.0, ans=0.2
2024-09-22 23:49:54,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=130209.33333333333, ans=0.1
2024-09-22 23:50:04,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=130209.33333333333, ans=0.0
2024-09-22 23:50:09,719 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-22 23:50:11,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=130256.0, ans=0.125
2024-09-22 23:50:20,278 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.396e+02 1.531e+02 1.850e+02 5.586e+02, threshold=3.061e+02, percent-clipped=2.0
2024-09-22 23:50:23,504 INFO [train.py:1198] (1/4) Epoch 8, batch 650, loss[loss=0.3063, ctc_loss=0.219, cr_loss=0.4365, over 16619.00 frames. ], tot_loss[loss=0.2694, ctc_loss=0.1905, cr_loss=0.3947, over 3224993.70 frames. ], batch size: 66, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:50:27,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=130302.66666666667, ans=0.125
2024-09-22 23:50:43,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=130349.33333333333, ans=0.1
2024-09-22 23:51:09,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=130396.0, ans=0.125
2024-09-22 23:51:10,318 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.70 vs. limit=6.0
2024-09-22 23:51:48,929 INFO [train.py:1198] (1/4) Epoch 8, batch 700, loss[loss=0.2614, ctc_loss=0.1858, cr_loss=0.378, over 17010.00 frames. ], tot_loss[loss=0.2689, ctc_loss=0.1901, cr_loss=0.3939, over 3257023.55 frames. ], batch size: 44, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:52:02,610 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.07 vs. limit=12.0
2024-09-22 23:52:14,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=130582.66666666667, ans=0.125
2024-09-22 23:52:15,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=130582.66666666667, ans=0.04949747468305833
2024-09-22 23:52:16,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=130582.66666666667, ans=0.125
2024-09-22 23:52:19,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=130629.33333333333, ans=0.1
2024-09-22 23:52:32,452 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.53 vs. limit=15.0
2024-09-22 23:52:35,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=130629.33333333333, ans=0.125
2024-09-22 23:52:55,123 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.16 vs. limit=15.0
2024-09-22 23:52:56,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=130722.66666666667, ans=0.0
2024-09-22 23:52:59,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=130722.66666666667, ans=0.2
2024-09-22 23:53:07,234 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.368e+02 1.561e+02 1.822e+02 3.141e+02, threshold=3.123e+02, percent-clipped=1.0
2024-09-22 23:53:10,432 INFO [train.py:1198] (1/4) Epoch 8, batch 750, loss[loss=0.2757, ctc_loss=0.1939, cr_loss=0.4093, over 17089.00 frames. ], tot_loss[loss=0.2677, ctc_loss=0.1893, cr_loss=0.392, over 3272260.98 frames. ], batch size: 49, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:53:11,160 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.73 vs. limit=15.0
2024-09-22 23:53:32,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=130816.0, ans=0.125
2024-09-22 23:53:41,279 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.48 vs. limit=15.0
2024-09-22 23:53:53,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=130862.66666666667, ans=0.025
2024-09-22 23:54:08,189 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.75 vs. limit=12.0
2024-09-22 23:54:27,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=130956.0, ans=0.125
2024-09-22 23:54:33,312 INFO [train.py:1198] (1/4) Epoch 8, batch 800, loss[loss=0.2497, ctc_loss=0.1769, cr_loss=0.3638, over 17358.00 frames. ], tot_loss[loss=0.266, ctc_loss=0.188, cr_loss=0.3899, over 3290875.39 frames. ], batch size: 48, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:54:36,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=131002.66666666667, ans=0.1
2024-09-22 23:54:44,769 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.66 vs. limit=22.5
2024-09-22 23:54:46,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=131002.66666666667, ans=0.125
2024-09-22 23:55:10,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=131096.0, ans=0.025
2024-09-22 23:55:12,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=131096.0, ans=0.125
2024-09-22 23:55:54,938 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.424e+02 1.641e+02 1.911e+02 2.980e+02, threshold=3.282e+02, percent-clipped=0.0
2024-09-22 23:55:58,176 INFO [train.py:1198] (1/4) Epoch 8, batch 850, loss[loss=0.2459, ctc_loss=0.1688, cr_loss=0.3856, over 17291.00 frames. ], tot_loss[loss=0.2649, ctc_loss=0.1872, cr_loss=0.3883, over 3311382.85 frames. ], batch size: 49, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:56:04,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=131236.0, ans=0.0
2024-09-22 23:56:11,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=131236.0, ans=0.07
2024-09-22 23:56:29,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten.whitening_limit, batch_count=131282.66666666666, ans=15.0
2024-09-22 23:56:35,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=131329.33333333334, ans=0.025
2024-09-22 23:56:46,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=131329.33333333334, ans=0.1
2024-09-22 23:56:48,735 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.44 vs. limit=6.0
2024-09-22 23:57:08,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=131422.66666666666, ans=0.0
2024-09-22 23:57:21,043 INFO [train.py:1198] (1/4) Epoch 8, batch 900, loss[loss=0.2879, ctc_loss=0.2044, cr_loss=0.4177, over 16636.00 frames. ], tot_loss[loss=0.2652, ctc_loss=0.1874, cr_loss=0.3888, over 3327529.10 frames. ], batch size: 66, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:57:23,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=17.87 vs. limit=22.5
2024-09-22 23:57:26,536 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.25 vs. limit=15.0
2024-09-22 23:57:59,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=131562.66666666666, ans=0.0
2024-09-22 23:58:30,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=131656.0, ans=0.125
2024-09-22 23:58:37,889 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.417e+02 1.553e+02 1.784e+02 2.618e+02, threshold=3.106e+02, percent-clipped=0.0
2024-09-22 23:58:41,089 INFO [train.py:1198] (1/4) Epoch 8, batch 950, loss[loss=0.2187, ctc_loss=0.1543, cr_loss=0.3221, over 17165.00 frames. ], tot_loss[loss=0.2673, ctc_loss=0.1891, cr_loss=0.391, over 3328259.60 frames. ], batch size: 41, lr: 1.53e-02, grad_scale: 32.0
2024-09-22 23:59:01,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=131749.33333333334, ans=0.1
2024-09-22 23:59:18,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=131796.0, ans=0.2
2024-09-22 23:59:47,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=131889.33333333334, ans=0.2
2024-09-23 00:00:05,373 INFO [train.py:1198] (1/4) Epoch 8, batch 1000, loss[loss=0.365, ctc_loss=0.2706, cr_loss=0.4718, over 11718.00 frames. ], tot_loss[loss=0.267, ctc_loss=0.1889, cr_loss=0.3905, over 3330926.68 frames. ], batch size: 124, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:00:45,247 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.87 vs. limit=10.0
2024-09-23 00:01:08,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=132076.0, ans=0.0
2024-09-23 00:01:26,819 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.410e+02 1.544e+02 1.747e+02 2.558e+02, threshold=3.087e+02, percent-clipped=0.0
2024-09-23 00:01:29,992 INFO [train.py:1198] (1/4) Epoch 8, batch 1050, loss[loss=0.269, ctc_loss=0.1835, cr_loss=0.4274, over 17135.00 frames. ], tot_loss[loss=0.2689, ctc_loss=0.1903, cr_loss=0.3929, over 3331918.23 frames. ], batch size: 48, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:01:31,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=132169.33333333334, ans=0.0
2024-09-23 00:02:02,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.26 vs. limit=15.0
2024-09-23 00:02:09,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=132262.66666666666, ans=0.125
2024-09-23 00:02:26,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=132309.33333333334, ans=0.125
2024-09-23 00:02:28,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=132309.33333333334, ans=0.125
2024-09-23 00:02:35,361 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=7.31 vs. limit=12.0
2024-09-23 00:02:49,159 INFO [train.py:1198] (1/4) Epoch 8, batch 1100, loss[loss=0.2827, ctc_loss=0.2044, cr_loss=0.3913, over 16960.00 frames. ], tot_loss[loss=0.2679, ctc_loss=0.1896, cr_loss=0.3915, over 3342571.40 frames. ], batch size: 58, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:03:31,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.72 vs. limit=15.0
2024-09-23 00:03:45,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=132542.66666666666, ans=0.125
2024-09-23 00:03:46,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=132542.66666666666, ans=0.0
2024-09-23 00:04:08,242 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.381e+02 1.495e+02 1.697e+02 2.213e+02, threshold=2.991e+02, percent-clipped=0.0
2024-09-23 00:04:11,435 INFO [train.py:1198] (1/4) Epoch 8, batch 1150, loss[loss=0.2483, ctc_loss=0.1704, cr_loss=0.3898, over 17086.00 frames. ], tot_loss[loss=0.2678, ctc_loss=0.1894, cr_loss=0.392, over 3354205.17 frames. ], batch size: 43, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:04:26,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=132682.66666666666, ans=0.125
2024-09-23 00:04:27,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=132682.66666666666, ans=0.125
2024-09-23 00:04:40,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=132682.66666666666, ans=0.5
2024-09-23 00:04:45,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=132729.33333333334, ans=0.0
2024-09-23 00:05:21,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.17 vs. limit=22.5
2024-09-23 00:05:33,698 INFO [train.py:1198] (1/4) Epoch 8, batch 1200, loss[loss=0.2419, ctc_loss=0.1724, cr_loss=0.3479, over 17030.00 frames. ], tot_loss[loss=0.2677, ctc_loss=0.1894, cr_loss=0.3918, over 3365123.93 frames. ], batch size: 44, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:05:34,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=132869.33333333334, ans=0.2
2024-09-23 00:05:42,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=132869.33333333334, ans=0.125
2024-09-23 00:05:46,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=132869.33333333334, ans=0.0
2024-09-23 00:06:07,648 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.77 vs. limit=22.5
2024-09-23 00:06:19,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=132962.66666666666, ans=0.5
2024-09-23 00:06:20,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=132962.66666666666, ans=0.125
2024-09-23 00:06:43,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=133056.0, ans=0.125
2024-09-23 00:06:57,396 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.352e+02 1.465e+02 1.619e+02 2.564e+02, threshold=2.930e+02, percent-clipped=0.0
2024-09-23 00:06:58,990 INFO [train.py:1198] (1/4) Epoch 8, batch 1250, loss[loss=0.288, ctc_loss=0.1977, cr_loss=0.4517, over 17026.00 frames. ], tot_loss[loss=0.2676, ctc_loss=0.1893, cr_loss=0.3919, over 3368355.81 frames. ], batch size: 44, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:07:04,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=133102.66666666666, ans=0.125
2024-09-23 00:07:27,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=133149.33333333334, ans=0.125
2024-09-23 00:07:48,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=133242.66666666666, ans=0.125
2024-09-23 00:08:06,521 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.66 vs. limit=15.0
2024-09-23 00:08:10,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=133289.33333333334, ans=0.125
2024-09-23 00:08:18,295 INFO [train.py:1198] (1/4) Epoch 8, batch 1300, loss[loss=0.2878, ctc_loss=0.2071, cr_loss=0.4035, over 16438.00 frames. ], tot_loss[loss=0.2682, ctc_loss=0.1897, cr_loss=0.3923, over 3361482.83 frames. ], batch size: 66, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:08:29,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=133336.0, ans=0.07
2024-09-23 00:08:34,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=133382.66666666666, ans=0.125
2024-09-23 00:08:41,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=133382.66666666666, ans=0.09899494936611666
2024-09-23 00:09:03,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=133429.33333333334, ans=0.2
2024-09-23 00:09:20,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=133476.0, ans=0.125
2024-09-23 00:09:38,133 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.339e+02 1.481e+02 1.642e+02 2.207e+02, threshold=2.961e+02, percent-clipped=0.0
2024-09-23 00:09:38,945 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.92 vs. limit=15.0
2024-09-23 00:09:39,797 INFO [train.py:1198] (1/4) Epoch 8, batch 1350, loss[loss=0.2353, ctc_loss=0.1636, cr_loss=0.3585, over 16294.00 frames. ], tot_loss[loss=0.2672, ctc_loss=0.1888, cr_loss=0.3921, over 3367806.50 frames. ], batch size: 36, lr: 1.52e-02, grad_scale: 32.0
2024-09-23 00:09:40,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=133569.33333333334, ans=0.125
2024-09-23 00:10:08,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=133616.0, ans=0.125
2024-09-23 00:10:18,064 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.81 vs. limit=6.0
2024-09-23 00:10:22,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=133662.66666666666, ans=0.2
2024-09-23 00:10:49,263 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=7.29 vs. limit=15.0
2024-09-23 00:10:50,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=133756.0, ans=0.1
2024-09-23 00:11:06,821 INFO [train.py:1198] (1/4) Epoch 8, batch 1400, loss[loss=0.2486, ctc_loss=0.1767, cr_loss=0.3595, over 17059.00 frames. ], tot_loss[loss=0.2669, ctc_loss=0.1885, cr_loss=0.392, over 3370877.08 frames. ], batch size: 39, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:12:12,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=133989.33333333334, ans=0.1
2024-09-23 00:12:15,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=133989.33333333334, ans=0.025
2024-09-23 00:12:24,402 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.304e+02 1.414e+02 1.596e+02 2.535e+02, threshold=2.828e+02, percent-clipped=0.0
2024-09-23 00:12:25,993 INFO [train.py:1198] (1/4) Epoch 8, batch 1450, loss[loss=0.2409, ctc_loss=0.169, cr_loss=0.3597, over 17269.00 frames. ], tot_loss[loss=0.267, ctc_loss=0.1885, cr_loss=0.3924, over 3371338.03 frames. ], batch size: 42, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:13:13,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=134176.0, ans=0.0
2024-09-23 00:13:29,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=134222.66666666666, ans=0.1
2024-09-23 00:13:48,011 INFO [train.py:1198] (1/4) Epoch 8, batch 1500, loss[loss=0.2958, ctc_loss=0.21, cr_loss=0.4291, over 17001.00 frames. ], tot_loss[loss=0.2661, ctc_loss=0.1878, cr_loss=0.3914, over 3368551.15 frames. ], batch size: 53, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:13:59,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=134269.33333333334, ans=0.1
2024-09-23 00:14:26,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_abs, batch_count=134362.66666666666, ans=0.5
2024-09-23 00:14:47,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=134409.33333333334, ans=0.0
2024-09-23 00:14:56,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=134456.0, ans=0.1
2024-09-23 00:15:01,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=134456.0, ans=0.07
2024-09-23 00:15:09,201 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.316e+02 1.415e+02 1.558e+02 2.036e+02, threshold=2.830e+02, percent-clipped=0.0
2024-09-23 00:15:10,821 INFO [train.py:1198] (1/4) Epoch 8, batch 1550, loss[loss=0.2175, ctc_loss=0.1519, cr_loss=0.3277, over 17262.00 frames. ], tot_loss[loss=0.2667, ctc_loss=0.1883, cr_loss=0.3918, over 3370781.02 frames. ], batch size: 42, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:15:19,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=134502.66666666666, ans=0.1
2024-09-23 00:15:58,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=134596.0, ans=0.1
2024-09-23 00:16:35,828 INFO [train.py:1198] (1/4) Epoch 8, batch 1600, loss[loss=0.2483, ctc_loss=0.1743, cr_loss=0.3703, over 17256.00 frames. ], tot_loss[loss=0.2672, ctc_loss=0.1888, cr_loss=0.3922, over 3369474.38 frames. ], batch size: 44, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:16:36,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=134736.0, ans=0.0
2024-09-23 00:16:45,983 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.92 vs. limit=22.5
2024-09-23 00:16:56,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=134782.66666666666, ans=0.05
2024-09-23 00:17:07,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=134829.33333333334, ans=0.025
2024-09-23 00:17:17,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=134829.33333333334, ans=0.125
2024-09-23 00:17:19,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=134829.33333333334, ans=0.125
2024-09-23 00:17:23,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=134876.0, ans=0.125
2024-09-23 00:17:28,903 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 00:17:49,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=134922.66666666666, ans=0.2
2024-09-23 00:17:54,116 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.357e+02 1.515e+02 1.820e+02 2.997e+02, threshold=3.030e+02, percent-clipped=2.0
2024-09-23 00:17:55,797 INFO [train.py:1198] (1/4) Epoch 8, batch 1650, loss[loss=0.2467, ctc_loss=0.1747, cr_loss=0.3602, over 17290.00 frames. ], tot_loss[loss=0.2649, ctc_loss=0.187, cr_loss=0.3893, over 3369415.95 frames. ], batch size: 51, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:17:57,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=134969.33333333334, ans=0.0
2024-09-23 00:18:10,670 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.29 vs. limit=22.5
2024-09-23 00:18:16,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=135016.0, ans=0.125
2024-09-23 00:18:26,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=135062.66666666666, ans=0.0
2024-09-23 00:18:36,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=135062.66666666666, ans=0.125
2024-09-23 00:18:38,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=135062.66666666666, ans=0.0
2024-09-23 00:18:45,828 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 00:18:49,554 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.67 vs. limit=15.0
2024-09-23 00:18:57,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=135109.33333333334, ans=0.2
2024-09-23 00:19:13,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=135156.0, ans=0.95
2024-09-23 00:19:17,537 INFO [train.py:1198] (1/4) Epoch 8, batch 1700, loss[loss=0.2584, ctc_loss=0.1825, cr_loss=0.3792, over 16723.00 frames. ], tot_loss[loss=0.2656, ctc_loss=0.1875, cr_loss=0.3905, over 3371689.61 frames. ], batch size: 37, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:19:53,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=135296.0, ans=0.1
2024-09-23 00:20:34,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.45 vs. limit=15.0
2024-09-23 00:20:35,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=135389.33333333334, ans=10.0
2024-09-23 00:20:40,291 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.382e+02 1.518e+02 1.704e+02 3.525e+02, threshold=3.037e+02, percent-clipped=1.0
2024-09-23 00:20:41,977 INFO [train.py:1198] (1/4) Epoch 8, batch 1750, loss[loss=0.3045, ctc_loss=0.219, cr_loss=0.4277, over 16670.00 frames. ], tot_loss[loss=0.2665, ctc_loss=0.1881, cr_loss=0.3922, over 3374455.64 frames. ], batch size: 61, lr: 1.51e-02, grad_scale: 32.0
2024-09-23 00:20:45,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=135436.0, ans=0.025
2024-09-23 00:20:46,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=135436.0, ans=0.0
2024-09-23 00:20:50,416 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.16 vs. limit=15.0
2024-09-23 00:20:53,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=135436.0, ans=0.125
2024-09-23 00:21:03,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=135482.66666666666, ans=0.0
2024-09-23 00:21:13,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=135482.66666666666, ans=0.025
2024-09-23 00:21:19,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=135529.33333333334, ans=0.125
2024-09-23 00:21:25,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=135529.33333333334, ans=0.2
2024-09-23 00:21:32,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=135576.0, ans=0.125
2024-09-23 00:21:48,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.39 vs. limit=22.5
2024-09-23 00:22:03,864 INFO [train.py:1198] (1/4) Epoch 8, batch 1800, loss[loss=0.2077, ctc_loss=0.1443, cr_loss=0.317, over 17109.00 frames. ], tot_loss[loss=0.2661, ctc_loss=0.1877, cr_loss=0.3919, over 3369467.39 frames. ], batch size: 40, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:22:11,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=135669.33333333334, ans=0.125
2024-09-23 00:22:16,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=135669.33333333334, ans=0.0
2024-09-23 00:22:31,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=135716.0, ans=0.125
2024-09-23 00:22:45,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=135762.66666666666, ans=0.125
2024-09-23 00:22:53,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=135809.33333333334, ans=0.2
2024-09-23 00:22:54,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=135809.33333333334, ans=0.1
2024-09-23 00:23:12,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten.whitening_limit, batch_count=135856.0, ans=15.0
2024-09-23 00:23:21,450 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.352e+02 1.514e+02 1.696e+02 3.020e+02, threshold=3.028e+02, percent-clipped=0.0
2024-09-23 00:23:23,111 INFO [train.py:1198] (1/4) Epoch 8, batch 1850, loss[loss=0.2737, ctc_loss=0.191, cr_loss=0.4135, over 17038.00 frames. ], tot_loss[loss=0.2654, ctc_loss=0.1872, cr_loss=0.3908, over 3362237.25 frames. ], batch size: 52, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:24:26,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=136042.66666666666, ans=0.125
2024-09-23 00:24:29,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=136089.33333333334, ans=0.1
2024-09-23 00:24:29,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.92 vs. limit=22.5
2024-09-23 00:24:32,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=136089.33333333334, ans=0.0
2024-09-23 00:24:47,780 INFO [train.py:1198] (1/4) Epoch 8, batch 1900, loss[loss=0.2294, ctc_loss=0.1612, cr_loss=0.3414, over 17006.00 frames. ], tot_loss[loss=0.264, ctc_loss=0.1862, cr_loss=0.3887, over 3360880.18 frames. ], batch size: 44, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:24:55,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=136136.0, ans=0.07
2024-09-23 00:24:58,094 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=19.78 vs. limit=22.5
2024-09-23 00:25:26,308 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.00 vs. limit=15.0
2024-09-23 00:25:27,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=136229.33333333334, ans=0.1
2024-09-23 00:26:01,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=136322.66666666666, ans=0.07
2024-09-23 00:26:02,219 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=8.15 vs. limit=15.0
2024-09-23 00:26:08,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=136322.66666666666, ans=0.125
2024-09-23 00:26:10,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2.whitening_limit, batch_count=136322.66666666666, ans=15.0
2024-09-23 00:26:11,059 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.376e+02 1.564e+02 1.775e+02 2.559e+02, threshold=3.128e+02, percent-clipped=0.0
2024-09-23 00:26:12,647 INFO [train.py:1198] (1/4) Epoch 8, batch 1950, loss[loss=0.2875, ctc_loss=0.1998, cr_loss=0.4387, over 17234.00 frames. ], tot_loss[loss=0.265, ctc_loss=0.187, cr_loss=0.3899, over 3360274.38 frames. ], batch size: 55, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:26:28,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=136416.0, ans=0.2
2024-09-23 00:26:55,304 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=14.62 vs. limit=15.0
2024-09-23 00:27:00,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=136509.33333333334, ans=0.1
2024-09-23 00:27:32,278 INFO [train.py:1198] (1/4) Epoch 8, batch 2000, loss[loss=0.2614, ctc_loss=0.1853, cr_loss=0.3805, over 17291.00 frames. ], tot_loss[loss=0.2648, ctc_loss=0.1869, cr_loss=0.3895, over 3356354.61 frames. ], batch size: 49, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:28:06,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=136696.0, ans=0.125
2024-09-23 00:28:28,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.99 vs. limit=12.0
2024-09-23 00:28:31,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=136742.66666666666, ans=0.125
2024-09-23 00:28:37,116 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.35 vs. limit=15.0
2024-09-23 00:28:53,819 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.381e+02 1.478e+02 1.758e+02 2.844e+02, threshold=2.957e+02, percent-clipped=0.0
2024-09-23 00:28:55,407 INFO [train.py:1198] (1/4) Epoch 8, batch 2050, loss[loss=0.2887, ctc_loss=0.2043, cr_loss=0.4218, over 17007.00 frames. ], tot_loss[loss=0.2651, ctc_loss=0.187, cr_loss=0.3903, over 3360757.95 frames. ], batch size: 56, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:29:30,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=136929.33333333334, ans=0.1
2024-09-23 00:30:12,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=137022.66666666666, ans=0.025
2024-09-23 00:30:16,953 INFO [train.py:1198] (1/4) Epoch 8, batch 2100, loss[loss=0.2386, ctc_loss=0.1677, cr_loss=0.3544, over 17167.00 frames. ], tot_loss[loss=0.2648, ctc_loss=0.1868, cr_loss=0.39, over 3357066.60 frames. ], batch size: 41, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:30:31,468 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.79 vs. limit=15.0
2024-09-23 00:30:37,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=137116.0, ans=0.125
2024-09-23 00:30:50,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=137116.0, ans=0.1
2024-09-23 00:30:52,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=137162.66666666666, ans=0.2
2024-09-23 00:31:25,251 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.13 vs. limit=22.5
2024-09-23 00:31:25,514 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.99 vs. limit=15.0
2024-09-23 00:31:39,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=137256.0, ans=0.0
2024-09-23 00:31:40,396 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.375e+02 1.520e+02 1.695e+02 2.810e+02, threshold=3.041e+02, percent-clipped=0.0
2024-09-23 00:31:41,917 INFO [train.py:1198] (1/4) Epoch 8, batch 2150, loss[loss=0.2642, ctc_loss=0.1851, cr_loss=0.3955, over 17252.00 frames. ], tot_loss[loss=0.2645, ctc_loss=0.1865, cr_loss=0.3903, over 3350533.67 frames. ], batch size: 44, lr: 1.50e-02, grad_scale: 32.0
2024-09-23 00:32:27,796 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.71 vs. limit=10.0
2024-09-23 00:32:38,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=137442.66666666666, ans=0.0
2024-09-23 00:32:44,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=137489.33333333334, ans=0.0
2024-09-23 00:33:01,805 INFO [train.py:1198] (1/4) Epoch 8, batch 2200, loss[loss=0.2905, ctc_loss=0.2096, cr_loss=0.4045, over 16887.00 frames. ], tot_loss[loss=0.265, ctc_loss=0.1869, cr_loss=0.3905, over 3351461.10 frames. ], batch size: 58, lr: 1.49e-02, grad_scale: 32.0
2024-09-23 00:33:40,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=137629.33333333334, ans=0.125
2024-09-23 00:33:50,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=137676.0, ans=0.125
2024-09-23 00:34:13,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=137722.66666666666, ans=0.0
2024-09-23 00:34:16,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=137722.66666666666, ans=0.025
2024-09-23 00:34:21,855 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.429e+02 1.621e+02 1.891e+02 2.462e+02, threshold=3.242e+02, percent-clipped=0.0
2024-09-23 00:34:23,498 INFO [train.py:1198] (1/4) Epoch 8, batch 2250, loss[loss=0.3081, ctc_loss=0.2159, cr_loss=0.4611, over 16915.00 frames. ], tot_loss[loss=0.2642, ctc_loss=0.1862, cr_loss=0.39, over 3354432.90 frames. ], batch size: 58, lr: 1.49e-02, grad_scale: 32.0
2024-09-23 00:34:25,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=137769.33333333334, ans=0.2
2024-09-23 00:34:34,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=137769.33333333334, ans=0.125
2024-09-23 00:34:52,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=137816.0, ans=0.0
2024-09-23 00:34:55,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=137816.0, ans=0.0
2024-09-23 00:35:02,180 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.62 vs. limit=15.0
2024-09-23 00:35:13,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=137862.66666666666, ans=10.0
2024-09-23 00:35:29,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=137909.33333333334, ans=0.0
2024-09-23 00:35:34,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=137956.0, ans=0.2
2024-09-23 00:35:50,812 INFO [train.py:1198] (1/4) Epoch 8, batch 2300, loss[loss=0.269, ctc_loss=0.1917, cr_loss=0.3866, over 17238.00 frames. ], tot_loss[loss=0.2639, ctc_loss=0.1859, cr_loss=0.39, over 3365226.09 frames. ], batch size: 44, lr: 1.49e-02, grad_scale: 32.0
2024-09-23 00:35:57,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=138002.66666666666, ans=0.125
2024-09-23 00:36:18,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=138049.33333333334, ans=0.0
2024-09-23 00:36:26,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=138096.0, ans=0.125
2024-09-23 00:37:04,078 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.50 vs. limit=6.0
2024-09-23 00:37:09,110 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.383e+02 1.498e+02 1.682e+02 2.493e+02, threshold=2.995e+02, percent-clipped=0.0
2024-09-23 00:37:10,794 INFO [train.py:1198] (1/4) Epoch 8, batch 2350, loss[loss=0.2734, ctc_loss=0.1914, cr_loss=0.4096, over 17027.00 frames. ], tot_loss[loss=0.2642, ctc_loss=0.186, cr_loss=0.3909, over 3370523.14 frames. ], batch size: 44, lr: 1.49e-02, grad_scale: 32.0
2024-09-23 00:37:28,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=138282.66666666666, ans=0.125
2024-09-23 00:37:28,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=138282.66666666666, ans=0.0
2024-09-23 00:37:31,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=138282.66666666666, ans=0.05
2024-09-23 00:38:16,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=138422.66666666666, ans=0.125
2024-09-23 00:38:17,158 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.57 vs. limit=15.0
2024-09-23 00:38:33,070 INFO [train.py:1198] (1/4) Epoch 8, batch 2400, loss[loss=0.2743, ctc_loss=0.1939, cr_loss=0.4019, over 17351.00 frames. ], tot_loss[loss=0.2646, ctc_loss=0.1864, cr_loss=0.3913, over 3370380.93 frames. ], batch size: 48, lr: 1.49e-02, grad_scale: 32.0
2024-09-23 00:38:47,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=138516.0, ans=0.0
2024-09-23 00:38:55,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=138516.0, ans=0.125
2024-09-23 00:38:56,068 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.37 vs. limit=15.0
2024-09-23 00:39:21,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=138609.33333333334, ans=0.0
2024-09-23 00:39:35,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=138609.33333333334, ans=0.025
2024-09-23 00:39:45,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=138656.0, ans=0.125
2024-09-23 00:39:49,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=138656.0, ans=0.0
2024-09-23 00:39:53,485 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.369e+02 1.492e+02 1.774e+02 2.629e+02, threshold=2.985e+02, percent-clipped=0.0
2024-09-23 00:39:55,153 INFO [train.py:1198] (1/4) Epoch 8, batch 2450, loss[loss=0.2483, ctc_loss=0.1744, cr_loss=0.3693, over 17344.00 frames. ], tot_loss[loss=0.2658, ctc_loss=0.1872, cr_loss=0.3931, over 3373129.18 frames. ], batch size: 48, lr: 1.49e-02, grad_scale: 32.0
2024-09-23 00:40:26,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=138749.33333333334, ans=0.0
2024-09-23 00:40:32,215 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.73 vs. limit=10.0
2024-09-23 00:41:01,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=138842.66666666666, ans=0.0
2024-09-23 00:41:20,429 INFO [train.py:1198] (1/4) Epoch 8, batch 2500, loss[loss=0.2706, ctc_loss=0.1904, cr_loss=0.4007, over 17316.00 frames. ], tot_loss[loss=0.266, ctc_loss=0.1874, cr_loss=0.3931, over 3362379.36 frames. ], batch size: 46, lr: 1.49e-02, grad_scale: 32.0
2024-09-23 00:41:38,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=138982.66666666666, ans=0.0
2024-09-23 00:41:49,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=138982.66666666666, ans=0.05
2024-09-23 00:42:10,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=139076.0, ans=0.1
2024-09-23 00:42:36,076 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.57 vs. limit=15.0
2024-09-23 00:42:40,013 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.309e+02 1.454e+02 1.673e+02 2.570e+02, threshold=2.909e+02, percent-clipped=0.0
2024-09-23 00:42:40,037 INFO [train.py:1198] (1/4) Epoch 8, batch 2550, loss[loss=0.2667, ctc_loss=0.1835, cr_loss=0.4157, over 17289.00 frames. ], tot_loss[loss=0.2653, ctc_loss=0.1868, cr_loss=0.3921, over 3365185.10 frames. ], batch size: 49, lr: 1.49e-02, grad_scale: 16.0
2024-09-23 00:42:46,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=139169.33333333334, ans=0.125
2024-09-23 00:43:19,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=139262.66666666666, ans=0.1
2024-09-23 00:43:27,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer_na.min_abs, batch_count=139262.66666666666, ans=0.02
2024-09-23 00:43:34,636 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.15 vs. limit=15.0
2024-09-23 00:43:39,235 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=16.23 vs. limit=22.5
2024-09-23 00:43:48,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=139356.0, ans=0.125
2024-09-23 00:43:57,231 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.29 vs. limit=10.0
2024-09-23 00:44:02,587 INFO [train.py:1198] (1/4) Epoch 8, batch 2600, loss[loss=0.2406, ctc_loss=0.1663, cr_loss=0.3715, over 17169.00 frames. ], tot_loss[loss=0.2647, ctc_loss=0.1865, cr_loss=0.3911, over 3367165.22 frames. ], batch size: 41, lr: 1.48e-02, grad_scale: 16.0
2024-09-23 00:44:02,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=139402.66666666666, ans=0.125
2024-09-23 00:44:43,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=139496.0, ans=0.0
2024-09-23 00:44:46,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=139496.0, ans=0.2
2024-09-23 00:45:27,487 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.349e+02 1.476e+02 1.822e+02 2.637e+02, threshold=2.952e+02, percent-clipped=0.0
2024-09-23 00:45:27,512 INFO [train.py:1198] (1/4) Epoch 8, batch 2650, loss[loss=0.2905, ctc_loss=0.2109, cr_loss=0.3982, over 17209.00 frames. ], tot_loss[loss=0.2639, ctc_loss=0.1859, cr_loss=0.3897, over 3373595.04 frames. ], batch size: 55, lr: 1.48e-02, grad_scale: 16.0
2024-09-23 00:45:27,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=139636.0, ans=0.125
2024-09-23 00:45:36,092 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.70 vs. limit=15.0
2024-09-23 00:45:58,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=139682.66666666666, ans=0.2
2024-09-23 00:45:58,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=139682.66666666666, ans=0.125
2024-09-23 00:46:05,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=139729.33333333334, ans=0.2
2024-09-23 00:46:49,528 INFO [train.py:1198] (1/4) Epoch 8, batch 2700, loss[loss=0.244, ctc_loss=0.1701, cr_loss=0.3698, over 17180.00 frames. ], tot_loss[loss=0.2649, ctc_loss=0.1866, cr_loss=0.3913, over 3367200.96 frames. ], batch size: 45, lr: 1.48e-02, grad_scale: 16.0
2024-09-23 00:46:59,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=139869.33333333334, ans=0.0
2024-09-23 00:47:25,379 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.16 vs. limit=22.5
2024-09-23 00:47:32,902 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 00:47:42,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=140009.33333333334, ans=0.0
2024-09-23 00:48:01,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=140056.0, ans=0.1
2024-09-23 00:48:11,966 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.374e+02 1.576e+02 1.752e+02 3.056e+02, threshold=3.153e+02, percent-clipped=1.0
2024-09-23 00:48:11,991 INFO [train.py:1198] (1/4) Epoch 8, batch 2750, loss[loss=0.2405, ctc_loss=0.1645, cr_loss=0.38, over 17090.00 frames. ], tot_loss[loss=0.2657, ctc_loss=0.1872, cr_loss=0.3926, over 3367299.72 frames. ], batch size: 43, lr: 1.48e-02, grad_scale: 16.0
2024-09-23 00:48:20,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=140102.66666666666, ans=0.0
2024-09-23 00:48:45,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=140196.0, ans=0.0
2024-09-23 00:49:00,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=140242.66666666666, ans=0.05
2024-09-23 00:49:03,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=140242.66666666666, ans=0.0
2024-09-23 00:49:22,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=140289.33333333334, ans=0.0
2024-09-23 00:49:34,153 INFO [train.py:1198] (1/4) Epoch 8, batch 2800, loss[loss=0.3014, ctc_loss=0.2108, cr_loss=0.453, over 17305.00 frames. ], tot_loss[loss=0.2651, ctc_loss=0.1867, cr_loss=0.3918, over 3360392.51 frames. ], batch size: 49, lr: 1.48e-02, grad_scale: 32.0
2024-09-23 00:49:42,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=140336.0, ans=0.025
2024-09-23 00:49:44,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=140336.0, ans=0.0
2024-09-23 00:49:50,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=140382.66666666666, ans=0.2
2024-09-23 00:50:55,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=140522.66666666666, ans=0.0
2024-09-23 00:50:58,772 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.406e+02 1.667e+02 1.997e+02 3.785e+02, threshold=3.334e+02, percent-clipped=1.0
2024-09-23 00:50:58,797 INFO [train.py:1198] (1/4) Epoch 8, batch 2850, loss[loss=0.2355, ctc_loss=0.1616, cr_loss=0.3694, over 16320.00 frames. ], tot_loss[loss=0.263, ctc_loss=0.1852, cr_loss=0.389, over 3356767.96 frames. ], batch size: 36, lr: 1.48e-02, grad_scale: 32.0
2024-09-23 00:51:35,224 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=6.87 vs. limit=8.0
2024-09-23 00:51:38,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=140662.66666666666, ans=0.1
2024-09-23 00:52:11,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=140756.0, ans=0.1
2024-09-23 00:52:18,842 INFO [train.py:1198] (1/4) Epoch 8, batch 2900, loss[loss=0.294, ctc_loss=0.2057, cr_loss=0.4415, over 17218.00 frames. ], tot_loss[loss=0.2638, ctc_loss=0.1858, cr_loss=0.3904, over 3355122.33 frames. ], batch size: 55, lr: 1.48e-02, grad_scale: 32.0
2024-09-23 00:52:20,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=140802.66666666666, ans=0.2
2024-09-23 00:52:46,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=140849.33333333334, ans=0.125
2024-09-23 00:52:57,630 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=140896.0, ans=0.125
2024-09-23 00:52:59,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=140896.0, ans=0.0
2024-09-23 00:52:59,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=140896.0, ans=0.125
2024-09-23 00:53:02,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=140896.0, ans=0.0
2024-09-23 00:53:13,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=140942.66666666666, ans=0.025
2024-09-23 00:53:39,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=141036.0, ans=0.0
2024-09-23 00:53:41,170 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.334e+02 1.488e+02 1.595e+02 2.196e+02, threshold=2.977e+02, percent-clipped=0.0
2024-09-23 00:53:41,194 INFO [train.py:1198] (1/4) Epoch 8, batch 2950, loss[loss=0.2732, ctc_loss=0.1934, cr_loss=0.3989, over 17011.00 frames. ], tot_loss[loss=0.2641, ctc_loss=0.1859, cr_loss=0.391, over 3352961.54 frames. ], batch size: 52, lr: 1.48e-02, grad_scale: 32.0
2024-09-23 00:53:49,844 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.22 vs. limit=15.0
2024-09-23 00:54:43,387 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.91 vs. limit=15.0
2024-09-23 00:55:02,757 INFO [train.py:1198] (1/4) Epoch 8, batch 3000, loss[loss=0.2987, ctc_loss=0.2187, cr_loss=0.3999, over 15215.00 frames. ], tot_loss[loss=0.2638, ctc_loss=0.1856, cr_loss=0.3909, over 3367303.63 frames. ], batch size: 89, lr: 1.48e-02, grad_scale: 32.0
2024-09-23 00:55:02,757 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 00:55:14,959 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([3.3820, 3.0844, 2.5590, 2.9530], device='cuda:1')
2024-09-23 00:55:18,811 INFO [train.py:1230] (1/4) Epoch 8, validation: loss=0.05304, ctc_loss=0.05304, cr_loss=7.247e-15, over 944034.00 frames. 
2024-09-23 00:55:18,812 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 00:55:22,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=141269.33333333334, ans=0.0
2024-09-23 00:55:23,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=141269.33333333334, ans=0.0
2024-09-23 00:55:53,922 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 00:56:09,327 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.41 vs. limit=22.5
2024-09-23 00:56:18,941 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.27 vs. limit=15.0
2024-09-23 00:56:26,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.12 vs. limit=12.0
2024-09-23 00:56:27,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=141456.0, ans=0.1
2024-09-23 00:56:39,932 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.352e+02 1.499e+02 1.786e+02 3.736e+02, threshold=2.997e+02, percent-clipped=4.0
2024-09-23 00:56:39,956 INFO [train.py:1198] (1/4) Epoch 8, batch 3050, loss[loss=0.2674, ctc_loss=0.1898, cr_loss=0.3881, over 17096.00 frames. ], tot_loss[loss=0.265, ctc_loss=0.1867, cr_loss=0.3918, over 3361741.89 frames. ], batch size: 49, lr: 1.47e-02, grad_scale: 32.0
2024-09-23 00:56:48,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=141502.66666666666, ans=0.0
2024-09-23 00:56:49,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=141502.66666666666, ans=0.035
2024-09-23 00:56:52,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=141502.66666666666, ans=0.125
2024-09-23 00:57:02,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=141549.33333333334, ans=0.0
2024-09-23 00:57:18,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=141596.0, ans=0.035
2024-09-23 00:57:31,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.25 vs. limit=15.0
2024-09-23 00:57:58,665 INFO [train.py:1198] (1/4) Epoch 8, batch 3100, loss[loss=0.2502, ctc_loss=0.1784, cr_loss=0.359, over 17082.00 frames. ], tot_loss[loss=0.2636, ctc_loss=0.1856, cr_loss=0.39, over 3364379.51 frames. ], batch size: 43, lr: 1.47e-02, grad_scale: 16.0
2024-09-23 00:58:03,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=141736.0, ans=0.0
2024-09-23 00:58:05,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=141736.0, ans=0.125
2024-09-23 00:58:08,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=141736.0, ans=0.0
2024-09-23 00:58:09,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=141736.0, ans=0.1
2024-09-23 00:58:09,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=141736.0, ans=0.2
2024-09-23 00:58:39,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=141829.33333333334, ans=0.125
2024-09-23 00:58:44,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=141876.0, ans=0.0
2024-09-23 00:58:45,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=141876.0, ans=0.125
2024-09-23 00:58:50,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=141876.0, ans=0.2
2024-09-23 00:58:52,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.54 vs. limit=15.0
2024-09-23 00:59:03,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.87 vs. limit=22.5
2024-09-23 00:59:16,309 INFO [train.py:1198] (1/4) Epoch 8, batch 3150, loss[loss=0.2424, ctc_loss=0.1709, cr_loss=0.3577, over 17161.00 frames. ], tot_loss[loss=0.2632, ctc_loss=0.1854, cr_loss=0.3893, over 3363691.53 frames. ], batch size: 45, lr: 1.47e-02, grad_scale: 16.0
2024-09-23 00:59:17,877 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.325e+02 1.474e+02 1.672e+02 3.223e+02, threshold=2.948e+02, percent-clipped=1.0
2024-09-23 00:59:21,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=141969.33333333334, ans=0.125
2024-09-23 00:59:34,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.93 vs. limit=15.0
2024-09-23 00:59:44,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=142016.0, ans=0.1
2024-09-23 00:59:46,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=142062.66666666666, ans=0.0
2024-09-23 01:00:02,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=142109.33333333334, ans=0.0
2024-09-23 01:00:29,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=142156.0, ans=0.0
2024-09-23 01:00:34,386 INFO [train.py:1198] (1/4) Epoch 8, batch 3200, loss[loss=0.2354, ctc_loss=0.1628, cr_loss=0.3631, over 17090.00 frames. ], tot_loss[loss=0.265, ctc_loss=0.1868, cr_loss=0.3911, over 3364469.18 frames. ], batch size: 40, lr: 1.47e-02, grad_scale: 32.0
2024-09-23 01:00:36,490 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.17 vs. limit=15.0
2024-09-23 01:01:07,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=142296.0, ans=0.0
2024-09-23 01:01:31,258 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=16.35 vs. limit=15.0
2024-09-23 01:01:45,263 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.80 vs. limit=22.5
2024-09-23 01:01:46,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=142389.33333333334, ans=0.125
2024-09-23 01:01:54,764 INFO [train.py:1198] (1/4) Epoch 8, batch 3250, loss[loss=0.2466, ctc_loss=0.173, cr_loss=0.3678, over 17285.00 frames. ], tot_loss[loss=0.2652, ctc_loss=0.1868, cr_loss=0.3917, over 3358470.81 frames. ], batch size: 46, lr: 1.47e-02, grad_scale: 32.0
2024-09-23 01:01:56,369 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.501e+02 1.666e+02 1.949e+02 2.835e+02, threshold=3.332e+02, percent-clipped=0.0
2024-09-23 01:02:12,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=142482.66666666666, ans=0.125
2024-09-23 01:03:12,271 INFO [train.py:1198] (1/4) Epoch 8, batch 3300, loss[loss=0.323, ctc_loss=0.2458, cr_loss=0.386, over 11915.00 frames. ], tot_loss[loss=0.2667, ctc_loss=0.1882, cr_loss=0.3924, over 3335671.83 frames. ], batch size: 123, lr: 1.47e-02, grad_scale: 32.0
2024-09-23 01:03:22,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.00 vs. limit=15.0
2024-09-23 01:03:24,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=142669.33333333334, ans=0.0
2024-09-23 01:04:04,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=142809.33333333334, ans=0.035
2024-09-23 01:04:18,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.99 vs. limit=15.0
2024-09-23 01:04:30,462 INFO [train.py:1198] (1/4) Epoch 8, batch 3350, loss[loss=0.2598, ctc_loss=0.1834, cr_loss=0.382, over 17061.00 frames. ], tot_loss[loss=0.2654, ctc_loss=0.1872, cr_loss=0.3911, over 3342490.65 frames. ], batch size: 46, lr: 1.47e-02, grad_scale: 16.0
2024-09-23 01:04:33,559 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.197e+02 1.393e+02 1.569e+02 1.774e+02 3.394e+02, threshold=3.137e+02, percent-clipped=1.0
2024-09-23 01:04:51,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=142949.33333333334, ans=0.125
2024-09-23 01:05:05,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=142996.0, ans=0.1
2024-09-23 01:05:10,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=142996.0, ans=0.125
2024-09-23 01:05:21,702 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.56 vs. limit=15.0
2024-09-23 01:05:27,683 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 01:05:29,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=143042.66666666666, ans=0.1
2024-09-23 01:05:33,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=143089.33333333334, ans=0.125
2024-09-23 01:05:37,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=143089.33333333334, ans=0.125
2024-09-23 01:05:38,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=143089.33333333334, ans=0.1
2024-09-23 01:05:52,730 INFO [train.py:1198] (1/4) Epoch 8, batch 3400, loss[loss=0.28, ctc_loss=0.1943, cr_loss=0.4287, over 17222.00 frames. ], tot_loss[loss=0.267, ctc_loss=0.1884, cr_loss=0.3929, over 3335215.67 frames. ], batch size: 55, lr: 1.47e-02, grad_scale: 16.0
2024-09-23 01:06:01,130 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=15.75 vs. limit=15.0
2024-09-23 01:06:27,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=143229.33333333334, ans=0.125
2024-09-23 01:06:35,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=143229.33333333334, ans=0.0
2024-09-23 01:06:53,317 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=11.98 vs. limit=15.0
2024-09-23 01:06:57,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=143322.66666666666, ans=0.1
2024-09-23 01:07:09,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=143322.66666666666, ans=0.125
2024-09-23 01:07:12,535 INFO [train.py:1198] (1/4) Epoch 8, batch 3450, loss[loss=0.2599, ctc_loss=0.1839, cr_loss=0.3802, over 17299.00 frames. ], tot_loss[loss=0.2656, ctc_loss=0.1873, cr_loss=0.3913, over 3345812.27 frames. ], batch size: 46, lr: 1.47e-02, grad_scale: 16.0
2024-09-23 01:07:12,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=143369.33333333334, ans=0.125
2024-09-23 01:07:15,656 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.386e+02 1.521e+02 1.778e+02 2.541e+02, threshold=3.041e+02, percent-clipped=0.0
2024-09-23 01:07:15,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=143369.33333333334, ans=0.025
2024-09-23 01:07:22,446 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 01:07:47,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=143462.66666666666, ans=0.0
2024-09-23 01:07:56,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=143462.66666666666, ans=0.125
2024-09-23 01:08:10,527 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=5.360e-03
2024-09-23 01:08:22,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=143556.0, ans=0.125
2024-09-23 01:08:23,301 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.45 vs. limit=12.0
2024-09-23 01:08:30,382 INFO [train.py:1198] (1/4) Epoch 8, batch 3500, loss[loss=0.2669, ctc_loss=0.1906, cr_loss=0.3817, over 17224.00 frames. ], tot_loss[loss=0.2654, ctc_loss=0.1873, cr_loss=0.3907, over 3338904.08 frames. ], batch size: 47, lr: 1.46e-02, grad_scale: 16.0
2024-09-23 01:08:32,769 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.84 vs. limit=15.0
2024-09-23 01:09:36,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=143789.33333333334, ans=0.09899494936611666
2024-09-23 01:09:48,668 INFO [train.py:1198] (1/4) Epoch 8, batch 3550, loss[loss=0.2928, ctc_loss=0.2069, cr_loss=0.4294, over 17014.00 frames. ], tot_loss[loss=0.2649, ctc_loss=0.1869, cr_loss=0.39, over 3340223.22 frames. ], batch size: 51, lr: 1.46e-02, grad_scale: 16.0
2024-09-23 01:09:51,761 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.323e+02 1.423e+02 1.598e+02 2.580e+02, threshold=2.846e+02, percent-clipped=0.0
2024-09-23 01:10:37,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=143976.0, ans=0.0
2024-09-23 01:10:58,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=144022.66666666666, ans=0.0
2024-09-23 01:10:58,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=144022.66666666666, ans=6.0
2024-09-23 01:11:00,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=144022.66666666666, ans=0.0
2024-09-23 01:11:03,979 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=2.73 vs. limit=15.0
2024-09-23 01:11:06,358 INFO [train.py:1198] (1/4) Epoch 8, batch 3600, loss[loss=0.251, ctc_loss=0.1768, cr_loss=0.3714, over 17178.00 frames. ], tot_loss[loss=0.2632, ctc_loss=0.1855, cr_loss=0.3887, over 3353163.90 frames. ], batch size: 41, lr: 1.46e-02, grad_scale: 32.0
2024-09-23 01:11:21,025 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 01:11:36,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=144116.0, ans=0.0
2024-09-23 01:12:27,262 INFO [train.py:1198] (1/4) Epoch 8, batch 3650, loss[loss=0.2714, ctc_loss=0.1972, cr_loss=0.3707, over 17120.00 frames. ], tot_loss[loss=0.2636, ctc_loss=0.1857, cr_loss=0.3895, over 3359962.25 frames. ], batch size: 48, lr: 1.46e-02, grad_scale: 32.0
2024-09-23 01:12:30,476 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.372e+02 1.519e+02 1.739e+02 2.361e+02, threshold=3.037e+02, percent-clipped=0.0
2024-09-23 01:13:17,492 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.56 vs. limit=15.0
2024-09-23 01:13:34,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=144489.33333333334, ans=0.125
2024-09-23 01:13:45,448 INFO [train.py:1198] (1/4) Epoch 8, batch 3700, loss[loss=0.313, ctc_loss=0.2235, cr_loss=0.4473, over 17360.00 frames. ], tot_loss[loss=0.2632, ctc_loss=0.1853, cr_loss=0.3892, over 3365765.04 frames. ], batch size: 48, lr: 1.46e-02, grad_scale: 32.0
2024-09-23 01:13:53,849 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.91 vs. limit=15.0
2024-09-23 01:13:58,697 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.75 vs. limit=6.0
2024-09-23 01:14:20,035 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.12 vs. limit=10.0
2024-09-23 01:14:36,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=144676.0, ans=0.0
2024-09-23 01:14:45,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=144676.0, ans=0.0
2024-09-23 01:14:53,802 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.21 vs. limit=15.0
2024-09-23 01:15:03,785 INFO [train.py:1198] (1/4) Epoch 8, batch 3750, loss[loss=0.2812, ctc_loss=0.2007, cr_loss=0.4026, over 17025.00 frames. ], tot_loss[loss=0.2641, ctc_loss=0.1859, cr_loss=0.3907, over 3366403.57 frames. ], batch size: 52, lr: 1.46e-02, grad_scale: 32.0
2024-09-23 01:15:06,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=144769.33333333334, ans=0.2
2024-09-23 01:15:07,862 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.371e+02 1.582e+02 1.824e+02 4.757e+02, threshold=3.165e+02, percent-clipped=1.0
2024-09-23 01:15:17,834 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.84 vs. limit=6.0
2024-09-23 01:15:31,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=144816.0, ans=0.0
2024-09-23 01:16:01,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=144909.33333333334, ans=0.0
2024-09-23 01:16:17,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=144956.0, ans=0.025
2024-09-23 01:16:23,490 INFO [train.py:1198] (1/4) Epoch 8, batch 3800, loss[loss=0.2589, ctc_loss=0.1877, cr_loss=0.3562, over 17304.00 frames. ], tot_loss[loss=0.2659, ctc_loss=0.1875, cr_loss=0.3918, over 3344122.54 frames. ], batch size: 46, lr: 1.46e-02, grad_scale: 32.0
2024-09-23 01:16:23,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=145002.66666666666, ans=0.125
2024-09-23 01:16:47,551 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.76 vs. limit=15.0
2024-09-23 01:16:47,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten.whitening_limit, batch_count=145049.33333333334, ans=15.0
2024-09-23 01:17:16,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=145142.66666666666, ans=0.125
2024-09-23 01:17:18,659 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.86 vs. limit=10.0
2024-09-23 01:17:21,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=145142.66666666666, ans=0.035
2024-09-23 01:17:24,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=145189.33333333334, ans=0.125
2024-09-23 01:17:41,069 INFO [train.py:1198] (1/4) Epoch 8, batch 3850, loss[loss=0.2442, ctc_loss=0.1707, cr_loss=0.3677, over 16980.00 frames. ], tot_loss[loss=0.2702, ctc_loss=0.1915, cr_loss=0.3934, over 3269830.10 frames. ], batch size: 42, lr: 1.46e-02, grad_scale: 16.0
2024-09-23 01:17:45,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.439e+02 1.582e+02 1.881e+02 4.076e+02, threshold=3.165e+02, percent-clipped=1.0
2024-09-23 01:18:05,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=145282.66666666666, ans=0.0
2024-09-23 01:19:42,987 INFO [train.py:1198] (1/4) Epoch 9, batch 0, loss[loss=0.2782, ctc_loss=0.1954, cr_loss=0.4139, over 17304.00 frames. ], tot_loss[loss=0.2782, ctc_loss=0.1954, cr_loss=0.4139, over 17304.00 frames. ], batch size: 51, lr: 1.38e-02, grad_scale: 32.0
2024-09-23 01:19:42,987 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 01:19:58,862 INFO [train.py:1230] (1/4) Epoch 9, validation: loss=0.05451, ctc_loss=0.05451, cr_loss=7.076e-15, over 944034.00 frames. 
2024-09-23 01:19:58,863 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 01:20:04,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=145450.66666666666, ans=0.125
2024-09-23 01:20:28,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=145497.33333333334, ans=0.125
2024-09-23 01:20:42,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=145544.0, ans=0.025
2024-09-23 01:20:56,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=145590.66666666666, ans=0.0
2024-09-23 01:21:07,186 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.55 vs. limit=10.0
2024-09-23 01:21:08,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer_ff3.min_abs, batch_count=145637.33333333334, ans=0.2
2024-09-23 01:21:12,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.35 vs. limit=10.0
2024-09-23 01:21:23,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=145637.33333333334, ans=0.0
2024-09-23 01:21:26,165 INFO [train.py:1198] (1/4) Epoch 9, batch 50, loss[loss=0.2605, ctc_loss=0.1818, cr_loss=0.3936, over 17320.00 frames. ], tot_loss[loss=0.2624, ctc_loss=0.1843, cr_loss=0.3908, over 761820.10 frames. ], batch size: 51, lr: 1.38e-02, grad_scale: 32.0
2024-09-23 01:21:37,366 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.512e+02 1.709e+02 2.026e+02 3.260e+02, threshold=3.417e+02, percent-clipped=2.0
2024-09-23 01:21:43,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=145730.66666666666, ans=0.0
2024-09-23 01:21:46,120 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.16 vs. limit=10.0
2024-09-23 01:22:06,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=145777.33333333334, ans=0.0
2024-09-23 01:22:28,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=145870.66666666666, ans=0.05
2024-09-23 01:22:33,551 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.42 vs. limit=6.0
2024-09-23 01:22:47,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=145917.33333333334, ans=0.1
2024-09-23 01:22:48,404 INFO [train.py:1198] (1/4) Epoch 9, batch 100, loss[loss=0.3078, ctc_loss=0.2233, cr_loss=0.4227, over 16991.00 frames. ], tot_loss[loss=0.2597, ctc_loss=0.1822, cr_loss=0.3877, over 1335461.41 frames. ], batch size: 56, lr: 1.38e-02, grad_scale: 32.0
2024-09-23 01:22:54,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=145917.33333333334, ans=0.125
2024-09-23 01:22:56,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=145917.33333333334, ans=0.125
2024-09-23 01:23:03,714 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=11.02 vs. limit=15.0
2024-09-23 01:23:04,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=145964.0, ans=0.125
2024-09-23 01:23:04,827 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.83 vs. limit=22.5
2024-09-23 01:23:15,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=145964.0, ans=0.0
2024-09-23 01:24:08,182 INFO [train.py:1198] (1/4) Epoch 9, batch 150, loss[loss=0.3226, ctc_loss=0.2437, cr_loss=0.3943, over 11899.00 frames. ], tot_loss[loss=0.2595, ctc_loss=0.1823, cr_loss=0.3859, over 1781587.15 frames. ], batch size: 124, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:24:19,530 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.178e+02 1.303e+02 1.427e+02 1.679e+02 2.380e+02, threshold=2.853e+02, percent-clipped=0.0
2024-09-23 01:24:25,333 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.28 vs. limit=10.0
2024-09-23 01:24:27,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=146197.33333333334, ans=0.125
2024-09-23 01:25:05,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=146290.66666666666, ans=0.0
2024-09-23 01:25:08,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=146290.66666666666, ans=0.0
2024-09-23 01:25:22,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=146337.33333333334, ans=0.1
2024-09-23 01:25:27,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=146337.33333333334, ans=0.125
2024-09-23 01:25:33,201 INFO [train.py:1198] (1/4) Epoch 9, batch 200, loss[loss=0.2682, ctc_loss=0.1863, cr_loss=0.4095, over 17098.00 frames. ], tot_loss[loss=0.2612, ctc_loss=0.1835, cr_loss=0.3884, over 2134147.06 frames. ], batch size: 49, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:25:53,335 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.83 vs. limit=5.0
2024-09-23 01:26:19,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=146524.0, ans=0.2
2024-09-23 01:26:40,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=146570.66666666666, ans=0.025
2024-09-23 01:26:40,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=146570.66666666666, ans=0.125
2024-09-23 01:26:54,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=146617.33333333334, ans=0.0
2024-09-23 01:26:55,831 INFO [train.py:1198] (1/4) Epoch 9, batch 250, loss[loss=0.2369, ctc_loss=0.1634, cr_loss=0.3672, over 16268.00 frames. ], tot_loss[loss=0.2623, ctc_loss=0.1844, cr_loss=0.3897, over 2404219.18 frames. ], batch size: 36, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:26:57,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=146617.33333333334, ans=0.125
2024-09-23 01:27:06,848 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.339e+02 1.502e+02 1.713e+02 2.875e+02, threshold=3.003e+02, percent-clipped=1.0
2024-09-23 01:27:29,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=146710.66666666666, ans=0.2
2024-09-23 01:27:43,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.27 vs. limit=22.5
2024-09-23 01:27:44,525 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.87 vs. limit=15.0
2024-09-23 01:27:45,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=146757.33333333334, ans=0.2
2024-09-23 01:27:46,726 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.04 vs. limit=15.0
2024-09-23 01:27:55,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=146757.33333333334, ans=0.125
2024-09-23 01:28:02,580 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.12 vs. limit=22.5
2024-09-23 01:28:17,488 INFO [train.py:1198] (1/4) Epoch 9, batch 300, loss[loss=0.2697, ctc_loss=0.1891, cr_loss=0.4032, over 17022.00 frames. ], tot_loss[loss=0.2612, ctc_loss=0.1834, cr_loss=0.389, over 2619598.92 frames. ], batch size: 44, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:28:38,029 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=10.88 vs. limit=15.0
2024-09-23 01:28:46,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=146897.33333333334, ans=0.025
2024-09-23 01:29:02,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=146944.0, ans=0.1
2024-09-23 01:29:32,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=147037.33333333334, ans=0.125
2024-09-23 01:29:37,417 INFO [train.py:1198] (1/4) Epoch 9, batch 350, loss[loss=0.2413, ctc_loss=0.1678, cr_loss=0.3674, over 17094.00 frames. ], tot_loss[loss=0.2618, ctc_loss=0.184, cr_loss=0.3888, over 2773379.01 frames. ], batch size: 43, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:29:48,845 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.338e+02 1.473e+02 1.743e+02 2.948e+02, threshold=2.946e+02, percent-clipped=0.0
2024-09-23 01:30:24,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=147177.33333333334, ans=0.125
2024-09-23 01:30:37,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=147224.0, ans=0.0
2024-09-23 01:30:39,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=147224.0, ans=0.125
2024-09-23 01:30:43,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=147224.0, ans=0.09899494936611666
2024-09-23 01:31:02,653 INFO [train.py:1198] (1/4) Epoch 9, batch 400, loss[loss=0.2528, ctc_loss=0.1766, cr_loss=0.381, over 17025.00 frames. ], tot_loss[loss=0.261, ctc_loss=0.1835, cr_loss=0.3875, over 2898589.79 frames. ], batch size: 44, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:31:09,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=147317.33333333334, ans=0.09899494936611666
2024-09-23 01:31:14,719 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.36 vs. limit=6.0
2024-09-23 01:31:26,904 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.82 vs. limit=15.0
2024-09-23 01:31:34,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=147364.0, ans=0.95
2024-09-23 01:31:37,787 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.03 vs. limit=15.0
2024-09-23 01:31:38,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=147410.66666666666, ans=0.0
2024-09-23 01:31:41,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.41 vs. limit=22.5
2024-09-23 01:31:41,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=7.13 vs. limit=15.0
2024-09-23 01:32:01,540 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 01:32:13,234 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.17 vs. limit=10.0
2024-09-23 01:32:25,078 INFO [train.py:1198] (1/4) Epoch 9, batch 450, loss[loss=0.226, ctc_loss=0.1541, cr_loss=0.3595, over 17088.00 frames. ], tot_loss[loss=0.2615, ctc_loss=0.1838, cr_loss=0.3883, over 3001091.79 frames. ], batch size: 43, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:32:31,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=147550.66666666666, ans=0.1
2024-09-23 01:32:37,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=147550.66666666666, ans=0.125
2024-09-23 01:32:38,951 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.318e+02 1.484e+02 1.790e+02 2.979e+02, threshold=2.969e+02, percent-clipped=1.0
2024-09-23 01:32:50,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=147597.33333333334, ans=0.125
2024-09-23 01:33:32,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=147737.33333333334, ans=0.125
2024-09-23 01:33:38,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=147737.33333333334, ans=0.07
2024-09-23 01:33:41,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=147737.33333333334, ans=0.025
2024-09-23 01:33:42,013 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.50 vs. limit=22.5
2024-09-23 01:33:46,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=147784.0, ans=10.0
2024-09-23 01:33:47,559 INFO [train.py:1198] (1/4) Epoch 9, batch 500, loss[loss=0.3304, ctc_loss=0.2392, cr_loss=0.4562, over 14947.00 frames. ], tot_loss[loss=0.2618, ctc_loss=0.184, cr_loss=0.3888, over 3079001.69 frames. ], batch size: 89, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:34:30,250 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.40 vs. limit=15.0
2024-09-23 01:34:43,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=147924.0, ans=0.0
2024-09-23 01:35:09,827 INFO [train.py:1198] (1/4) Epoch 9, batch 550, loss[loss=0.2506, ctc_loss=0.1789, cr_loss=0.3584, over 17092.00 frames. ], tot_loss[loss=0.26, ctc_loss=0.1825, cr_loss=0.3874, over 3147361.18 frames. ], batch size: 43, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:35:23,367 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.335e+02 1.443e+02 1.626e+02 2.688e+02, threshold=2.885e+02, percent-clipped=0.0
2024-09-23 01:35:57,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.16 vs. limit=10.0
2024-09-23 01:36:01,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=148157.33333333334, ans=0.125
2024-09-23 01:36:04,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=148157.33333333334, ans=0.0
2024-09-23 01:36:23,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=148204.0, ans=0.1
2024-09-23 01:36:34,062 INFO [train.py:1198] (1/4) Epoch 9, batch 600, loss[loss=0.2669, ctc_loss=0.1879, cr_loss=0.3948, over 17293.00 frames. ], tot_loss[loss=0.2603, ctc_loss=0.1826, cr_loss=0.3883, over 3195502.12 frames. ], batch size: 42, lr: 1.37e-02, grad_scale: 32.0
2024-09-23 01:36:37,824 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.11 vs. limit=15.0
2024-09-23 01:36:45,961 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.18 vs. limit=15.0
2024-09-23 01:37:06,070 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 01:37:26,591 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.53 vs. limit=8.0
2024-09-23 01:37:38,415 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=22.03 vs. limit=22.5
2024-09-23 01:37:56,745 INFO [train.py:1198] (1/4) Epoch 9, batch 650, loss[loss=0.2535, ctc_loss=0.1769, cr_loss=0.3829, over 17020.00 frames. ], tot_loss[loss=0.2606, ctc_loss=0.1828, cr_loss=0.3888, over 3229817.57 frames. ], batch size: 51, lr: 1.36e-02, grad_scale: 16.0
2024-09-23 01:37:57,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=148484.0, ans=0.2
2024-09-23 01:38:05,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=148484.0, ans=0.04949747468305833
2024-09-23 01:38:09,415 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.361e+02 1.468e+02 1.616e+02 2.267e+02, threshold=2.935e+02, percent-clipped=0.0
2024-09-23 01:38:25,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=148530.66666666666, ans=0.0
2024-09-23 01:38:39,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=148577.33333333334, ans=0.0
2024-09-23 01:39:01,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=148670.66666666666, ans=0.0
2024-09-23 01:39:15,865 INFO [train.py:1198] (1/4) Epoch 9, batch 700, loss[loss=0.2533, ctc_loss=0.1765, cr_loss=0.3842, over 17354.00 frames. ], tot_loss[loss=0.261, ctc_loss=0.1831, cr_loss=0.3897, over 3266118.12 frames. ], batch size: 48, lr: 1.36e-02, grad_scale: 16.0
2024-09-23 01:40:10,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=148857.33333333334, ans=0.125
2024-09-23 01:40:11,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=148857.33333333334, ans=0.0
2024-09-23 01:40:21,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=148857.33333333334, ans=0.125
2024-09-23 01:40:40,041 INFO [train.py:1198] (1/4) Epoch 9, batch 750, loss[loss=0.2357, ctc_loss=0.1643, cr_loss=0.3569, over 17306.00 frames. ], tot_loss[loss=0.2607, ctc_loss=0.183, cr_loss=0.3889, over 3291039.11 frames. ], batch size: 46, lr: 1.36e-02, grad_scale: 16.0
2024-09-23 01:40:52,737 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.365e+02 1.514e+02 1.779e+02 2.634e+02, threshold=3.027e+02, percent-clipped=0.0
2024-09-23 01:41:06,127 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 01:41:32,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=149090.66666666666, ans=0.1
2024-09-23 01:41:33,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.44 vs. limit=15.0
2024-09-23 01:42:02,474 INFO [train.py:1198] (1/4) Epoch 9, batch 800, loss[loss=0.3047, ctc_loss=0.2121, cr_loss=0.4628, over 17012.00 frames. ], tot_loss[loss=0.2602, ctc_loss=0.1827, cr_loss=0.3876, over 3293577.94 frames. ], batch size: 53, lr: 1.36e-02, grad_scale: 32.0
2024-09-23 01:42:31,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=149230.66666666666, ans=0.0
2024-09-23 01:42:31,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=149230.66666666666, ans=0.1
2024-09-23 01:42:38,063 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=10.53 vs. limit=12.0
2024-09-23 01:42:42,474 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=149277.33333333334, ans=0.125
2024-09-23 01:42:45,928 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.95 vs. limit=10.0
2024-09-23 01:43:13,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=149370.66666666666, ans=0.025
2024-09-23 01:43:27,282 INFO [train.py:1198] (1/4) Epoch 9, batch 850, loss[loss=0.2327, ctc_loss=0.1604, cr_loss=0.3614, over 17106.00 frames. ], tot_loss[loss=0.2594, ctc_loss=0.1821, cr_loss=0.3867, over 3312383.73 frames. ], batch size: 40, lr: 1.36e-02, grad_scale: 32.0
2024-09-23 01:43:38,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=149417.33333333334, ans=0.0
2024-09-23 01:43:39,941 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.394e+02 1.570e+02 1.849e+02 3.011e+02, threshold=3.140e+02, percent-clipped=0.0
2024-09-23 01:43:41,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=149464.0, ans=0.125
2024-09-23 01:44:41,505 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.37 vs. limit=6.0
2024-09-23 01:44:49,131 INFO [train.py:1198] (1/4) Epoch 9, batch 900, loss[loss=0.248, ctc_loss=0.1734, cr_loss=0.373, over 17313.00 frames. ], tot_loss[loss=0.26, ctc_loss=0.1826, cr_loss=0.3873, over 3317296.47 frames. ], batch size: 46, lr: 1.36e-02, grad_scale: 32.0
2024-09-23 01:45:16,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=149697.33333333334, ans=0.125
2024-09-23 01:45:24,705 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.86 vs. limit=15.0
2024-09-23 01:45:29,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=149744.0, ans=0.0
2024-09-23 01:45:40,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=149790.66666666666, ans=0.125
2024-09-23 01:45:53,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=149790.66666666666, ans=15.0
2024-09-23 01:45:54,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=149837.33333333334, ans=0.07
2024-09-23 01:46:14,333 INFO [train.py:1198] (1/4) Epoch 9, batch 950, loss[loss=0.2275, ctc_loss=0.157, cr_loss=0.3523, over 16839.00 frames. ], tot_loss[loss=0.2603, ctc_loss=0.1826, cr_loss=0.3882, over 3331460.11 frames. ], batch size: 37, lr: 1.36e-02, grad_scale: 32.0
2024-09-23 01:46:27,021 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.338e+02 1.417e+02 1.570e+02 2.385e+02, threshold=2.834e+02, percent-clipped=0.0
2024-09-23 01:46:27,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=149884.0, ans=0.125
2024-09-23 01:46:54,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=149977.33333333334, ans=0.2
2024-09-23 01:47:18,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=150070.66666666666, ans=0.1
2024-09-23 01:47:37,173 INFO [train.py:1198] (1/4) Epoch 9, batch 1000, loss[loss=0.2568, ctc_loss=0.1827, cr_loss=0.3707, over 17215.00 frames. ], tot_loss[loss=0.2595, ctc_loss=0.1821, cr_loss=0.387, over 3342258.83 frames. ], batch size: 47, lr: 1.36e-02, grad_scale: 32.0
2024-09-23 01:48:12,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=150210.66666666666, ans=0.2
2024-09-23 01:48:23,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=150257.33333333334, ans=0.125
2024-09-23 01:48:28,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=150257.33333333334, ans=0.125
2024-09-23 01:48:31,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=150257.33333333334, ans=0.05
2024-09-23 01:48:40,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=7.46 vs. limit=15.0
2024-09-23 01:48:57,105 INFO [train.py:1198] (1/4) Epoch 9, batch 1050, loss[loss=0.2985, ctc_loss=0.211, cr_loss=0.4374, over 17019.00 frames. ], tot_loss[loss=0.26, ctc_loss=0.1824, cr_loss=0.3882, over 3348396.91 frames. ], batch size: 51, lr: 1.36e-02, grad_scale: 32.0
2024-09-23 01:48:57,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=150350.66666666666, ans=0.125
2024-09-23 01:49:01,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=150350.66666666666, ans=0.025
2024-09-23 01:49:09,786 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.329e+02 1.425e+02 1.705e+02 2.859e+02, threshold=2.851e+02, percent-clipped=1.0
2024-09-23 01:49:23,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=150397.33333333334, ans=0.125
2024-09-23 01:49:32,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=150444.0, ans=0.0
2024-09-23 01:50:18,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=150537.33333333334, ans=0.025
2024-09-23 01:50:22,529 INFO [train.py:1198] (1/4) Epoch 9, batch 1100, loss[loss=0.2156, ctc_loss=0.147, cr_loss=0.3432, over 17187.00 frames. ], tot_loss[loss=0.26, ctc_loss=0.1823, cr_loss=0.3881, over 3351363.30 frames. ], batch size: 41, lr: 1.36e-02, grad_scale: 32.0
2024-09-23 01:50:51,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=150630.66666666666, ans=0.1
2024-09-23 01:51:13,808 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.72 vs. limit=22.5
2024-09-23 01:51:21,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=150724.0, ans=0.125
2024-09-23 01:51:25,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.66 vs. limit=15.0
2024-09-23 01:51:27,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=150770.66666666666, ans=0.025
2024-09-23 01:51:45,050 INFO [train.py:1198] (1/4) Epoch 9, batch 1150, loss[loss=0.2212, ctc_loss=0.1562, cr_loss=0.3249, over 17126.00 frames. ], tot_loss[loss=0.2583, ctc_loss=0.1811, cr_loss=0.3861, over 3360036.80 frames. ], batch size: 40, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 01:51:57,742 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.320e+02 1.454e+02 1.675e+02 2.569e+02, threshold=2.907e+02, percent-clipped=0.0
2024-09-23 01:52:17,564 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.45 vs. limit=15.0
2024-09-23 01:52:18,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=150910.66666666666, ans=0.125
2024-09-23 01:52:36,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=150957.33333333334, ans=0.0
2024-09-23 01:52:38,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=150957.33333333334, ans=0.125
2024-09-23 01:52:42,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.05 vs. limit=15.0
2024-09-23 01:52:43,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=150957.33333333334, ans=0.0
2024-09-23 01:52:59,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer_ff3.min_abs, batch_count=151004.0, ans=0.2
2024-09-23 01:53:01,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=151004.0, ans=0.0
2024-09-23 01:53:07,106 INFO [train.py:1198] (1/4) Epoch 9, batch 1200, loss[loss=0.3063, ctc_loss=0.217, cr_loss=0.4467, over 15945.00 frames. ], tot_loss[loss=0.2587, ctc_loss=0.1814, cr_loss=0.3865, over 3354671.34 frames. ], batch size: 74, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 01:53:21,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=151097.33333333334, ans=0.0
2024-09-23 01:54:04,581 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.19 vs. limit=22.5
2024-09-23 01:54:19,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=151237.33333333334, ans=0.125
2024-09-23 01:54:24,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=151237.33333333334, ans=0.125
2024-09-23 01:54:27,529 INFO [train.py:1198] (1/4) Epoch 9, batch 1250, loss[loss=0.2961, ctc_loss=0.2085, cr_loss=0.4381, over 17061.00 frames. ], tot_loss[loss=0.2588, ctc_loss=0.1816, cr_loss=0.3864, over 3345671.81 frames. ], batch size: 52, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 01:54:42,634 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.349e+02 1.478e+02 1.607e+02 2.970e+02, threshold=2.955e+02, percent-clipped=1.0
2024-09-23 01:54:46,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=151330.66666666666, ans=0.0
2024-09-23 01:55:06,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=151377.33333333334, ans=0.5
2024-09-23 01:55:24,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=151424.0, ans=0.0
2024-09-23 01:55:36,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=151470.66666666666, ans=0.125
2024-09-23 01:55:39,707 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.62 vs. limit=15.0
2024-09-23 01:55:51,559 INFO [train.py:1198] (1/4) Epoch 9, batch 1300, loss[loss=0.2523, ctc_loss=0.1769, cr_loss=0.3767, over 17001.00 frames. ], tot_loss[loss=0.2585, ctc_loss=0.1812, cr_loss=0.3861, over 3351789.37 frames. ], batch size: 56, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 01:57:01,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=151704.0, ans=0.125
2024-09-23 01:57:01,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=151704.0, ans=0.125
2024-09-23 01:57:09,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=151704.0, ans=0.1
2024-09-23 01:57:13,855 INFO [train.py:1198] (1/4) Epoch 9, batch 1350, loss[loss=0.2927, ctc_loss=0.2072, cr_loss=0.4279, over 17020.00 frames. ], tot_loss[loss=0.2594, ctc_loss=0.1821, cr_loss=0.3866, over 3346811.91 frames. ], batch size: 52, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 01:57:29,083 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.345e+02 1.485e+02 1.651e+02 2.569e+02, threshold=2.970e+02, percent-clipped=0.0
2024-09-23 01:57:32,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=151797.33333333334, ans=0.2
2024-09-23 01:58:17,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=151890.66666666666, ans=0.2
2024-09-23 01:58:35,901 INFO [train.py:1198] (1/4) Epoch 9, batch 1400, loss[loss=0.2347, ctc_loss=0.1583, cr_loss=0.3816, over 16232.00 frames. ], tot_loss[loss=0.2597, ctc_loss=0.1823, cr_loss=0.3869, over 3350344.92 frames. ], batch size: 36, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 01:58:47,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=151984.0, ans=0.125
2024-09-23 01:59:31,709 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 01:59:45,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=152170.66666666666, ans=0.0
2024-09-23 02:00:01,048 INFO [train.py:1198] (1/4) Epoch 9, batch 1450, loss[loss=0.2191, ctc_loss=0.1507, cr_loss=0.3424, over 16660.00 frames. ], tot_loss[loss=0.2597, ctc_loss=0.1823, cr_loss=0.3869, over 3343214.27 frames. ], batch size: 37, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 02:00:01,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=152217.33333333334, ans=0.125
2024-09-23 02:00:12,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=152217.33333333334, ans=0.125
2024-09-23 02:00:13,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.70 vs. limit=15.0
2024-09-23 02:00:13,884 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.414e+02 1.556e+02 1.752e+02 2.841e+02, threshold=3.113e+02, percent-clipped=0.0
2024-09-23 02:00:23,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=152264.0, ans=0.1
2024-09-23 02:00:24,438 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.05 vs. limit=15.0
2024-09-23 02:00:25,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=152264.0, ans=0.1
2024-09-23 02:00:27,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=152264.0, ans=0.125
2024-09-23 02:00:31,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=152310.66666666666, ans=0.125
2024-09-23 02:00:33,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=152310.66666666666, ans=0.1
2024-09-23 02:00:59,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=152357.33333333334, ans=0.2
2024-09-23 02:01:05,347 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.74 vs. limit=12.0
2024-09-23 02:01:21,036 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.77 vs. limit=15.0
2024-09-23 02:01:22,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=152450.66666666666, ans=0.125
2024-09-23 02:01:23,505 INFO [train.py:1198] (1/4) Epoch 9, batch 1500, loss[loss=0.2808, ctc_loss=0.1986, cr_loss=0.4109, over 17015.00 frames. ], tot_loss[loss=0.2591, ctc_loss=0.1817, cr_loss=0.387, over 3350650.40 frames. ], batch size: 52, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 02:01:23,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=152450.66666666666, ans=0.1
2024-09-23 02:01:49,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=152497.33333333334, ans=0.2
2024-09-23 02:02:01,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=152544.0, ans=0.125
2024-09-23 02:02:23,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=152590.66666666666, ans=0.125
2024-09-23 02:02:42,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=152637.33333333334, ans=0.0
2024-09-23 02:02:45,183 INFO [train.py:1198] (1/4) Epoch 9, batch 1550, loss[loss=0.2614, ctc_loss=0.1826, cr_loss=0.3942, over 17256.00 frames. ], tot_loss[loss=0.259, ctc_loss=0.1818, cr_loss=0.3864, over 3339671.63 frames. ], batch size: 44, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 02:02:51,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=152684.0, ans=0.1
2024-09-23 02:02:57,972 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.340e+02 1.461e+02 1.652e+02 2.342e+02, threshold=2.922e+02, percent-clipped=0.0
2024-09-23 02:03:18,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=152777.33333333334, ans=0.0
2024-09-23 02:03:23,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=152777.33333333334, ans=0.125
2024-09-23 02:03:25,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=152777.33333333334, ans=0.125
2024-09-23 02:03:26,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.66 vs. limit=6.0
2024-09-23 02:03:46,568 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.88 vs. limit=22.5
2024-09-23 02:03:50,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=152870.66666666666, ans=0.0
2024-09-23 02:04:04,956 INFO [train.py:1198] (1/4) Epoch 9, batch 1600, loss[loss=0.2857, ctc_loss=0.2012, cr_loss=0.4227, over 17096.00 frames. ], tot_loss[loss=0.2583, ctc_loss=0.1813, cr_loss=0.3852, over 3337336.75 frames. ], batch size: 49, lr: 1.35e-02, grad_scale: 32.0
2024-09-23 02:04:18,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=152917.33333333334, ans=0.0
2024-09-23 02:04:21,539 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=13.05 vs. limit=22.5
2024-09-23 02:04:34,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=152964.0, ans=0.125
2024-09-23 02:05:27,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=153104.0, ans=0.125
2024-09-23 02:05:30,267 INFO [train.py:1198] (1/4) Epoch 9, batch 1650, loss[loss=0.2591, ctc_loss=0.1749, cr_loss=0.4208, over 17061.00 frames. ], tot_loss[loss=0.2588, ctc_loss=0.1815, cr_loss=0.3865, over 3339065.49 frames. ], batch size: 39, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:05:42,186 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.77 vs. limit=15.0
2024-09-23 02:05:42,928 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.315e+02 1.418e+02 1.618e+02 2.447e+02, threshold=2.836e+02, percent-clipped=0.0
2024-09-23 02:06:31,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=153290.66666666666, ans=0.125
2024-09-23 02:06:39,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=153337.33333333334, ans=0.125
2024-09-23 02:06:52,589 INFO [train.py:1198] (1/4) Epoch 9, batch 1700, loss[loss=0.307, ctc_loss=0.2192, cr_loss=0.4389, over 15924.00 frames. ], tot_loss[loss=0.2582, ctc_loss=0.181, cr_loss=0.3857, over 3343214.98 frames. ], batch size: 74, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:06:56,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=153384.0, ans=0.125
2024-09-23 02:07:06,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.55 vs. limit=12.0
2024-09-23 02:07:19,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.min_positive, batch_count=153430.66666666666, ans=0.05
2024-09-23 02:07:35,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=153477.33333333334, ans=0.125
2024-09-23 02:07:38,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=153477.33333333334, ans=0.125
2024-09-23 02:08:14,978 INFO [train.py:1198] (1/4) Epoch 9, batch 1750, loss[loss=0.2656, ctc_loss=0.1837, cr_loss=0.4095, over 17132.00 frames. ], tot_loss[loss=0.2578, ctc_loss=0.1806, cr_loss=0.3856, over 3346014.83 frames. ], batch size: 48, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:08:27,630 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.306e+02 1.435e+02 1.601e+02 2.532e+02, threshold=2.871e+02, percent-clipped=0.0
2024-09-23 02:08:32,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=153664.0, ans=0.05
2024-09-23 02:08:34,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=153664.0, ans=0.0
2024-09-23 02:09:13,049 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.56 vs. limit=6.0
2024-09-23 02:09:37,154 INFO [train.py:1198] (1/4) Epoch 9, batch 1800, loss[loss=0.233, ctc_loss=0.1651, cr_loss=0.3394, over 17145.00 frames. ], tot_loss[loss=0.2578, ctc_loss=0.1807, cr_loss=0.3858, over 3353046.58 frames. ], batch size: 48, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:10:05,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=153897.33333333334, ans=0.05
2024-09-23 02:10:33,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=153990.66666666666, ans=0.1
2024-09-23 02:10:36,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=153990.66666666666, ans=0.0
2024-09-23 02:10:53,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=154037.33333333334, ans=0.125
2024-09-23 02:10:54,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=154037.33333333334, ans=0.1
2024-09-23 02:10:54,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=154037.33333333334, ans=0.125
2024-09-23 02:11:02,433 INFO [train.py:1198] (1/4) Epoch 9, batch 1850, loss[loss=0.2625, ctc_loss=0.1819, cr_loss=0.4029, over 16821.00 frames. ], tot_loss[loss=0.2578, ctc_loss=0.1806, cr_loss=0.386, over 3352045.25 frames. ], batch size: 58, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:11:15,289 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.311e+02 1.455e+02 1.599e+02 2.363e+02, threshold=2.909e+02, percent-clipped=0.0
2024-09-23 02:11:17,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=154130.66666666666, ans=0.1
2024-09-23 02:11:36,771 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.79 vs. limit=15.0
2024-09-23 02:11:37,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=154177.33333333334, ans=0.0
2024-09-23 02:11:58,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=154224.0, ans=0.2
2024-09-23 02:12:24,532 INFO [train.py:1198] (1/4) Epoch 9, batch 1900, loss[loss=0.2194, ctc_loss=0.153, cr_loss=0.3317, over 16957.00 frames. ], tot_loss[loss=0.2577, ctc_loss=0.1806, cr_loss=0.3856, over 3341437.47 frames. ], batch size: 42, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:12:46,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=154364.0, ans=0.2
2024-09-23 02:12:56,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=154410.66666666666, ans=0.125
2024-09-23 02:13:21,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=154457.33333333334, ans=0.125
2024-09-23 02:13:23,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=154457.33333333334, ans=0.1
2024-09-23 02:13:43,681 INFO [train.py:1198] (1/4) Epoch 9, batch 1950, loss[loss=0.255, ctc_loss=0.1775, cr_loss=0.3871, over 17053.00 frames. ], tot_loss[loss=0.2587, ctc_loss=0.1814, cr_loss=0.3865, over 3333202.01 frames. ], batch size: 52, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:13:44,076 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 02:13:50,693 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=4.41 vs. limit=12.0
2024-09-23 02:13:56,415 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.387e+02 1.540e+02 1.769e+02 3.177e+02, threshold=3.081e+02, percent-clipped=1.0
2024-09-23 02:15:09,133 INFO [train.py:1198] (1/4) Epoch 9, batch 2000, loss[loss=0.2387, ctc_loss=0.1658, cr_loss=0.3643, over 17261.00 frames. ], tot_loss[loss=0.2578, ctc_loss=0.1805, cr_loss=0.3862, over 3348172.48 frames. ], batch size: 44, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:15:09,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=154784.0, ans=0.125
2024-09-23 02:15:10,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=154784.0, ans=0.2
2024-09-23 02:15:14,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=154784.0, ans=0.0
2024-09-23 02:15:17,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=154784.0, ans=0.025
2024-09-23 02:15:38,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=154830.66666666666, ans=0.125
2024-09-23 02:15:48,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=154877.33333333334, ans=0.125
2024-09-23 02:16:01,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=154924.0, ans=0.125
2024-09-23 02:16:07,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=154924.0, ans=0.1
2024-09-23 02:16:31,230 INFO [train.py:1198] (1/4) Epoch 9, batch 2050, loss[loss=0.2968, ctc_loss=0.2126, cr_loss=0.4209, over 16911.00 frames. ], tot_loss[loss=0.2595, ctc_loss=0.182, cr_loss=0.3876, over 3350942.95 frames. ], batch size: 58, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:16:38,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=155017.33333333334, ans=0.0
2024-09-23 02:16:43,966 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.371e+02 1.510e+02 1.685e+02 3.292e+02, threshold=3.020e+02, percent-clipped=1.0
2024-09-23 02:16:52,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=155064.0, ans=0.125
2024-09-23 02:17:15,977 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=15.53 vs. limit=22.5
2024-09-23 02:17:53,734 INFO [train.py:1198] (1/4) Epoch 9, batch 2100, loss[loss=0.2478, ctc_loss=0.1743, cr_loss=0.3676, over 17182.00 frames. ], tot_loss[loss=0.2583, ctc_loss=0.1809, cr_loss=0.3868, over 3361553.86 frames. ], batch size: 41, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:17:58,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=155250.66666666666, ans=0.125
2024-09-23 02:18:43,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=155390.66666666666, ans=0.025
2024-09-23 02:18:46,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=155390.66666666666, ans=0.025
2024-09-23 02:19:12,972 INFO [train.py:1198] (1/4) Epoch 9, batch 2150, loss[loss=0.2299, ctc_loss=0.1601, cr_loss=0.3491, over 17145.00 frames. ], tot_loss[loss=0.2582, ctc_loss=0.1809, cr_loss=0.3869, over 3360965.22 frames. ], batch size: 45, lr: 1.34e-02, grad_scale: 32.0
2024-09-23 02:19:27,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.57 vs. limit=12.0
2024-09-23 02:19:28,282 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.379e+02 1.514e+02 1.800e+02 2.768e+02, threshold=3.028e+02, percent-clipped=0.0
2024-09-23 02:19:28,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=155484.0, ans=0.125
2024-09-23 02:19:28,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer_ff2.min_abs, batch_count=155484.0, ans=0.1
2024-09-23 02:19:38,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=155530.66666666666, ans=0.125
2024-09-23 02:20:09,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.00 vs. limit=15.0
2024-09-23 02:20:22,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=155670.66666666666, ans=0.2
2024-09-23 02:20:32,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=155670.66666666666, ans=0.0
2024-09-23 02:20:32,725 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.75 vs. limit=15.0
2024-09-23 02:20:36,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=155717.33333333334, ans=0.0
2024-09-23 02:20:38,244 INFO [train.py:1198] (1/4) Epoch 9, batch 2200, loss[loss=0.1906, ctc_loss=0.129, cr_loss=0.3081, over 17037.00 frames. ], tot_loss[loss=0.2573, ctc_loss=0.18, cr_loss=0.3863, over 3369811.33 frames. ], batch size: 39, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:20:42,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=155717.33333333334, ans=0.5
2024-09-23 02:20:56,097 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.74 vs. limit=10.0
2024-09-23 02:21:01,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=155764.0, ans=0.0
2024-09-23 02:21:18,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=155810.66666666666, ans=0.025
2024-09-23 02:21:22,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=155810.66666666666, ans=10.0
2024-09-23 02:21:38,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=155857.33333333334, ans=0.125
2024-09-23 02:21:45,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=155904.0, ans=0.1
2024-09-23 02:22:03,746 INFO [train.py:1198] (1/4) Epoch 9, batch 2250, loss[loss=0.2288, ctc_loss=0.158, cr_loss=0.354, over 17247.00 frames. ], tot_loss[loss=0.2584, ctc_loss=0.181, cr_loss=0.3874, over 3361256.25 frames. ], batch size: 44, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:22:13,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=155950.66666666666, ans=0.2
2024-09-23 02:22:16,468 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.340e+02 1.524e+02 1.728e+02 3.121e+02, threshold=3.047e+02, percent-clipped=1.0
2024-09-23 02:22:27,053 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.64 vs. limit=22.5
2024-09-23 02:22:47,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=156044.0, ans=0.125
2024-09-23 02:23:08,274 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.74 vs. limit=15.0
2024-09-23 02:23:17,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer_ff2.min_abs, batch_count=156137.33333333334, ans=0.1
2024-09-23 02:23:23,327 INFO [train.py:1198] (1/4) Epoch 9, batch 2300, loss[loss=0.2446, ctc_loss=0.1706, cr_loss=0.3702, over 17011.00 frames. ], tot_loss[loss=0.2589, ctc_loss=0.1814, cr_loss=0.3875, over 3358851.12 frames. ], batch size: 51, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:23:36,818 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.72 vs. limit=15.0
2024-09-23 02:23:49,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=156230.66666666666, ans=0.2
2024-09-23 02:24:00,572 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 02:24:33,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=156370.66666666666, ans=0.125
2024-09-23 02:24:48,596 INFO [train.py:1198] (1/4) Epoch 9, batch 2350, loss[loss=0.1786, ctc_loss=0.1194, cr_loss=0.296, over 17026.00 frames. ], tot_loss[loss=0.2592, ctc_loss=0.1815, cr_loss=0.3885, over 3358996.31 frames. ], batch size: 39, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:25:01,204 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.341e+02 1.440e+02 1.582e+02 2.935e+02, threshold=2.879e+02, percent-clipped=0.0
2024-09-23 02:25:01,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=156417.33333333334, ans=0.0
2024-09-23 02:25:03,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=156464.0, ans=0.0
2024-09-23 02:25:17,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=156464.0, ans=0.125
2024-09-23 02:25:38,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=156557.33333333334, ans=0.1
2024-09-23 02:25:45,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=156557.33333333334, ans=0.125
2024-09-23 02:26:02,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=156604.0, ans=0.125
2024-09-23 02:26:10,108 INFO [train.py:1198] (1/4) Epoch 9, batch 2400, loss[loss=0.2657, ctc_loss=0.185, cr_loss=0.4037, over 16052.00 frames. ], tot_loss[loss=0.2578, ctc_loss=0.1805, cr_loss=0.3869, over 3361589.84 frames. ], batch size: 74, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:26:17,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.66 vs. limit=15.0
2024-09-23 02:26:32,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=156697.33333333334, ans=0.125
2024-09-23 02:26:46,008 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.42 vs. limit=15.0
2024-09-23 02:27:02,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=156790.66666666666, ans=0.125
2024-09-23 02:27:07,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=156790.66666666666, ans=0.0
2024-09-23 02:27:26,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=156837.33333333334, ans=0.125
2024-09-23 02:27:29,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=156837.33333333334, ans=0.2
2024-09-23 02:27:31,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=156884.0, ans=0.0
2024-09-23 02:27:32,764 INFO [train.py:1198] (1/4) Epoch 9, batch 2450, loss[loss=0.2406, ctc_loss=0.1676, cr_loss=0.3651, over 17015.00 frames. ], tot_loss[loss=0.2572, ctc_loss=0.1798, cr_loss=0.3868, over 3368597.92 frames. ], batch size: 51, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:27:41,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=156884.0, ans=0.1
2024-09-23 02:27:45,479 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.502e+02 1.626e+02 1.858e+02 2.761e+02, threshold=3.252e+02, percent-clipped=0.0
2024-09-23 02:28:27,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=157024.0, ans=0.125
2024-09-23 02:28:37,650 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.11 vs. limit=12.0
2024-09-23 02:28:40,976 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.73 vs. limit=15.0
2024-09-23 02:28:41,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=157070.66666666666, ans=0.125
2024-09-23 02:28:43,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=157070.66666666666, ans=0.025
2024-09-23 02:28:46,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=157070.66666666666, ans=0.2
2024-09-23 02:28:52,842 INFO [train.py:1198] (1/4) Epoch 9, batch 2500, loss[loss=0.2604, ctc_loss=0.1792, cr_loss=0.406, over 17214.00 frames. ], tot_loss[loss=0.2576, ctc_loss=0.1803, cr_loss=0.3866, over 3359115.10 frames. ], batch size: 47, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:28:53,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=157117.33333333334, ans=0.0
2024-09-23 02:29:18,463 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.55 vs. limit=15.0
2024-09-23 02:29:35,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=157210.66666666666, ans=0.0
2024-09-23 02:29:38,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=157210.66666666666, ans=15.0
2024-09-23 02:29:55,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=157257.33333333334, ans=0.0
2024-09-23 02:29:59,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=157257.33333333334, ans=0.125
2024-09-23 02:30:00,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=157304.0, ans=0.1
2024-09-23 02:30:04,278 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.05 vs. limit=15.0
2024-09-23 02:30:13,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=157304.0, ans=0.125
2024-09-23 02:30:18,044 INFO [train.py:1198] (1/4) Epoch 9, batch 2550, loss[loss=0.289, ctc_loss=0.2037, cr_loss=0.4267, over 17005.00 frames. ], tot_loss[loss=0.2575, ctc_loss=0.1801, cr_loss=0.387, over 3351952.80 frames. ], batch size: 53, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:30:23,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=157350.66666666666, ans=0.1
2024-09-23 02:30:30,729 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.377e+02 1.519e+02 1.754e+02 2.605e+02, threshold=3.038e+02, percent-clipped=0.0
2024-09-23 02:30:54,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=157444.0, ans=0.2
2024-09-23 02:30:59,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=157444.0, ans=0.0
2024-09-23 02:31:01,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=16.30 vs. limit=22.5
2024-09-23 02:31:03,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=157444.0, ans=0.5
2024-09-23 02:31:08,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=157490.66666666666, ans=0.07
2024-09-23 02:31:40,940 INFO [train.py:1198] (1/4) Epoch 9, batch 2600, loss[loss=0.2632, ctc_loss=0.183, cr_loss=0.401, over 16849.00 frames. ], tot_loss[loss=0.2593, ctc_loss=0.1816, cr_loss=0.3885, over 3348718.82 frames. ], batch size: 58, lr: 1.33e-02, grad_scale: 32.0
2024-09-23 02:31:50,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=157584.0, ans=0.0
2024-09-23 02:31:50,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=157584.0, ans=0.1
2024-09-23 02:31:53,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.52 vs. limit=15.0
2024-09-23 02:32:03,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=157630.66666666666, ans=0.1
2024-09-23 02:32:07,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=157630.66666666666, ans=0.1
2024-09-23 02:32:18,289 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.60 vs. limit=15.0
2024-09-23 02:33:04,040 INFO [train.py:1198] (1/4) Epoch 9, batch 2650, loss[loss=0.2484, ctc_loss=0.1685, cr_loss=0.3994, over 17263.00 frames. ], tot_loss[loss=0.2588, ctc_loss=0.1813, cr_loss=0.3875, over 3352266.61 frames. ], batch size: 42, lr: 1.33e-02, grad_scale: 64.0
2024-09-23 02:33:16,814 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.311e+02 1.423e+02 1.645e+02 2.492e+02, threshold=2.847e+02, percent-clipped=0.0
2024-09-23 02:34:14,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=158004.0, ans=0.125
2024-09-23 02:34:17,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=158004.0, ans=0.2
2024-09-23 02:34:26,716 INFO [train.py:1198] (1/4) Epoch 9, batch 2700, loss[loss=0.269, ctc_loss=0.1928, cr_loss=0.3809, over 16786.00 frames. ], tot_loss[loss=0.2583, ctc_loss=0.1809, cr_loss=0.387, over 3357860.49 frames. ], batch size: 61, lr: 1.32e-02, grad_scale: 64.0
2024-09-23 02:34:50,519 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=15.21 vs. limit=15.0
2024-09-23 02:34:53,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=158097.33333333334, ans=0.1
2024-09-23 02:35:12,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=158144.0, ans=0.125
2024-09-23 02:35:23,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=158190.66666666666, ans=0.2
2024-09-23 02:35:25,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=158190.66666666666, ans=0.2
2024-09-23 02:35:39,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=158237.33333333334, ans=0.1
2024-09-23 02:35:48,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=158237.33333333334, ans=0.2
2024-09-23 02:35:51,927 INFO [train.py:1198] (1/4) Epoch 9, batch 2750, loss[loss=0.2821, ctc_loss=0.1967, cr_loss=0.427, over 17214.00 frames. ], tot_loss[loss=0.2582, ctc_loss=0.1808, cr_loss=0.3869, over 3345044.52 frames. ], batch size: 50, lr: 1.32e-02, grad_scale: 32.0
2024-09-23 02:36:06,252 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.343e+02 1.485e+02 1.822e+02 3.173e+02, threshold=2.970e+02, percent-clipped=1.0
2024-09-23 02:36:33,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=158377.33333333334, ans=0.025
2024-09-23 02:36:57,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=158470.66666666666, ans=0.125
2024-09-23 02:37:11,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=158470.66666666666, ans=0.2
2024-09-23 02:37:14,489 INFO [train.py:1198] (1/4) Epoch 9, batch 2800, loss[loss=0.2181, ctc_loss=0.1463, cr_loss=0.3591, over 17163.00 frames. ], tot_loss[loss=0.2571, ctc_loss=0.1799, cr_loss=0.3862, over 3351440.69 frames. ], batch size: 41, lr: 1.32e-02, grad_scale: 32.0
2024-09-23 02:37:32,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=158564.0, ans=0.125
2024-09-23 02:37:45,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=158610.66666666666, ans=0.0
2024-09-23 02:37:47,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=158610.66666666666, ans=0.1
2024-09-23 02:37:53,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=158610.66666666666, ans=0.125
2024-09-23 02:38:01,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=158657.33333333334, ans=0.125
2024-09-23 02:38:14,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=158657.33333333334, ans=0.0
2024-09-23 02:38:14,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=158657.33333333334, ans=0.125
2024-09-23 02:38:18,298 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.76 vs. limit=15.0
2024-09-23 02:38:30,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=158704.0, ans=0.125
2024-09-23 02:38:34,672 INFO [train.py:1198] (1/4) Epoch 9, batch 2850, loss[loss=0.2194, ctc_loss=0.1526, cr_loss=0.3341, over 17245.00 frames. ], tot_loss[loss=0.2563, ctc_loss=0.1792, cr_loss=0.3854, over 3354273.58 frames. ], batch size: 44, lr: 1.32e-02, grad_scale: 16.0
2024-09-23 02:38:42,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=158750.66666666666, ans=0.0
2024-09-23 02:38:50,462 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.314e+02 1.411e+02 1.546e+02 2.607e+02, threshold=2.821e+02, percent-clipped=0.0
2024-09-23 02:38:51,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=158797.33333333334, ans=15.0
2024-09-23 02:39:11,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=158844.0, ans=0.025
2024-09-23 02:39:12,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=158844.0, ans=0.05
2024-09-23 02:39:19,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=158844.0, ans=0.1
2024-09-23 02:39:25,727 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.51 vs. limit=15.0
2024-09-23 02:39:27,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.61 vs. limit=6.0
2024-09-23 02:39:59,460 INFO [train.py:1198] (1/4) Epoch 9, batch 2900, loss[loss=0.2128, ctc_loss=0.1463, cr_loss=0.3323, over 16225.00 frames. ], tot_loss[loss=0.2561, ctc_loss=0.1791, cr_loss=0.385, over 3357928.42 frames. ], batch size: 36, lr: 1.32e-02, grad_scale: 16.0
2024-09-23 02:40:00,447 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.76 vs. limit=22.5
2024-09-23 02:40:20,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=159030.66666666666, ans=0.0
2024-09-23 02:40:20,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=159030.66666666666, ans=0.1
2024-09-23 02:41:21,662 INFO [train.py:1198] (1/4) Epoch 9, batch 2950, loss[loss=0.2515, ctc_loss=0.1701, cr_loss=0.4067, over 17200.00 frames. ], tot_loss[loss=0.257, ctc_loss=0.1797, cr_loss=0.3863, over 3358379.52 frames. ], batch size: 47, lr: 1.32e-02, grad_scale: 16.0
2024-09-23 02:41:40,277 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.351e+02 1.516e+02 1.686e+02 2.482e+02, threshold=3.032e+02, percent-clipped=0.0
2024-09-23 02:41:45,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.66 vs. limit=6.0
2024-09-23 02:42:02,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=159310.66666666666, ans=0.0
2024-09-23 02:42:15,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=159357.33333333334, ans=0.1
2024-09-23 02:42:43,437 INFO [train.py:1198] (1/4) Epoch 9, batch 3000, loss[loss=0.2671, ctc_loss=0.1881, cr_loss=0.395, over 16939.00 frames. ], tot_loss[loss=0.2568, ctc_loss=0.1796, cr_loss=0.3859, over 3355337.31 frames. ], batch size: 58, lr: 1.32e-02, grad_scale: 16.0
2024-09-23 02:42:43,438 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 02:42:56,352 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.5.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([1.9897, 4.2272, 4.0185, 4.3423], device='cuda:1')
2024-09-23 02:42:57,328 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.1.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.3597, 4.1748, 3.6985, 4.1725], device='cuda:1')
2024-09-23 02:42:59,050 INFO [train.py:1230] (1/4) Epoch 9, validation: loss=0.05024, ctc_loss=0.05024, cr_loss=7.059e-15, over 944034.00 frames. 
2024-09-23 02:42:59,051 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 02:42:59,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=159450.66666666666, ans=0.0
2024-09-23 02:43:46,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=159590.66666666666, ans=0.0
2024-09-23 02:44:08,954 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.66 vs. limit=6.0
2024-09-23 02:44:17,778 INFO [train.py:1198] (1/4) Epoch 9, batch 3050, loss[loss=0.2769, ctc_loss=0.199, cr_loss=0.3897, over 17295.00 frames. ], tot_loss[loss=0.2575, ctc_loss=0.1802, cr_loss=0.3867, over 3354045.20 frames. ], batch size: 51, lr: 1.32e-02, grad_scale: 16.0
2024-09-23 02:44:33,651 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.314e+02 1.416e+02 1.662e+02 2.316e+02, threshold=2.832e+02, percent-clipped=0.0
2024-09-23 02:44:54,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=159777.33333333334, ans=0.125
2024-09-23 02:45:35,916 INFO [train.py:1198] (1/4) Epoch 9, batch 3100, loss[loss=0.2673, ctc_loss=0.1859, cr_loss=0.4072, over 16996.00 frames. ], tot_loss[loss=0.2573, ctc_loss=0.18, cr_loss=0.3863, over 3354285.12 frames. ], batch size: 51, lr: 1.32e-02, grad_scale: 16.0
2024-09-23 02:46:16,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=160010.66666666666, ans=0.1
2024-09-23 02:46:20,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=160010.66666666666, ans=0.0
2024-09-23 02:46:24,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=160057.33333333334, ans=0.125
2024-09-23 02:46:25,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=160057.33333333334, ans=0.125
2024-09-23 02:46:36,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=160057.33333333334, ans=0.09899494936611666
2024-09-23 02:46:59,396 INFO [train.py:1198] (1/4) Epoch 9, batch 3150, loss[loss=0.2608, ctc_loss=0.1827, cr_loss=0.3902, over 17135.00 frames. ], tot_loss[loss=0.2579, ctc_loss=0.1805, cr_loss=0.387, over 3358993.50 frames. ], batch size: 48, lr: 1.32e-02, grad_scale: 16.0
2024-09-23 02:47:15,043 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.398e+02 1.504e+02 1.697e+02 2.388e+02, threshold=3.008e+02, percent-clipped=0.0
2024-09-23 02:47:49,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=160290.66666666666, ans=0.125
2024-09-23 02:48:13,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=160337.33333333334, ans=0.125
2024-09-23 02:48:13,403 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 02:48:17,649 INFO [train.py:1198] (1/4) Epoch 9, batch 3200, loss[loss=0.2143, ctc_loss=0.1447, cr_loss=0.3479, over 16953.00 frames. ], tot_loss[loss=0.2573, ctc_loss=0.18, cr_loss=0.3863, over 3369747.29 frames. ], batch size: 42, lr: 1.32e-02, grad_scale: 32.0
2024-09-23 02:48:17,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=160384.0, ans=0.125
2024-09-23 02:48:18,451 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=8.38 vs. limit=15.0
2024-09-23 02:48:39,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=160430.66666666666, ans=0.1
2024-09-23 02:48:58,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=160477.33333333334, ans=0.09899494936611666
2024-09-23 02:49:03,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=160477.33333333334, ans=0.125
2024-09-23 02:49:23,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=160570.66666666666, ans=0.125
2024-09-23 02:49:34,347 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 02:49:38,713 INFO [train.py:1198] (1/4) Epoch 9, batch 3250, loss[loss=0.2748, ctc_loss=0.1921, cr_loss=0.4137, over 17057.00 frames. ], tot_loss[loss=0.2567, ctc_loss=0.1795, cr_loss=0.3863, over 3372654.48 frames. ], batch size: 46, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 02:49:54,281 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.347e+02 1.464e+02 1.659e+02 3.194e+02, threshold=2.929e+02, percent-clipped=1.0
2024-09-23 02:49:57,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=160664.0, ans=0.125
2024-09-23 02:50:12,176 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=9.90 vs. limit=15.0
2024-09-23 02:50:50,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=160804.0, ans=0.125
2024-09-23 02:50:56,581 INFO [train.py:1198] (1/4) Epoch 9, batch 3300, loss[loss=0.277, ctc_loss=0.1923, cr_loss=0.4235, over 17039.00 frames. ], tot_loss[loss=0.2566, ctc_loss=0.1794, cr_loss=0.3859, over 3370438.70 frames. ], batch size: 56, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 02:51:00,510 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.03 vs. limit=15.0
2024-09-23 02:51:12,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=160897.33333333334, ans=0.2
2024-09-23 02:51:22,470 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=5.06 vs. limit=12.0
2024-09-23 02:51:24,319 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.23 vs. limit=12.0
2024-09-23 02:51:26,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=160897.33333333334, ans=0.0
2024-09-23 02:51:43,738 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=3.180e-02
2024-09-23 02:51:48,884 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.36 vs. limit=6.0
2024-09-23 02:51:59,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=161037.33333333334, ans=0.125
2024-09-23 02:52:13,725 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.45 vs. limit=10.0
2024-09-23 02:52:15,711 INFO [train.py:1198] (1/4) Epoch 9, batch 3350, loss[loss=0.2528, ctc_loss=0.1747, cr_loss=0.3904, over 16865.00 frames. ], tot_loss[loss=0.2576, ctc_loss=0.1802, cr_loss=0.3873, over 3368215.50 frames. ], batch size: 58, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 02:52:15,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=161084.0, ans=0.0
2024-09-23 02:52:19,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=161084.0, ans=0.0
2024-09-23 02:52:31,265 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.345e+02 1.546e+02 1.825e+02 3.271e+02, threshold=3.093e+02, percent-clipped=1.0
2024-09-23 02:52:36,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=161130.66666666666, ans=0.125
2024-09-23 02:52:39,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=161130.66666666666, ans=0.025
2024-09-23 02:52:46,550 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=5.01 vs. limit=5.0
2024-09-23 02:53:04,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=161224.0, ans=0.125
2024-09-23 02:53:10,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=161224.0, ans=0.125
2024-09-23 02:53:15,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=161224.0, ans=0.09899494936611666
2024-09-23 02:53:26,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=161270.66666666666, ans=0.125
2024-09-23 02:53:31,192 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.03 vs. limit=12.0
2024-09-23 02:53:33,968 INFO [train.py:1198] (1/4) Epoch 9, batch 3400, loss[loss=0.2679, ctc_loss=0.1847, cr_loss=0.4165, over 17042.00 frames. ], tot_loss[loss=0.2576, ctc_loss=0.1802, cr_loss=0.3872, over 3366833.27 frames. ], batch size: 56, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 02:53:58,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=161364.0, ans=0.125
2024-09-23 02:54:05,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=161410.66666666666, ans=0.0
2024-09-23 02:54:25,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=161457.33333333334, ans=0.2
2024-09-23 02:54:48,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=161504.0, ans=0.125
2024-09-23 02:54:51,710 INFO [train.py:1198] (1/4) Epoch 9, batch 3450, loss[loss=0.2648, ctc_loss=0.1848, cr_loss=0.4, over 17221.00 frames. ], tot_loss[loss=0.2567, ctc_loss=0.1795, cr_loss=0.3861, over 3368854.83 frames. ], batch size: 50, lr: 1.31e-02, grad_scale: 16.0
2024-09-23 02:55:08,998 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.381e+02 1.554e+02 1.871e+02 2.951e+02, threshold=3.107e+02, percent-clipped=0.0
2024-09-23 02:55:23,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=161644.0, ans=0.0
2024-09-23 02:55:35,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=161644.0, ans=0.2
2024-09-23 02:56:12,426 INFO [train.py:1198] (1/4) Epoch 9, batch 3500, loss[loss=0.2618, ctc_loss=0.1828, cr_loss=0.3953, over 17072.00 frames. ], tot_loss[loss=0.2549, ctc_loss=0.178, cr_loss=0.3846, over 3373199.40 frames. ], batch size: 46, lr: 1.31e-02, grad_scale: 16.0
2024-09-23 02:56:18,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=161784.0, ans=0.025
2024-09-23 02:56:21,172 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.89 vs. limit=22.5
2024-09-23 02:56:47,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=161877.33333333334, ans=0.2
2024-09-23 02:57:17,096 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 02:57:18,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=161970.66666666666, ans=0.125
2024-09-23 02:57:29,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=161970.66666666666, ans=0.125
2024-09-23 02:57:32,419 INFO [train.py:1198] (1/4) Epoch 9, batch 3550, loss[loss=0.2786, ctc_loss=0.196, cr_loss=0.4132, over 17349.00 frames. ], tot_loss[loss=0.2554, ctc_loss=0.1783, cr_loss=0.3852, over 3363925.20 frames. ], batch size: 48, lr: 1.31e-02, grad_scale: 16.0
2024-09-23 02:57:32,924 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.74 vs. limit=12.0
2024-09-23 02:57:38,049 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.07 vs. limit=15.0
2024-09-23 02:57:49,779 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.397e+02 1.544e+02 1.862e+02 4.630e+02, threshold=3.088e+02, percent-clipped=2.0
2024-09-23 02:57:51,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=162064.0, ans=0.025
2024-09-23 02:58:24,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=162157.33333333334, ans=0.1
2024-09-23 02:58:41,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=162204.0, ans=0.2
2024-09-23 02:58:50,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=162250.66666666666, ans=0.0
2024-09-23 02:58:52,213 INFO [train.py:1198] (1/4) Epoch 9, batch 3600, loss[loss=0.2255, ctc_loss=0.1562, cr_loss=0.3465, over 16945.00 frames. ], tot_loss[loss=0.2546, ctc_loss=0.1777, cr_loss=0.3842, over 3372215.91 frames. ], batch size: 42, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 02:58:52,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=162250.66666666666, ans=0.2
2024-09-23 02:58:53,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.93 vs. limit=15.0
2024-09-23 02:58:54,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=162250.66666666666, ans=0.1
2024-09-23 02:59:19,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=162297.33333333334, ans=0.0
2024-09-23 02:59:36,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten.whitening_limit, batch_count=162344.0, ans=15.0
2024-09-23 02:59:37,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=162390.66666666666, ans=0.2
2024-09-23 02:59:39,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=162390.66666666666, ans=0.025
2024-09-23 03:00:05,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=162437.33333333334, ans=0.1
2024-09-23 03:00:10,090 INFO [train.py:1198] (1/4) Epoch 9, batch 3650, loss[loss=0.3085, ctc_loss=0.2197, cr_loss=0.444, over 17209.00 frames. ], tot_loss[loss=0.2553, ctc_loss=0.1783, cr_loss=0.3848, over 3360958.69 frames. ], batch size: 50, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 03:00:16,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=162484.0, ans=0.125
2024-09-23 03:00:27,444 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.388e+02 1.522e+02 1.765e+02 2.573e+02, threshold=3.044e+02, percent-clipped=0.0
2024-09-23 03:00:44,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=162577.33333333334, ans=0.125
2024-09-23 03:00:56,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=162577.33333333334, ans=0.125
2024-09-23 03:01:07,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=162624.0, ans=0.0
2024-09-23 03:01:16,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=162670.66666666666, ans=0.0
2024-09-23 03:01:30,906 INFO [train.py:1198] (1/4) Epoch 9, batch 3700, loss[loss=0.26, ctc_loss=0.1813, cr_loss=0.3935, over 17022.00 frames. ], tot_loss[loss=0.2551, ctc_loss=0.1783, cr_loss=0.3841, over 3363253.70 frames. ], batch size: 52, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 03:02:06,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten.whitening_limit, batch_count=162810.66666666666, ans=15.0
2024-09-23 03:02:18,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=162857.33333333334, ans=0.125
2024-09-23 03:02:39,630 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=162904.0, ans=0.09899494936611666
2024-09-23 03:02:45,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=162904.0, ans=0.0
2024-09-23 03:02:48,723 INFO [train.py:1198] (1/4) Epoch 9, batch 3750, loss[loss=0.2909, ctc_loss=0.201, cr_loss=0.4496, over 16898.00 frames. ], tot_loss[loss=0.2563, ctc_loss=0.1793, cr_loss=0.385, over 3340744.07 frames. ], batch size: 58, lr: 1.31e-02, grad_scale: 32.0
2024-09-23 03:02:50,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=162950.66666666666, ans=0.125
2024-09-23 03:03:05,821 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.286e+02 1.440e+02 1.620e+02 2.372e+02, threshold=2.880e+02, percent-clipped=0.0
2024-09-23 03:03:35,092 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=6.52 vs. limit=15.0
2024-09-23 03:03:41,271 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.22 vs. limit=15.0
2024-09-23 03:03:55,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=163137.33333333334, ans=0.125
2024-09-23 03:04:07,105 INFO [train.py:1198] (1/4) Epoch 9, batch 3800, loss[loss=0.2456, ctc_loss=0.1699, cr_loss=0.3781, over 16924.00 frames. ], tot_loss[loss=0.258, ctc_loss=0.1808, cr_loss=0.3861, over 3316718.16 frames. ], batch size: 42, lr: 1.30e-02, grad_scale: 32.0
2024-09-23 03:04:08,029 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.67 vs. limit=15.0
2024-09-23 03:04:36,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=163277.33333333334, ans=0.0
2024-09-23 03:05:10,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=163370.66666666666, ans=0.0
2024-09-23 03:05:18,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=163370.66666666666, ans=0.125
2024-09-23 03:05:19,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=163370.66666666666, ans=0.125
2024-09-23 03:05:25,696 INFO [train.py:1198] (1/4) Epoch 9, batch 3850, loss[loss=0.3257, ctc_loss=0.2415, cr_loss=0.4214, over 11909.00 frames. ], tot_loss[loss=0.2603, ctc_loss=0.1828, cr_loss=0.3878, over 3283152.37 frames. ], batch size: 123, lr: 1.30e-02, grad_scale: 32.0
2024-09-23 03:05:27,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_abs, batch_count=163417.33333333334, ans=0.5
2024-09-23 03:05:27,933 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=13.01 vs. limit=22.5
2024-09-23 03:05:42,523 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.391e+02 1.511e+02 1.701e+02 2.274e+02, threshold=3.022e+02, percent-clipped=0.0
2024-09-23 03:05:55,146 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 03:05:59,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=163510.66666666666, ans=0.2
2024-09-23 03:06:06,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=163510.66666666666, ans=0.125
2024-09-23 03:06:14,303 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=9.81 vs. limit=15.0
2024-09-23 03:06:20,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=163557.33333333334, ans=0.125
2024-09-23 03:07:26,834 INFO [train.py:1198] (1/4) Epoch 10, batch 0, loss[loss=0.2591, ctc_loss=0.1825, cr_loss=0.3831, over 17296.00 frames. ], tot_loss[loss=0.2591, ctc_loss=0.1825, cr_loss=0.3831, over 17296.00 frames. ], batch size: 46, lr: 1.24e-02, grad_scale: 32.0
2024-09-23 03:07:26,835 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 03:07:41,776 INFO [train.py:1230] (1/4) Epoch 10, validation: loss=0.05143, ctc_loss=0.05143, cr_loss=7.705e-15, over 944034.00 frames. 
2024-09-23 03:07:41,776 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 03:07:56,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=163678.66666666666, ans=0.0
2024-09-23 03:08:04,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=163678.66666666666, ans=0.125
2024-09-23 03:08:18,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=163725.33333333334, ans=0.125
2024-09-23 03:08:35,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=163772.0, ans=0.035
2024-09-23 03:08:36,673 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.84 vs. limit=15.0
2024-09-23 03:09:04,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=163865.33333333334, ans=0.0
2024-09-23 03:09:05,466 INFO [train.py:1198] (1/4) Epoch 10, batch 50, loss[loss=0.2407, ctc_loss=0.1633, cr_loss=0.3869, over 17085.00 frames. ], tot_loss[loss=0.2568, ctc_loss=0.1798, cr_loss=0.3851, over 748163.33 frames. ], batch size: 43, lr: 1.24e-02, grad_scale: 32.0
2024-09-23 03:09:11,092 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.16 vs. limit=22.5
2024-09-23 03:09:15,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=163865.33333333334, ans=0.0
2024-09-23 03:09:29,247 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.322e+02 1.460e+02 1.757e+02 2.503e+02, threshold=2.921e+02, percent-clipped=0.0
2024-09-23 03:09:51,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=164005.33333333334, ans=0.125
2024-09-23 03:10:24,754 INFO [train.py:1198] (1/4) Epoch 10, batch 100, loss[loss=0.2346, ctc_loss=0.1659, cr_loss=0.3437, over 17225.00 frames. ], tot_loss[loss=0.2588, ctc_loss=0.1814, cr_loss=0.3873, over 1323815.66 frames. ], batch size: 50, lr: 1.24e-02, grad_scale: 32.0
2024-09-23 03:10:24,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=164098.66666666666, ans=0.0
2024-09-23 03:10:27,458 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.47 vs. limit=15.0
2024-09-23 03:10:49,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=164145.33333333334, ans=0.0
2024-09-23 03:11:16,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=164238.66666666666, ans=0.125
2024-09-23 03:11:18,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=164238.66666666666, ans=0.0
2024-09-23 03:11:35,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=164285.33333333334, ans=0.0
2024-09-23 03:11:46,742 INFO [train.py:1198] (1/4) Epoch 10, batch 150, loss[loss=0.2325, ctc_loss=0.1602, cr_loss=0.3615, over 17097.00 frames. ], tot_loss[loss=0.257, ctc_loss=0.1799, cr_loss=0.3856, over 1757001.35 frames. ], batch size: 40, lr: 1.24e-02, grad_scale: 32.0
2024-09-23 03:11:48,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=164332.0, ans=0.125
2024-09-23 03:12:13,436 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.322e+02 1.418e+02 1.649e+02 2.765e+02, threshold=2.835e+02, percent-clipped=0.0
2024-09-23 03:12:13,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=164378.66666666666, ans=0.0
2024-09-23 03:12:13,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=164378.66666666666, ans=0.125
2024-09-23 03:12:26,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=164425.33333333334, ans=0.125
2024-09-23 03:12:27,035 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=9.30 vs. limit=15.0
2024-09-23 03:12:32,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=164425.33333333334, ans=0.2
2024-09-23 03:12:41,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.25 vs. limit=12.0
2024-09-23 03:12:47,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=164472.0, ans=0.2
2024-09-23 03:12:58,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=164518.66666666666, ans=0.035
2024-09-23 03:13:09,148 INFO [train.py:1198] (1/4) Epoch 10, batch 200, loss[loss=0.264, ctc_loss=0.1872, cr_loss=0.3838, over 17302.00 frames. ], tot_loss[loss=0.2564, ctc_loss=0.1792, cr_loss=0.3865, over 2102821.60 frames. ], batch size: 49, lr: 1.24e-02, grad_scale: 32.0
2024-09-23 03:13:12,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=164565.33333333334, ans=0.0
2024-09-23 03:13:20,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=164565.33333333334, ans=0.2
2024-09-23 03:13:46,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=164658.66666666666, ans=0.125
2024-09-23 03:13:46,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=164658.66666666666, ans=0.125
2024-09-23 03:13:51,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=164658.66666666666, ans=0.025
2024-09-23 03:14:04,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=164705.33333333334, ans=0.025
2024-09-23 03:14:09,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1.whitening_limit, batch_count=164705.33333333334, ans=10.0
2024-09-23 03:14:29,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=164752.0, ans=0.125
2024-09-23 03:14:33,529 INFO [train.py:1198] (1/4) Epoch 10, batch 250, loss[loss=0.2437, ctc_loss=0.1699, cr_loss=0.369, over 17154.00 frames. ], tot_loss[loss=0.2567, ctc_loss=0.1795, cr_loss=0.3863, over 2378175.57 frames. ], batch size: 48, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:14:46,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=164798.66666666666, ans=0.025
2024-09-23 03:14:54,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=164845.33333333334, ans=0.125
2024-09-23 03:14:57,018 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.299e+02 1.403e+02 1.575e+02 2.434e+02, threshold=2.805e+02, percent-clipped=0.0
2024-09-23 03:15:02,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=164845.33333333334, ans=0.2
2024-09-23 03:15:10,357 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.69 vs. limit=15.0
2024-09-23 03:15:11,604 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=2.502e-03
2024-09-23 03:15:26,053 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.80 vs. limit=15.0
2024-09-23 03:15:54,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=165032.0, ans=0.125
2024-09-23 03:15:55,550 INFO [train.py:1198] (1/4) Epoch 10, batch 300, loss[loss=0.2515, ctc_loss=0.1746, cr_loss=0.3845, over 17062.00 frames. ], tot_loss[loss=0.2547, ctc_loss=0.1778, cr_loss=0.3841, over 2599064.85 frames. ], batch size: 46, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:16:57,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=165218.66666666666, ans=0.125
2024-09-23 03:17:03,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten.whitening_limit, batch_count=165218.66666666666, ans=15.0
2024-09-23 03:17:17,894 INFO [train.py:1198] (1/4) Epoch 10, batch 350, loss[loss=0.2763, ctc_loss=0.1965, cr_loss=0.3991, over 15930.00 frames. ], tot_loss[loss=0.2551, ctc_loss=0.1781, cr_loss=0.385, over 2765263.96 frames. ], batch size: 74, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:17:21,589 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 03:17:42,143 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.372e+02 1.514e+02 1.677e+02 2.269e+02, threshold=3.028e+02, percent-clipped=0.0
2024-09-23 03:17:50,306 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 03:17:59,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=165358.66666666666, ans=0.0
2024-09-23 03:18:24,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.57 vs. limit=15.0
2024-09-23 03:18:36,083 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.90 vs. limit=15.0
2024-09-23 03:18:43,353 INFO [train.py:1198] (1/4) Epoch 10, batch 400, loss[loss=0.2599, ctc_loss=0.1817, cr_loss=0.3909, over 17009.00 frames. ], tot_loss[loss=0.2561, ctc_loss=0.1788, cr_loss=0.3867, over 2894260.52 frames. ], batch size: 44, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:18:43,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=165498.66666666666, ans=0.125
2024-09-23 03:18:45,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=165498.66666666666, ans=0.125
2024-09-23 03:18:56,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=165498.66666666666, ans=0.2
2024-09-23 03:19:26,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=165592.0, ans=0.0
2024-09-23 03:20:02,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=165732.0, ans=0.1
2024-09-23 03:20:03,260 INFO [train.py:1198] (1/4) Epoch 10, batch 450, loss[loss=0.2635, ctc_loss=0.1797, cr_loss=0.4187, over 17066.00 frames. ], tot_loss[loss=0.2561, ctc_loss=0.1788, cr_loss=0.3865, over 2988644.65 frames. ], batch size: 46, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:20:03,621 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=165732.0, ans=0.0
2024-09-23 03:20:16,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=165732.0, ans=0.0
2024-09-23 03:20:16,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=165732.0, ans=0.1
2024-09-23 03:20:17,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=165778.66666666666, ans=0.0
2024-09-23 03:20:20,473 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.48 vs. limit=22.5
2024-09-23 03:20:25,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=165778.66666666666, ans=0.125
2024-09-23 03:20:26,967 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.328e+02 1.495e+02 1.704e+02 3.618e+02, threshold=2.990e+02, percent-clipped=1.0
2024-09-23 03:20:32,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=165778.66666666666, ans=0.125
2024-09-23 03:20:39,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=165825.33333333334, ans=0.1
2024-09-23 03:20:49,214 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 03:21:14,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=165918.66666666666, ans=0.5
2024-09-23 03:21:25,284 INFO [train.py:1198] (1/4) Epoch 10, batch 500, loss[loss=0.2688, ctc_loss=0.1883, cr_loss=0.4027, over 17231.00 frames. ], tot_loss[loss=0.2548, ctc_loss=0.1778, cr_loss=0.3851, over 3071089.95 frames. ], batch size: 55, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:22:09,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=166058.66666666666, ans=0.125
2024-09-23 03:22:09,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=166058.66666666666, ans=0.0
2024-09-23 03:22:27,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=166105.33333333334, ans=0.2
2024-09-23 03:22:38,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=166152.0, ans=0.125
2024-09-23 03:22:40,701 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=7.15 vs. limit=12.0
2024-09-23 03:22:47,886 INFO [train.py:1198] (1/4) Epoch 10, batch 550, loss[loss=0.2716, ctc_loss=0.191, cr_loss=0.403, over 17103.00 frames. ], tot_loss[loss=0.2543, ctc_loss=0.1774, cr_loss=0.3847, over 3137657.81 frames. ], batch size: 49, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:22:53,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=166198.66666666666, ans=0.125
2024-09-23 03:23:04,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=166245.33333333334, ans=0.125
2024-09-23 03:23:07,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=166245.33333333334, ans=0.125
2024-09-23 03:23:11,787 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.308e+02 1.376e+02 1.532e+02 2.311e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-23 03:23:16,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=166245.33333333334, ans=0.0
2024-09-23 03:24:11,334 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.33 vs. limit=6.0
2024-09-23 03:24:13,174 INFO [train.py:1198] (1/4) Epoch 10, batch 600, loss[loss=0.2288, ctc_loss=0.1602, cr_loss=0.3427, over 17232.00 frames. ], tot_loss[loss=0.2547, ctc_loss=0.1777, cr_loss=0.385, over 3183705.17 frames. ], batch size: 50, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:24:27,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=166478.66666666666, ans=0.2
2024-09-23 03:24:39,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=166478.66666666666, ans=0.025
2024-09-23 03:24:42,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=166478.66666666666, ans=0.0
2024-09-23 03:24:43,172 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.86 vs. limit=5.0
2024-09-23 03:24:48,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=166525.33333333334, ans=0.2
2024-09-23 03:25:23,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=166618.66666666666, ans=0.0
2024-09-23 03:25:32,729 INFO [train.py:1198] (1/4) Epoch 10, batch 650, loss[loss=0.2325, ctc_loss=0.1589, cr_loss=0.368, over 17118.00 frames. ], tot_loss[loss=0.2537, ctc_loss=0.1768, cr_loss=0.3847, over 3229431.87 frames. ], batch size: 40, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:25:44,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=6.58 vs. limit=15.0
2024-09-23 03:25:59,417 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.346e+02 1.493e+02 1.797e+02 2.927e+02, threshold=2.987e+02, percent-clipped=1.0
2024-09-23 03:26:20,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=166758.66666666666, ans=0.125
2024-09-23 03:26:31,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=166805.33333333334, ans=0.1
2024-09-23 03:26:46,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=166852.0, ans=0.125
2024-09-23 03:26:55,505 INFO [train.py:1198] (1/4) Epoch 10, batch 700, loss[loss=0.2175, ctc_loss=0.1543, cr_loss=0.3159, over 17078.00 frames. ], tot_loss[loss=0.2527, ctc_loss=0.1759, cr_loss=0.3838, over 3266129.65 frames. ], batch size: 43, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:26:59,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.min_positive, batch_count=166898.66666666666, ans=0.025
2024-09-23 03:26:59,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=166898.66666666666, ans=0.125
2024-09-23 03:27:05,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=166898.66666666666, ans=0.125
2024-09-23 03:27:22,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=166945.33333333334, ans=0.125
2024-09-23 03:27:52,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=167038.66666666666, ans=0.0
2024-09-23 03:28:07,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer_na.min_abs, batch_count=167085.33333333334, ans=0.02
2024-09-23 03:28:20,933 INFO [train.py:1198] (1/4) Epoch 10, batch 750, loss[loss=0.2392, ctc_loss=0.1683, cr_loss=0.3543, over 17248.00 frames. ], tot_loss[loss=0.2518, ctc_loss=0.1752, cr_loss=0.3831, over 3291279.80 frames. ], batch size: 44, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:28:47,711 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.321e+02 1.498e+02 1.811e+02 2.765e+02, threshold=2.996e+02, percent-clipped=0.0
2024-09-23 03:28:57,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=167225.33333333334, ans=0.125
2024-09-23 03:28:57,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=167225.33333333334, ans=0.125
2024-09-23 03:29:22,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=167272.0, ans=0.125
2024-09-23 03:29:43,237 INFO [train.py:1198] (1/4) Epoch 10, batch 800, loss[loss=0.2798, ctc_loss=0.1966, cr_loss=0.4163, over 17051.00 frames. ], tot_loss[loss=0.2531, ctc_loss=0.1763, cr_loss=0.384, over 3300415.61 frames. ], batch size: 52, lr: 1.23e-02, grad_scale: 32.0
2024-09-23 03:29:56,973 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.86 vs. limit=10.0
2024-09-23 03:30:26,511 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 03:30:43,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=167505.33333333334, ans=0.0
2024-09-23 03:30:52,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=167552.0, ans=0.125
2024-09-23 03:31:03,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=167598.66666666666, ans=0.2
2024-09-23 03:31:05,258 INFO [train.py:1198] (1/4) Epoch 10, batch 850, loss[loss=0.2876, ctc_loss=0.2048, cr_loss=0.4141, over 16013.00 frames. ], tot_loss[loss=0.2527, ctc_loss=0.176, cr_loss=0.3836, over 3319230.27 frames. ], batch size: 74, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:31:12,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=167598.66666666666, ans=0.2
2024-09-23 03:31:29,265 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.338e+02 1.474e+02 1.669e+02 2.399e+02, threshold=2.948e+02, percent-clipped=0.0
2024-09-23 03:31:40,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=167692.0, ans=0.0
2024-09-23 03:31:58,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=167738.66666666666, ans=0.95
2024-09-23 03:32:21,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=167785.33333333334, ans=0.0
2024-09-23 03:32:27,740 INFO [train.py:1198] (1/4) Epoch 10, batch 900, loss[loss=0.2427, ctc_loss=0.1681, cr_loss=0.3729, over 17091.00 frames. ], tot_loss[loss=0.2515, ctc_loss=0.175, cr_loss=0.3824, over 3341924.45 frames. ], batch size: 43, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:32:27,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=167832.0, ans=0.1
2024-09-23 03:33:05,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.08 vs. limit=15.0
2024-09-23 03:33:18,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=167972.0, ans=0.1
2024-09-23 03:33:55,325 INFO [train.py:1198] (1/4) Epoch 10, batch 950, loss[loss=0.2585, ctc_loss=0.1802, cr_loss=0.392, over 17105.00 frames. ], tot_loss[loss=0.2507, ctc_loss=0.1743, cr_loss=0.382, over 3357843.40 frames. ], batch size: 49, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:33:57,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=168065.33333333334, ans=0.125
2024-09-23 03:34:03,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=168065.33333333334, ans=0.0
2024-09-23 03:34:11,464 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 03:34:18,979 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.276e+02 1.450e+02 1.704e+02 3.070e+02, threshold=2.900e+02, percent-clipped=1.0
2024-09-23 03:34:43,695 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.66 vs. limit=15.0
2024-09-23 03:34:59,644 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.28 vs. limit=22.5
2024-09-23 03:35:00,813 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=8.19 vs. limit=15.0
2024-09-23 03:35:02,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=168252.0, ans=0.2
2024-09-23 03:35:02,992 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.09 vs. limit=12.0
2024-09-23 03:35:14,461 INFO [train.py:1198] (1/4) Epoch 10, batch 1000, loss[loss=0.2781, ctc_loss=0.193, cr_loss=0.4255, over 17039.00 frames. ], tot_loss[loss=0.2506, ctc_loss=0.1743, cr_loss=0.3819, over 3362049.23 frames. ], batch size: 52, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:35:46,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=168345.33333333334, ans=0.1
2024-09-23 03:35:47,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=168392.0, ans=0.025
2024-09-23 03:36:16,568 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.33 vs. limit=15.0
2024-09-23 03:36:33,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=168485.33333333334, ans=0.0
2024-09-23 03:36:36,364 INFO [train.py:1198] (1/4) Epoch 10, batch 1050, loss[loss=0.2011, ctc_loss=0.1364, cr_loss=0.3236, over 17290.00 frames. ], tot_loss[loss=0.2517, ctc_loss=0.175, cr_loss=0.3835, over 3365590.28 frames. ], batch size: 42, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:36:56,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=168578.66666666666, ans=0.125
2024-09-23 03:36:57,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=168578.66666666666, ans=0.0
2024-09-23 03:36:58,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.08 vs. limit=10.0
2024-09-23 03:37:00,706 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.301e+02 1.448e+02 1.658e+02 2.854e+02, threshold=2.897e+02, percent-clipped=0.0
2024-09-23 03:37:16,311 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.16 vs. limit=15.0
2024-09-23 03:37:27,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=168672.0, ans=0.0
2024-09-23 03:37:51,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=168718.66666666666, ans=0.125
2024-09-23 03:37:58,857 INFO [train.py:1198] (1/4) Epoch 10, batch 1100, loss[loss=0.2468, ctc_loss=0.168, cr_loss=0.3937, over 17011.00 frames. ], tot_loss[loss=0.2518, ctc_loss=0.1751, cr_loss=0.3836, over 3363595.67 frames. ], batch size: 52, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:38:01,380 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.48 vs. limit=12.0
2024-09-23 03:38:25,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=168812.0, ans=0.0
2024-09-23 03:38:50,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=168905.33333333334, ans=0.0
2024-09-23 03:39:23,800 INFO [train.py:1198] (1/4) Epoch 10, batch 1150, loss[loss=0.2373, ctc_loss=0.1651, cr_loss=0.3611, over 17027.00 frames. ], tot_loss[loss=0.2532, ctc_loss=0.1763, cr_loss=0.3845, over 3347732.92 frames. ], batch size: 44, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:39:47,599 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.373e+02 1.514e+02 1.720e+02 2.403e+02, threshold=3.028e+02, percent-clipped=0.0
2024-09-23 03:40:21,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=169138.66666666666, ans=0.125
2024-09-23 03:40:37,068 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.56 vs. limit=15.0
2024-09-23 03:40:41,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=169185.33333333334, ans=0.2
2024-09-23 03:40:45,947 INFO [train.py:1198] (1/4) Epoch 10, batch 1200, loss[loss=0.2202, ctc_loss=0.1494, cr_loss=0.3541, over 15883.00 frames. ], tot_loss[loss=0.2534, ctc_loss=0.1765, cr_loss=0.3847, over 3341630.04 frames. ], batch size: 35, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:40:52,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=169232.0, ans=0.125
2024-09-23 03:40:57,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=169232.0, ans=0.1
2024-09-23 03:41:27,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=169325.33333333334, ans=0.125
2024-09-23 03:41:40,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=169372.0, ans=0.125
2024-09-23 03:42:07,870 INFO [train.py:1198] (1/4) Epoch 10, batch 1250, loss[loss=0.2885, ctc_loss=0.2049, cr_loss=0.4181, over 17009.00 frames. ], tot_loss[loss=0.2518, ctc_loss=0.1751, cr_loss=0.3833, over 3358816.48 frames. ], batch size: 52, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:42:14,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=169465.33333333334, ans=0.0
2024-09-23 03:42:27,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=169512.0, ans=0.0
2024-09-23 03:42:30,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=169512.0, ans=0.025
2024-09-23 03:42:31,957 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.271e+02 1.375e+02 1.546e+02 2.384e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-23 03:42:37,440 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.43 vs. limit=15.0
2024-09-23 03:42:40,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=169558.66666666666, ans=0.0
2024-09-23 03:42:51,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=169558.66666666666, ans=0.09899494936611666
2024-09-23 03:42:52,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=169558.66666666666, ans=0.2
2024-09-23 03:43:02,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=169605.33333333334, ans=0.025
2024-09-23 03:43:12,329 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.45 vs. limit=10.0
2024-09-23 03:43:25,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=169652.0, ans=10.0
2024-09-23 03:43:32,882 INFO [train.py:1198] (1/4) Epoch 10, batch 1300, loss[loss=0.2563, ctc_loss=0.1809, cr_loss=0.3768, over 17089.00 frames. ], tot_loss[loss=0.252, ctc_loss=0.1753, cr_loss=0.3833, over 3356524.80 frames. ], batch size: 49, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:43:39,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=169698.66666666666, ans=0.0
2024-09-23 03:43:42,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=169698.66666666666, ans=0.2
2024-09-23 03:43:44,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=169698.66666666666, ans=0.0
2024-09-23 03:43:50,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=169745.33333333334, ans=0.125
2024-09-23 03:44:07,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.04 vs. limit=15.0
2024-09-23 03:44:43,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=169885.33333333334, ans=0.125
2024-09-23 03:44:44,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=169885.33333333334, ans=0.2
2024-09-23 03:44:44,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=169885.33333333334, ans=0.1
2024-09-23 03:44:52,271 INFO [train.py:1198] (1/4) Epoch 10, batch 1350, loss[loss=0.2692, ctc_loss=0.1876, cr_loss=0.4084, over 16995.00 frames. ], tot_loss[loss=0.2529, ctc_loss=0.176, cr_loss=0.3845, over 3358890.38 frames. ], batch size: 53, lr: 1.22e-02, grad_scale: 32.0
2024-09-23 03:44:54,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=169932.0, ans=0.125
2024-09-23 03:44:57,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=169932.0, ans=0.125
2024-09-23 03:45:06,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=169978.66666666666, ans=0.1
2024-09-23 03:45:16,272 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.346e+02 1.479e+02 1.695e+02 2.554e+02, threshold=2.958e+02, percent-clipped=0.0
2024-09-23 03:45:33,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=170025.33333333334, ans=0.2
2024-09-23 03:45:48,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=170072.0, ans=10.0
2024-09-23 03:45:56,620 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.31 vs. limit=15.0
2024-09-23 03:46:14,376 INFO [train.py:1198] (1/4) Epoch 10, batch 1400, loss[loss=0.2385, ctc_loss=0.1644, cr_loss=0.3707, over 17298.00 frames. ], tot_loss[loss=0.2531, ctc_loss=0.1763, cr_loss=0.3838, over 3355423.22 frames. ], batch size: 46, lr: 1.22e-02, grad_scale: 16.0
2024-09-23 03:46:20,344 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.89 vs. limit=22.5
2024-09-23 03:46:51,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=170258.66666666666, ans=0.0
2024-09-23 03:46:56,408 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.40 vs. limit=22.5
2024-09-23 03:47:06,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=170305.33333333334, ans=0.1
2024-09-23 03:47:36,189 INFO [train.py:1198] (1/4) Epoch 10, batch 1450, loss[loss=0.1949, ctc_loss=0.1311, cr_loss=0.3191, over 17137.00 frames. ], tot_loss[loss=0.254, ctc_loss=0.1769, cr_loss=0.3853, over 3358781.11 frames. ], batch size: 40, lr: 1.22e-02, grad_scale: 16.0
2024-09-23 03:47:58,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=170445.33333333334, ans=0.025
2024-09-23 03:48:04,178 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.358e+02 1.487e+02 1.729e+02 2.482e+02, threshold=2.974e+02, percent-clipped=0.0
2024-09-23 03:48:14,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=170492.0, ans=0.0
2024-09-23 03:49:00,803 INFO [train.py:1198] (1/4) Epoch 10, batch 1500, loss[loss=0.2521, ctc_loss=0.1773, cr_loss=0.3741, over 17226.00 frames. ], tot_loss[loss=0.2533, ctc_loss=0.1764, cr_loss=0.3844, over 3354320.60 frames. ], batch size: 50, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 03:49:07,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=170632.0, ans=0.125
2024-09-23 03:49:13,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=170632.0, ans=0.125
2024-09-23 03:49:43,176 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.33 vs. limit=6.0
2024-09-23 03:50:22,863 INFO [train.py:1198] (1/4) Epoch 10, batch 1550, loss[loss=0.2093, ctc_loss=0.1395, cr_loss=0.3491, over 17129.00 frames. ], tot_loss[loss=0.2532, ctc_loss=0.1764, cr_loss=0.3839, over 3353420.22 frames. ], batch size: 40, lr: 1.21e-02, grad_scale: 8.0
2024-09-23 03:50:24,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=170865.33333333334, ans=0.95
2024-09-23 03:50:37,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=170912.0, ans=0.2
2024-09-23 03:50:49,522 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.376e+02 1.480e+02 1.669e+02 2.824e+02, threshold=2.960e+02, percent-clipped=0.0
2024-09-23 03:51:00,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=170958.66666666666, ans=0.1
2024-09-23 03:51:18,679 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.72 vs. limit=22.5
2024-09-23 03:51:19,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=171005.33333333334, ans=0.1
2024-09-23 03:51:37,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=171052.0, ans=0.125
2024-09-23 03:51:39,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=171052.0, ans=0.1
2024-09-23 03:51:41,965 INFO [train.py:1198] (1/4) Epoch 10, batch 1600, loss[loss=0.2542, ctc_loss=0.1804, cr_loss=0.369, over 17311.00 frames. ], tot_loss[loss=0.2525, ctc_loss=0.1758, cr_loss=0.3837, over 3360535.57 frames. ], batch size: 51, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 03:52:09,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.whiten.whitening_limit, batch_count=171145.33333333334, ans=12.0
2024-09-23 03:52:11,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=171145.33333333334, ans=0.1
2024-09-23 03:52:29,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=171192.0, ans=0.125
2024-09-23 03:53:03,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=171285.33333333334, ans=0.025
2024-09-23 03:53:05,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.06 vs. limit=22.5
2024-09-23 03:53:06,734 INFO [train.py:1198] (1/4) Epoch 10, batch 1650, loss[loss=0.2851, ctc_loss=0.2008, cr_loss=0.4214, over 16713.00 frames. ], tot_loss[loss=0.2526, ctc_loss=0.1758, cr_loss=0.3842, over 3354788.06 frames. ], batch size: 61, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 03:53:08,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=171332.0, ans=0.125
2024-09-23 03:53:24,489 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.44 vs. limit=15.0
2024-09-23 03:53:36,437 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.322e+02 1.432e+02 1.629e+02 2.855e+02, threshold=2.864e+02, percent-clipped=0.0
2024-09-23 03:53:45,301 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.84 vs. limit=15.0
2024-09-23 03:53:46,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=171425.33333333334, ans=0.025
2024-09-23 03:53:56,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=171472.0, ans=0.2
2024-09-23 03:54:07,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=171472.0, ans=0.0
2024-09-23 03:54:29,206 INFO [train.py:1198] (1/4) Epoch 10, batch 1700, loss[loss=0.3011, ctc_loss=0.2146, cr_loss=0.4327, over 17049.00 frames. ], tot_loss[loss=0.2528, ctc_loss=0.1759, cr_loss=0.3848, over 3355523.00 frames. ], batch size: 56, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 03:54:37,965 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.51 vs. limit=22.5
2024-09-23 03:54:45,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=171612.0, ans=0.1
2024-09-23 03:55:26,323 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.55 vs. limit=15.0
2024-09-23 03:55:50,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=171798.66666666666, ans=0.125
2024-09-23 03:55:51,290 INFO [train.py:1198] (1/4) Epoch 10, batch 1750, loss[loss=0.3253, ctc_loss=0.2447, cr_loss=0.403, over 12218.00 frames. ], tot_loss[loss=0.2528, ctc_loss=0.1759, cr_loss=0.3847, over 3347605.16 frames. ], batch size: 123, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 03:56:04,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=171798.66666666666, ans=0.2
2024-09-23 03:56:15,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=171845.33333333334, ans=0.1
2024-09-23 03:56:18,403 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.367e+02 1.590e+02 1.877e+02 2.998e+02, threshold=3.180e+02, percent-clipped=1.0
2024-09-23 03:56:53,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=171938.66666666666, ans=0.2
2024-09-23 03:57:07,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=171985.33333333334, ans=0.0
2024-09-23 03:57:07,923 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=15.45 vs. limit=15.0
2024-09-23 03:57:13,328 INFO [train.py:1198] (1/4) Epoch 10, batch 1800, loss[loss=0.3309, ctc_loss=0.2456, cr_loss=0.4266, over 12231.00 frames. ], tot_loss[loss=0.254, ctc_loss=0.177, cr_loss=0.3849, over 3327066.92 frames. ], batch size: 123, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 03:57:18,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=172032.0, ans=0.0
2024-09-23 03:57:31,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=172078.66666666666, ans=0.2
2024-09-23 03:57:49,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=172125.33333333334, ans=0.125
2024-09-23 03:57:50,031 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=21.72 vs. limit=22.5
2024-09-23 03:58:01,229 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.66 vs. limit=15.0
2024-09-23 03:58:29,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=172218.66666666666, ans=0.07
2024-09-23 03:58:37,573 INFO [train.py:1198] (1/4) Epoch 10, batch 1850, loss[loss=0.2764, ctc_loss=0.1917, cr_loss=0.4235, over 17141.00 frames. ], tot_loss[loss=0.2528, ctc_loss=0.1758, cr_loss=0.3848, over 3341388.25 frames. ], batch size: 48, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 03:58:51,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=172312.0, ans=0.0
2024-09-23 03:58:55,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=172312.0, ans=0.0
2024-09-23 03:59:04,289 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.347e+02 1.457e+02 1.655e+02 2.749e+02, threshold=2.915e+02, percent-clipped=0.0
2024-09-23 03:59:04,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=172312.0, ans=0.5
2024-09-23 03:59:40,617 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.79 vs. limit=12.0
2024-09-23 03:59:51,473 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=19.75 vs. limit=22.5
2024-09-23 03:59:57,222 INFO [train.py:1198] (1/4) Epoch 10, batch 1900, loss[loss=0.2698, ctc_loss=0.1914, cr_loss=0.3924, over 17311.00 frames. ], tot_loss[loss=0.2517, ctc_loss=0.175, cr_loss=0.3831, over 3344860.47 frames. ], batch size: 46, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 04:00:02,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.12 vs. limit=10.0
2024-09-23 04:01:12,527 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.41 vs. limit=15.0
2024-09-23 04:01:19,644 INFO [train.py:1198] (1/4) Epoch 10, batch 1950, loss[loss=0.2622, ctc_loss=0.1838, cr_loss=0.3918, over 17203.00 frames. ], tot_loss[loss=0.2534, ctc_loss=0.1766, cr_loss=0.3843, over 3337625.42 frames. ], batch size: 55, lr: 1.21e-02, grad_scale: 16.0
2024-09-23 04:01:35,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=172778.66666666666, ans=0.04949747468305833
2024-09-23 04:01:48,821 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.405e+02 1.574e+02 1.755e+02 2.409e+02, threshold=3.148e+02, percent-clipped=0.0
2024-09-23 04:01:52,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=172825.33333333334, ans=0.125
2024-09-23 04:02:13,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.09 vs. limit=22.5
2024-09-23 04:02:21,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=172872.0, ans=0.1
2024-09-23 04:02:30,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=172918.66666666666, ans=0.0
2024-09-23 04:02:43,973 INFO [train.py:1198] (1/4) Epoch 10, batch 2000, loss[loss=0.2561, ctc_loss=0.1784, cr_loss=0.3886, over 17047.00 frames. ], tot_loss[loss=0.2529, ctc_loss=0.1761, cr_loss=0.3841, over 3339978.09 frames. ], batch size: 52, lr: 1.21e-02, grad_scale: 32.0
2024-09-23 04:02:52,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=172965.33333333334, ans=0.2
2024-09-23 04:03:18,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=173058.66666666666, ans=0.125
2024-09-23 04:03:34,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=173105.33333333334, ans=0.025
2024-09-23 04:04:00,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=173152.0, ans=0.07
2024-09-23 04:04:06,412 INFO [train.py:1198] (1/4) Epoch 10, batch 2050, loss[loss=0.2337, ctc_loss=0.1616, cr_loss=0.3603, over 17158.00 frames. ], tot_loss[loss=0.2516, ctc_loss=0.175, cr_loss=0.3827, over 3352709.32 frames. ], batch size: 45, lr: 1.21e-02, grad_scale: 32.0
2024-09-23 04:04:16,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=173198.66666666666, ans=0.1
2024-09-23 04:04:34,982 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.316e+02 1.437e+02 1.596e+02 3.726e+02, threshold=2.874e+02, percent-clipped=1.0
2024-09-23 04:04:43,718 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=4.19 vs. limit=12.0
2024-09-23 04:04:49,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=173292.0, ans=0.1
2024-09-23 04:04:49,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=173292.0, ans=0.05
2024-09-23 04:05:03,687 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.68 vs. limit=15.0
2024-09-23 04:05:21,226 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.41 vs. limit=15.0
2024-09-23 04:05:22,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=173385.33333333334, ans=0.125
2024-09-23 04:05:28,496 INFO [train.py:1198] (1/4) Epoch 10, batch 2100, loss[loss=0.2487, ctc_loss=0.1711, cr_loss=0.3877, over 17012.00 frames. ], tot_loss[loss=0.2512, ctc_loss=0.1747, cr_loss=0.3827, over 3354418.03 frames. ], batch size: 53, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:05:33,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=173432.0, ans=0.025
2024-09-23 04:05:41,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=173432.0, ans=0.125
2024-09-23 04:05:45,917 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 04:06:06,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=173525.33333333334, ans=0.0
2024-09-23 04:06:39,713 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.31 vs. limit=6.0
2024-09-23 04:06:49,963 INFO [train.py:1198] (1/4) Epoch 10, batch 2150, loss[loss=0.2473, ctc_loss=0.1707, cr_loss=0.3832, over 16701.00 frames. ], tot_loss[loss=0.251, ctc_loss=0.1745, cr_loss=0.3826, over 3363132.81 frames. ], batch size: 61, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:06:53,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=173665.33333333334, ans=0.125
2024-09-23 04:06:58,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=173665.33333333334, ans=0.0
2024-09-23 04:07:05,188 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.96 vs. limit=15.0
2024-09-23 04:07:18,784 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.317e+02 1.458e+02 1.666e+02 3.135e+02, threshold=2.917e+02, percent-clipped=1.0
2024-09-23 04:07:31,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=173758.66666666666, ans=0.125
2024-09-23 04:07:32,216 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.56 vs. limit=15.0
2024-09-23 04:07:42,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=173805.33333333334, ans=0.04949747468305833
2024-09-23 04:07:48,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=173805.33333333334, ans=0.125
2024-09-23 04:08:03,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=173852.0, ans=0.125
2024-09-23 04:08:14,679 INFO [train.py:1198] (1/4) Epoch 10, batch 2200, loss[loss=0.2485, ctc_loss=0.1734, cr_loss=0.3756, over 17294.00 frames. ], tot_loss[loss=0.2511, ctc_loss=0.1746, cr_loss=0.3822, over 3363212.05 frames. ], batch size: 49, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:08:42,624 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.24 vs. limit=12.0
2024-09-23 04:08:43,927 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.22 vs. limit=15.0
2024-09-23 04:08:48,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=173992.0, ans=0.025
2024-09-23 04:08:56,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=173992.0, ans=0.0
2024-09-23 04:09:07,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=174038.66666666666, ans=0.125
2024-09-23 04:09:26,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=174085.33333333334, ans=0.125
2024-09-23 04:09:30,301 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.93 vs. limit=10.0
2024-09-23 04:09:34,448 INFO [train.py:1198] (1/4) Epoch 10, batch 2250, loss[loss=0.2649, ctc_loss=0.187, cr_loss=0.3893, over 17044.00 frames. ], tot_loss[loss=0.2518, ctc_loss=0.1752, cr_loss=0.3829, over 3365080.28 frames. ], batch size: 53, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:09:45,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=174132.0, ans=0.125
2024-09-23 04:09:58,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=174178.66666666666, ans=0.0
2024-09-23 04:10:05,826 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.382e+02 1.475e+02 1.727e+02 2.787e+02, threshold=2.949e+02, percent-clipped=0.0
2024-09-23 04:10:06,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.10 vs. limit=22.5
2024-09-23 04:10:17,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=174225.33333333334, ans=0.125
2024-09-23 04:10:18,122 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=15.59 vs. limit=22.5
2024-09-23 04:10:25,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=174272.0, ans=0.0
2024-09-23 04:10:25,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=174272.0, ans=0.05
2024-09-23 04:10:33,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=174272.0, ans=0.0
2024-09-23 04:10:36,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=174272.0, ans=0.0
2024-09-23 04:10:40,502 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.94 vs. limit=10.0
2024-09-23 04:10:42,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=174318.66666666666, ans=0.125
2024-09-23 04:10:56,693 INFO [train.py:1198] (1/4) Epoch 10, batch 2300, loss[loss=0.261, ctc_loss=0.1818, cr_loss=0.396, over 17288.00 frames. ], tot_loss[loss=0.251, ctc_loss=0.1745, cr_loss=0.3824, over 3365223.85 frames. ], batch size: 51, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:11:22,704 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=17.18 vs. limit=22.5
2024-09-23 04:11:42,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=174458.66666666666, ans=0.125
2024-09-23 04:12:04,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys.whitening_limit, batch_count=174552.0, ans=6.0
2024-09-23 04:12:09,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=174552.0, ans=0.2
2024-09-23 04:12:14,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=174552.0, ans=0.125
2024-09-23 04:12:19,121 INFO [train.py:1198] (1/4) Epoch 10, batch 2350, loss[loss=0.2386, ctc_loss=0.1635, cr_loss=0.3758, over 17299.00 frames. ], tot_loss[loss=0.2506, ctc_loss=0.1742, cr_loss=0.382, over 3365090.13 frames. ], batch size: 46, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:12:41,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=174645.33333333334, ans=0.1
2024-09-23 04:12:41,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.25 vs. limit=15.0
2024-09-23 04:12:42,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=174645.33333333334, ans=0.0
2024-09-23 04:12:50,300 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.308e+02 1.414e+02 1.626e+02 2.428e+02, threshold=2.828e+02, percent-clipped=0.0
2024-09-23 04:13:41,542 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.57 vs. limit=15.0
2024-09-23 04:13:43,456 INFO [train.py:1198] (1/4) Epoch 10, batch 2400, loss[loss=0.2151, ctc_loss=0.1481, cr_loss=0.3351, over 16949.00 frames. ], tot_loss[loss=0.2513, ctc_loss=0.1748, cr_loss=0.3828, over 3367462.91 frames. ], batch size: 42, lr: 1.20e-02, grad_scale: 32.0
2024-09-23 04:13:48,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=174832.0, ans=0.0
2024-09-23 04:13:48,935 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.49 vs. limit=22.5
2024-09-23 04:13:58,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=174878.66666666666, ans=0.125
2024-09-23 04:13:58,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=174878.66666666666, ans=0.125
2024-09-23 04:14:08,179 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=17.70 vs. limit=22.5
2024-09-23 04:14:18,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=174925.33333333334, ans=0.1
2024-09-23 04:14:22,709 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.17 vs. limit=6.0
2024-09-23 04:14:46,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=175018.66666666666, ans=0.125
2024-09-23 04:14:49,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=175018.66666666666, ans=0.125
2024-09-23 04:14:51,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=175018.66666666666, ans=0.125
2024-09-23 04:15:03,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=175018.66666666666, ans=0.125
2024-09-23 04:15:06,054 INFO [train.py:1198] (1/4) Epoch 10, batch 2450, loss[loss=0.2621, ctc_loss=0.1812, cr_loss=0.4044, over 17020.00 frames. ], tot_loss[loss=0.2522, ctc_loss=0.1754, cr_loss=0.3841, over 3368747.42 frames. ], batch size: 52, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:15:22,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=175112.0, ans=0.2
2024-09-23 04:15:36,388 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.295e+02 1.392e+02 1.667e+02 2.800e+02, threshold=2.783e+02, percent-clipped=0.0
2024-09-23 04:15:43,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=175158.66666666666, ans=0.1
2024-09-23 04:15:43,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=175158.66666666666, ans=0.025
2024-09-23 04:16:05,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=175205.33333333334, ans=0.125
2024-09-23 04:16:21,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=175252.0, ans=0.125
2024-09-23 04:16:25,984 INFO [train.py:1198] (1/4) Epoch 10, batch 2500, loss[loss=0.2141, ctc_loss=0.1464, cr_loss=0.3382, over 17169.00 frames. ], tot_loss[loss=0.2527, ctc_loss=0.1758, cr_loss=0.3849, over 3363136.03 frames. ], batch size: 41, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:16:26,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=175298.66666666666, ans=0.0
2024-09-23 04:16:39,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=175298.66666666666, ans=0.125
2024-09-23 04:16:40,789 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.26 vs. limit=22.5
2024-09-23 04:16:45,401 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.77 vs. limit=22.5
2024-09-23 04:17:12,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=175392.0, ans=0.04949747468305833
2024-09-23 04:17:21,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=175438.66666666666, ans=0.09899494936611666
2024-09-23 04:17:50,877 INFO [train.py:1198] (1/4) Epoch 10, batch 2550, loss[loss=0.2689, ctc_loss=0.1854, cr_loss=0.4175, over 15935.00 frames. ], tot_loss[loss=0.2523, ctc_loss=0.1753, cr_loss=0.3846, over 3360303.53 frames. ], batch size: 74, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:18:20,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=175578.66666666666, ans=0.05
2024-09-23 04:18:23,452 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.321e+02 1.441e+02 1.691e+02 2.698e+02, threshold=2.882e+02, percent-clipped=0.0
2024-09-23 04:18:41,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=175672.0, ans=0.035
2024-09-23 04:18:42,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=175672.0, ans=0.2
2024-09-23 04:18:55,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=175718.66666666666, ans=0.95
2024-09-23 04:19:00,667 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=6.72 vs. limit=15.0
2024-09-23 04:19:02,468 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.33 vs. limit=15.0
2024-09-23 04:19:12,531 INFO [train.py:1198] (1/4) Epoch 10, batch 2600, loss[loss=0.2488, ctc_loss=0.1714, cr_loss=0.387, over 17276.00 frames. ], tot_loss[loss=0.2515, ctc_loss=0.1747, cr_loss=0.3838, over 3358626.40 frames. ], batch size: 44, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:19:15,065 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.05 vs. limit=6.0
2024-09-23 04:19:27,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=175812.0, ans=0.0
2024-09-23 04:19:53,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=175858.66666666666, ans=0.125
2024-09-23 04:20:34,898 INFO [train.py:1198] (1/4) Epoch 10, batch 2650, loss[loss=0.2645, ctc_loss=0.1816, cr_loss=0.4148, over 17000.00 frames. ], tot_loss[loss=0.2525, ctc_loss=0.1755, cr_loss=0.3847, over 3355037.81 frames. ], batch size: 51, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:20:44,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=175998.66666666666, ans=0.1
2024-09-23 04:20:52,071 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.44 vs. limit=6.0
2024-09-23 04:20:52,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=176045.33333333334, ans=0.125
2024-09-23 04:20:54,299 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=176045.33333333334, ans=0.125
2024-09-23 04:21:03,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=176045.33333333334, ans=0.125
2024-09-23 04:21:05,299 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.318e+02 1.384e+02 1.550e+02 2.652e+02, threshold=2.768e+02, percent-clipped=0.0
2024-09-23 04:21:07,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=176092.0, ans=0.025
2024-09-23 04:21:09,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.28 vs. limit=15.0
2024-09-23 04:21:11,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=176092.0, ans=0.125
2024-09-23 04:21:17,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.08 vs. limit=10.0
2024-09-23 04:21:36,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=176138.66666666666, ans=0.125
2024-09-23 04:21:57,389 INFO [train.py:1198] (1/4) Epoch 10, batch 2700, loss[loss=0.2364, ctc_loss=0.1593, cr_loss=0.3855, over 17101.00 frames. ], tot_loss[loss=0.2521, ctc_loss=0.1753, cr_loss=0.3842, over 3347726.50 frames. ], batch size: 46, lr: 1.20e-02, grad_scale: 16.0
2024-09-23 04:21:59,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.08 vs. limit=15.0
2024-09-23 04:22:07,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=176232.0, ans=0.0
2024-09-23 04:22:10,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=176232.0, ans=0.125
2024-09-23 04:22:55,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=176372.0, ans=0.1
2024-09-23 04:23:21,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=176465.33333333334, ans=0.125
2024-09-23 04:23:22,732 INFO [train.py:1198] (1/4) Epoch 10, batch 2750, loss[loss=0.2656, ctc_loss=0.1864, cr_loss=0.396, over 17348.00 frames. ], tot_loss[loss=0.2509, ctc_loss=0.1743, cr_loss=0.3831, over 3357550.05 frames. ], batch size: 48, lr: 1.19e-02, grad_scale: 16.0
2024-09-23 04:23:53,132 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.357e+02 1.501e+02 1.751e+02 2.551e+02, threshold=3.001e+02, percent-clipped=0.0
2024-09-23 04:24:17,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=176605.33333333334, ans=0.2
2024-09-23 04:24:27,251 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.91 vs. limit=6.0
2024-09-23 04:24:33,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=176652.0, ans=0.025
2024-09-23 04:24:36,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=176652.0, ans=0.025
2024-09-23 04:24:41,486 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.40 vs. limit=15.0
2024-09-23 04:24:42,409 INFO [train.py:1198] (1/4) Epoch 10, batch 2800, loss[loss=0.2797, ctc_loss=0.1942, cr_loss=0.4275, over 17367.00 frames. ], tot_loss[loss=0.2518, ctc_loss=0.175, cr_loss=0.3838, over 3353051.31 frames. ], batch size: 48, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:24:46,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=176698.66666666666, ans=0.07
2024-09-23 04:25:01,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=176745.33333333334, ans=0.0
2024-09-23 04:25:04,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=176745.33333333334, ans=0.125
2024-09-23 04:25:09,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=176745.33333333334, ans=0.2
2024-09-23 04:25:27,447 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.07 vs. limit=22.5
2024-09-23 04:25:54,400 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.79 vs. limit=15.0
2024-09-23 04:26:05,052 INFO [train.py:1198] (1/4) Epoch 10, batch 2850, loss[loss=0.2338, ctc_loss=0.1613, cr_loss=0.3625, over 17256.00 frames. ], tot_loss[loss=0.2513, ctc_loss=0.1746, cr_loss=0.3834, over 3347374.98 frames. ], batch size: 44, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:26:21,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=176978.66666666666, ans=0.0
2024-09-23 04:26:37,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.79 vs. limit=15.0
2024-09-23 04:26:38,102 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.308e+02 1.482e+02 1.673e+02 2.195e+02, threshold=2.964e+02, percent-clipped=0.0
2024-09-23 04:26:43,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=177025.33333333334, ans=0.125
2024-09-23 04:26:44,000 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.07 vs. limit=15.0
2024-09-23 04:26:54,820 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 04:27:13,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=177118.66666666666, ans=0.125
2024-09-23 04:27:29,636 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.56 vs. limit=10.0
2024-09-23 04:27:30,213 INFO [train.py:1198] (1/4) Epoch 10, batch 2900, loss[loss=0.2698, ctc_loss=0.1884, cr_loss=0.4073, over 17026.00 frames. ], tot_loss[loss=0.2509, ctc_loss=0.1743, cr_loss=0.3834, over 3352313.47 frames. ], batch size: 52, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:27:55,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=177212.0, ans=0.1
2024-09-23 04:28:08,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=177258.66666666666, ans=0.0
2024-09-23 04:28:52,968 INFO [train.py:1198] (1/4) Epoch 10, batch 2950, loss[loss=0.216, ctc_loss=0.1463, cr_loss=0.3485, over 17208.00 frames. ], tot_loss[loss=0.25, ctc_loss=0.1736, cr_loss=0.382, over 3350803.90 frames. ], batch size: 41, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:29:07,932 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.46 vs. limit=22.5
2024-09-23 04:29:09,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=177445.33333333334, ans=0.0
2024-09-23 04:29:23,162 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.333e+02 1.430e+02 1.579e+02 2.314e+02, threshold=2.860e+02, percent-clipped=0.0
2024-09-23 04:29:44,547 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.78 vs. limit=15.0
2024-09-23 04:29:57,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=177585.33333333334, ans=0.125
2024-09-23 04:30:11,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=177585.33333333334, ans=0.0
2024-09-23 04:30:12,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=177632.0, ans=0.0
2024-09-23 04:30:14,432 INFO [train.py:1198] (1/4) Epoch 10, batch 3000, loss[loss=0.2552, ctc_loss=0.1749, cr_loss=0.4017, over 17156.00 frames. ], tot_loss[loss=0.2499, ctc_loss=0.1735, cr_loss=0.3823, over 3356526.02 frames. ], batch size: 48, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:30:14,432 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 04:30:27,210 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([3.4667, 4.0576, 3.8893, 4.2904, 3.8853, 3.9366, 3.8190, 3.6780],
       device='cuda:1')
2024-09-23 04:30:30,502 INFO [train.py:1230] (1/4) Epoch 10, validation: loss=0.04843, ctc_loss=0.04843, cr_loss=7.942e-15, over 944034.00 frames. 
2024-09-23 04:30:30,503 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 04:30:46,828 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.46 vs. limit=15.0
2024-09-23 04:31:01,056 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.39 vs. limit=15.0
2024-09-23 04:31:01,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=177725.33333333334, ans=0.125
2024-09-23 04:31:11,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=177725.33333333334, ans=0.2
2024-09-23 04:31:15,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=177772.0, ans=0.125
2024-09-23 04:31:28,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=177772.0, ans=0.1
2024-09-23 04:31:33,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=177818.66666666666, ans=0.05
2024-09-23 04:31:34,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=177818.66666666666, ans=0.2
2024-09-23 04:31:40,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=177818.66666666666, ans=0.125
2024-09-23 04:31:48,170 INFO [train.py:1198] (1/4) Epoch 10, batch 3050, loss[loss=0.2227, ctc_loss=0.1552, cr_loss=0.3373, over 17032.00 frames. ], tot_loss[loss=0.2505, ctc_loss=0.174, cr_loss=0.3824, over 3358504.65 frames. ], batch size: 39, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:32:01,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=177865.33333333334, ans=0.2
2024-09-23 04:32:14,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=177912.0, ans=0.125
2024-09-23 04:32:17,672 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.279e+02 1.408e+02 1.568e+02 2.746e+02, threshold=2.816e+02, percent-clipped=0.0
2024-09-23 04:32:28,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=177958.66666666666, ans=0.125
2024-09-23 04:32:30,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=177958.66666666666, ans=0.1
2024-09-23 04:32:47,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=178005.33333333334, ans=0.1
2024-09-23 04:33:06,403 INFO [train.py:1198] (1/4) Epoch 10, batch 3100, loss[loss=0.208, ctc_loss=0.1396, cr_loss=0.3417, over 17031.00 frames. ], tot_loss[loss=0.2499, ctc_loss=0.1735, cr_loss=0.3821, over 3363955.54 frames. ], batch size: 39, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:33:08,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=178098.66666666666, ans=0.04949747468305833
2024-09-23 04:33:19,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=178098.66666666666, ans=0.125
2024-09-23 04:33:27,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=178145.33333333334, ans=0.125
2024-09-23 04:33:54,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=178238.66666666666, ans=0.125
2024-09-23 04:33:55,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=178238.66666666666, ans=0.125
2024-09-23 04:34:03,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=178238.66666666666, ans=0.125
2024-09-23 04:34:10,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=178285.33333333334, ans=0.0
2024-09-23 04:34:10,963 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.35 vs. limit=15.0
2024-09-23 04:34:18,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=178285.33333333334, ans=0.125
2024-09-23 04:34:21,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=178285.33333333334, ans=0.125
2024-09-23 04:34:24,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=178285.33333333334, ans=0.1
2024-09-23 04:34:24,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=178285.33333333334, ans=0.1
2024-09-23 04:34:27,442 INFO [train.py:1198] (1/4) Epoch 10, batch 3150, loss[loss=0.2698, ctc_loss=0.1822, cr_loss=0.4383, over 17008.00 frames. ], tot_loss[loss=0.2509, ctc_loss=0.1742, cr_loss=0.3832, over 3352317.46 frames. ], batch size: 53, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:34:50,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=178378.66666666666, ans=0.125
2024-09-23 04:34:56,601 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.289e+02 1.420e+02 1.589e+02 2.247e+02, threshold=2.840e+02, percent-clipped=0.0
2024-09-23 04:35:10,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.68 vs. limit=12.0
2024-09-23 04:35:19,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=178472.0, ans=0.07
2024-09-23 04:35:27,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=178472.0, ans=0.2
2024-09-23 04:35:49,621 INFO [train.py:1198] (1/4) Epoch 10, batch 3200, loss[loss=0.2155, ctc_loss=0.1467, cr_loss=0.3439, over 16992.00 frames. ], tot_loss[loss=0.2508, ctc_loss=0.1743, cr_loss=0.3828, over 3355593.40 frames. ], batch size: 51, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:36:00,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=178565.33333333334, ans=0.1
2024-09-23 04:36:14,934 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.12 vs. limit=12.0
2024-09-23 04:36:24,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=178658.66666666666, ans=0.04949747468305833
2024-09-23 04:36:27,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=178658.66666666666, ans=0.125
2024-09-23 04:36:31,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=178658.66666666666, ans=0.125
2024-09-23 04:36:39,688 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 04:36:41,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.31 vs. limit=15.0
2024-09-23 04:36:56,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=178752.0, ans=0.0
2024-09-23 04:37:07,405 INFO [train.py:1198] (1/4) Epoch 10, batch 3250, loss[loss=0.2274, ctc_loss=0.1543, cr_loss=0.3655, over 17215.00 frames. ], tot_loss[loss=0.251, ctc_loss=0.1743, cr_loss=0.3834, over 3362350.98 frames. ], batch size: 41, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:37:10,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=10.36 vs. limit=15.0
2024-09-23 04:37:15,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=178798.66666666666, ans=0.125
2024-09-23 04:37:37,059 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.295e+02 1.395e+02 1.531e+02 3.942e+02, threshold=2.791e+02, percent-clipped=1.0
2024-09-23 04:37:37,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=178892.0, ans=0.125
2024-09-23 04:38:25,401 INFO [train.py:1198] (1/4) Epoch 10, batch 3300, loss[loss=0.272, ctc_loss=0.1928, cr_loss=0.3961, over 17015.00 frames. ], tot_loss[loss=0.2516, ctc_loss=0.1748, cr_loss=0.3841, over 3359846.55 frames. ], batch size: 56, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:38:28,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=179032.0, ans=0.0
2024-09-23 04:38:59,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=179125.33333333334, ans=0.125
2024-09-23 04:39:26,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=179218.66666666666, ans=0.07
2024-09-23 04:39:31,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=179218.66666666666, ans=0.0
2024-09-23 04:39:34,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=179218.66666666666, ans=0.125
2024-09-23 04:39:34,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=179218.66666666666, ans=0.025
2024-09-23 04:39:43,187 INFO [train.py:1198] (1/4) Epoch 10, batch 3350, loss[loss=0.257, ctc_loss=0.1803, cr_loss=0.3834, over 16040.00 frames. ], tot_loss[loss=0.2517, ctc_loss=0.1749, cr_loss=0.3841, over 3348347.22 frames. ], batch size: 74, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:39:53,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=179265.33333333334, ans=0.125
2024-09-23 04:40:09,708 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=16.03 vs. limit=22.5
2024-09-23 04:40:12,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=179312.0, ans=0.1
2024-09-23 04:40:15,061 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.385e+02 1.553e+02 1.795e+02 2.936e+02, threshold=3.106e+02, percent-clipped=1.0
2024-09-23 04:40:15,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=179358.66666666666, ans=0.0
2024-09-23 04:40:17,590 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.89 vs. limit=12.0
2024-09-23 04:41:03,094 INFO [train.py:1198] (1/4) Epoch 10, batch 3400, loss[loss=0.2558, ctc_loss=0.173, cr_loss=0.4138, over 17348.00 frames. ], tot_loss[loss=0.2516, ctc_loss=0.175, cr_loss=0.3829, over 3336227.90 frames. ], batch size: 48, lr: 1.19e-02, grad_scale: 32.0
2024-09-23 04:41:17,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=179545.33333333334, ans=0.125
2024-09-23 04:41:31,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=179545.33333333334, ans=0.1
2024-09-23 04:41:36,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=179592.0, ans=0.1
2024-09-23 04:41:51,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=179638.66666666666, ans=0.0
2024-09-23 04:41:58,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=179638.66666666666, ans=0.125
2024-09-23 04:42:21,370 INFO [train.py:1198] (1/4) Epoch 10, batch 3450, loss[loss=0.2711, ctc_loss=0.1901, cr_loss=0.4052, over 17256.00 frames. ], tot_loss[loss=0.2508, ctc_loss=0.1745, cr_loss=0.3816, over 3333692.72 frames. ], batch size: 55, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:42:21,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=179732.0, ans=0.1
2024-09-23 04:42:37,524 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.23 vs. limit=10.0
2024-09-23 04:42:41,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=179778.66666666666, ans=0.125
2024-09-23 04:42:46,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=179778.66666666666, ans=0.0
2024-09-23 04:42:50,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.316e+02 1.451e+02 1.743e+02 2.784e+02, threshold=2.902e+02, percent-clipped=0.0
2024-09-23 04:42:55,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=179825.33333333334, ans=0.1
2024-09-23 04:43:14,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=179872.0, ans=0.125
2024-09-23 04:43:23,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=179918.66666666666, ans=0.125
2024-09-23 04:43:25,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=179918.66666666666, ans=0.0
2024-09-23 04:43:26,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=179918.66666666666, ans=0.125
2024-09-23 04:43:32,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=179918.66666666666, ans=0.2
2024-09-23 04:43:37,916 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.94 vs. limit=15.0
2024-09-23 04:43:40,111 INFO [train.py:1198] (1/4) Epoch 10, batch 3500, loss[loss=0.2988, ctc_loss=0.2139, cr_loss=0.4242, over 15741.00 frames. ], tot_loss[loss=0.2503, ctc_loss=0.174, cr_loss=0.3815, over 3339415.44 frames. ], batch size: 74, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:43:44,223 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.51 vs. limit=15.0
2024-09-23 04:43:52,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=179965.33333333334, ans=0.1
2024-09-23 04:44:07,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=180012.0, ans=0.1
2024-09-23 04:44:21,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=180058.66666666666, ans=0.125
2024-09-23 04:44:26,452 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.06 vs. limit=10.0
2024-09-23 04:45:02,137 INFO [train.py:1198] (1/4) Epoch 10, batch 3550, loss[loss=0.2246, ctc_loss=0.1538, cr_loss=0.3537, over 16349.00 frames. ], tot_loss[loss=0.2506, ctc_loss=0.1742, cr_loss=0.3818, over 3341533.57 frames. ], batch size: 36, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:45:23,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=180245.33333333334, ans=0.0
2024-09-23 04:45:26,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=180245.33333333334, ans=0.125
2024-09-23 04:45:32,190 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.310e+02 1.457e+02 1.783e+02 3.691e+02, threshold=2.913e+02, percent-clipped=1.0
2024-09-23 04:46:02,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=180338.66666666666, ans=0.125
2024-09-23 04:46:03,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=180338.66666666666, ans=10.0
2024-09-23 04:46:18,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=180385.33333333334, ans=0.0
2024-09-23 04:46:20,829 INFO [train.py:1198] (1/4) Epoch 10, batch 3600, loss[loss=0.2726, ctc_loss=0.1883, cr_loss=0.4217, over 17301.00 frames. ], tot_loss[loss=0.2493, ctc_loss=0.173, cr_loss=0.3815, over 3356126.99 frames. ], batch size: 46, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:46:22,856 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.58 vs. limit=15.0
2024-09-23 04:46:28,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=180432.0, ans=0.125
2024-09-23 04:46:35,623 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=2.96 vs. limit=15.0
2024-09-23 04:47:21,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=180618.66666666666, ans=0.125
2024-09-23 04:47:28,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=180618.66666666666, ans=0.0
2024-09-23 04:47:30,414 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.98 vs. limit=22.5
2024-09-23 04:47:30,729 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.50 vs. limit=5.0
2024-09-23 04:47:38,885 INFO [train.py:1198] (1/4) Epoch 10, batch 3650, loss[loss=0.2689, ctc_loss=0.1892, cr_loss=0.3985, over 17213.00 frames. ], tot_loss[loss=0.2503, ctc_loss=0.1737, cr_loss=0.3829, over 3355696.77 frames. ], batch size: 55, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:48:03,961 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 04:48:08,242 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.307e+02 1.399e+02 1.525e+02 2.249e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-23 04:48:15,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=180758.66666666666, ans=0.1
2024-09-23 04:48:18,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=180758.66666666666, ans=0.2
2024-09-23 04:48:32,247 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.30 vs. limit=15.0
2024-09-23 04:48:46,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=180852.0, ans=0.125
2024-09-23 04:48:51,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=180852.0, ans=0.95
2024-09-23 04:48:57,313 INFO [train.py:1198] (1/4) Epoch 10, batch 3700, loss[loss=0.2625, ctc_loss=0.1836, cr_loss=0.3944, over 17035.00 frames. ], tot_loss[loss=0.2515, ctc_loss=0.1747, cr_loss=0.3843, over 3352537.98 frames. ], batch size: 51, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:48:59,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1.whitening_limit, batch_count=180898.66666666666, ans=10.0
2024-09-23 04:49:02,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=180898.66666666666, ans=0.2
2024-09-23 04:49:56,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=181038.66666666666, ans=0.2
2024-09-23 04:50:16,592 INFO [train.py:1198] (1/4) Epoch 10, batch 3750, loss[loss=0.2883, ctc_loss=0.2019, cr_loss=0.432, over 16996.00 frames. ], tot_loss[loss=0.2521, ctc_loss=0.1751, cr_loss=0.3846, over 3344408.31 frames. ], batch size: 53, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:50:46,085 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.346e+02 1.480e+02 1.709e+02 3.123e+02, threshold=2.960e+02, percent-clipped=1.0
2024-09-23 04:51:33,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=181365.33333333334, ans=0.04949747468305833
2024-09-23 04:51:34,558 INFO [train.py:1198] (1/4) Epoch 10, batch 3800, loss[loss=0.2217, ctc_loss=0.1525, cr_loss=0.3456, over 17098.00 frames. ], tot_loss[loss=0.2502, ctc_loss=0.1738, cr_loss=0.3824, over 3348506.94 frames. ], batch size: 40, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:51:39,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=181365.33333333334, ans=0.1
2024-09-23 04:51:44,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=181365.33333333334, ans=0.0
2024-09-23 04:51:58,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=181412.0, ans=0.1
2024-09-23 04:52:01,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=181412.0, ans=0.125
2024-09-23 04:52:19,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=181505.33333333334, ans=0.125
2024-09-23 04:52:38,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=181552.0, ans=0.07
2024-09-23 04:52:53,254 INFO [train.py:1198] (1/4) Epoch 10, batch 3850, loss[loss=0.2031, ctc_loss=0.1375, cr_loss=0.3277, over 16666.00 frames. ], tot_loss[loss=0.2505, ctc_loss=0.1741, cr_loss=0.3818, over 3323909.03 frames. ], batch size: 37, lr: 1.18e-02, grad_scale: 32.0
2024-09-23 04:52:56,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=181598.66666666666, ans=0.0
2024-09-23 04:53:12,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=181645.33333333334, ans=0.07
2024-09-23 04:53:18,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=181645.33333333334, ans=0.0
2024-09-23 04:53:22,407 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.346e+02 1.514e+02 1.745e+02 2.888e+02, threshold=3.027e+02, percent-clipped=0.0
2024-09-23 04:53:34,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=181692.0, ans=0.125
2024-09-23 04:53:53,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=181785.33333333334, ans=0.2
2024-09-23 04:54:55,685 INFO [train.py:1198] (1/4) Epoch 11, batch 0, loss[loss=0.2755, ctc_loss=0.1968, cr_loss=0.3938, over 15143.00 frames. ], tot_loss[loss=0.2755, ctc_loss=0.1968, cr_loss=0.3938, over 15143.00 frames. ], batch size: 89, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 04:54:55,685 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 04:55:11,312 INFO [train.py:1230] (1/4) Epoch 11, validation: loss=0.04963, ctc_loss=0.04963, cr_loss=7.372e-15, over 944034.00 frames. 
2024-09-23 04:55:11,313 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 04:55:19,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=181813.33333333334, ans=0.09899494936611666
2024-09-23 04:55:39,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.24 vs. limit=15.0
2024-09-23 04:55:39,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=181860.0, ans=0.0
2024-09-23 04:55:41,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=181860.0, ans=0.1
2024-09-23 04:55:43,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=181860.0, ans=0.125
2024-09-23 04:55:51,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=181906.66666666666, ans=0.025
2024-09-23 04:55:52,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=181906.66666666666, ans=10.0
2024-09-23 04:55:59,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=181906.66666666666, ans=0.125
2024-09-23 04:56:00,920 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 04:56:29,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=182000.0, ans=0.025
2024-09-23 04:56:34,326 INFO [train.py:1198] (1/4) Epoch 11, batch 50, loss[loss=0.2519, ctc_loss=0.1751, cr_loss=0.3841, over 17149.00 frames. ], tot_loss[loss=0.253, ctc_loss=0.1749, cr_loss=0.3905, over 765316.69 frames. ], batch size: 48, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 04:56:35,303 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.88 vs. limit=15.0
2024-09-23 04:56:44,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=182046.66666666666, ans=0.1
2024-09-23 04:56:58,697 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 04:57:12,614 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.435e+02 1.595e+02 1.830e+02 2.692e+02, threshold=3.191e+02, percent-clipped=0.0
2024-09-23 04:57:14,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=182140.0, ans=0.0
2024-09-23 04:57:14,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=182140.0, ans=0.125
2024-09-23 04:57:50,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.83 vs. limit=12.0
2024-09-23 04:57:52,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=182280.0, ans=0.125
2024-09-23 04:57:53,900 INFO [train.py:1198] (1/4) Epoch 11, batch 100, loss[loss=0.248, ctc_loss=0.1694, cr_loss=0.3929, over 17292.00 frames. ], tot_loss[loss=0.2494, ctc_loss=0.1724, cr_loss=0.3851, over 1351029.70 frames. ], batch size: 49, lr: 1.12e-02, grad_scale: 16.0
2024-09-23 04:58:13,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=182326.66666666666, ans=0.0
2024-09-23 04:58:18,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=182326.66666666666, ans=0.0
2024-09-23 04:58:36,248 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.26 vs. limit=22.5
2024-09-23 04:59:16,234 INFO [train.py:1198] (1/4) Epoch 11, batch 150, loss[loss=0.2746, ctc_loss=0.1924, cr_loss=0.4108, over 17010.00 frames. ], tot_loss[loss=0.2464, ctc_loss=0.1701, cr_loss=0.3817, over 1808388.85 frames. ], batch size: 56, lr: 1.12e-02, grad_scale: 16.0
2024-09-23 04:59:24,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=182513.33333333334, ans=0.125
2024-09-23 04:59:57,644 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.353e+02 1.511e+02 1.758e+02 2.701e+02, threshold=3.021e+02, percent-clipped=0.0
2024-09-23 05:00:18,189 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 05:00:24,921 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.87 vs. limit=6.0
2024-09-23 05:00:29,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=182700.0, ans=0.125
2024-09-23 05:00:39,349 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=8.96 vs. limit=15.0
2024-09-23 05:00:41,713 INFO [train.py:1198] (1/4) Epoch 11, batch 200, loss[loss=0.3004, ctc_loss=0.2145, cr_loss=0.4295, over 14740.00 frames. ], tot_loss[loss=0.2486, ctc_loss=0.1719, cr_loss=0.3833, over 2152463.42 frames. ], batch size: 89, lr: 1.12e-02, grad_scale: 16.0
2024-09-23 05:00:45,772 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.56 vs. limit=15.0
2024-09-23 05:00:59,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=182793.33333333334, ans=0.125
2024-09-23 05:01:23,512 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 05:01:24,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=182840.0, ans=0.2
2024-09-23 05:02:01,465 INFO [train.py:1198] (1/4) Epoch 11, batch 250, loss[loss=0.2596, ctc_loss=0.178, cr_loss=0.408, over 17205.00 frames. ], tot_loss[loss=0.2478, ctc_loss=0.1713, cr_loss=0.3828, over 2427434.24 frames. ], batch size: 55, lr: 1.12e-02, grad_scale: 16.0
2024-09-23 05:02:03,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=182980.0, ans=0.125
2024-09-23 05:02:39,687 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.357e+02 1.521e+02 1.743e+02 2.579e+02, threshold=3.043e+02, percent-clipped=0.0
2024-09-23 05:03:20,848 INFO [train.py:1198] (1/4) Epoch 11, batch 300, loss[loss=0.2481, ctc_loss=0.1704, cr_loss=0.3886, over 17014.00 frames. ], tot_loss[loss=0.246, ctc_loss=0.1699, cr_loss=0.3806, over 2637712.13 frames. ], batch size: 51, lr: 1.12e-02, grad_scale: 16.0
2024-09-23 05:03:27,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=183213.33333333334, ans=0.1
2024-09-23 05:03:43,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=183260.0, ans=0.025
2024-09-23 05:03:49,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=183260.0, ans=0.125
2024-09-23 05:04:15,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=183353.33333333334, ans=0.125
2024-09-23 05:04:26,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=183400.0, ans=0.2
2024-09-23 05:04:29,737 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.89 vs. limit=12.0
2024-09-23 05:04:35,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=183400.0, ans=0.0
2024-09-23 05:04:49,183 INFO [train.py:1198] (1/4) Epoch 11, batch 350, loss[loss=0.2645, ctc_loss=0.1868, cr_loss=0.3884, over 17026.00 frames. ], tot_loss[loss=0.2469, ctc_loss=0.1706, cr_loss=0.3816, over 2797912.82 frames. ], batch size: 53, lr: 1.12e-02, grad_scale: 16.0
2024-09-23 05:04:58,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=183446.66666666666, ans=0.125
2024-09-23 05:05:30,065 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.290e+02 1.386e+02 1.605e+02 2.385e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-23 05:05:47,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=183586.66666666666, ans=0.025
2024-09-23 05:06:01,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=183633.33333333334, ans=0.2
2024-09-23 05:06:02,893 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.79 vs. limit=6.0
2024-09-23 05:06:11,272 INFO [train.py:1198] (1/4) Epoch 11, batch 400, loss[loss=0.2943, ctc_loss=0.21, cr_loss=0.4214, over 17019.00 frames. ], tot_loss[loss=0.2472, ctc_loss=0.171, cr_loss=0.3813, over 2911231.12 frames. ], batch size: 52, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 05:06:23,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.21 vs. limit=10.0
2024-09-23 05:07:09,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=183820.0, ans=0.125
2024-09-23 05:07:14,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=183866.66666666666, ans=0.0
2024-09-23 05:07:31,441 INFO [train.py:1198] (1/4) Epoch 11, batch 450, loss[loss=0.2424, ctc_loss=0.1661, cr_loss=0.3814, over 17111.00 frames. ], tot_loss[loss=0.2476, ctc_loss=0.1713, cr_loss=0.3815, over 3005936.94 frames. ], batch size: 49, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 05:07:46,809 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.24 vs. limit=15.0
2024-09-23 05:07:49,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=183960.0, ans=0.125
2024-09-23 05:07:54,162 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 05:08:09,515 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.312e+02 1.426e+02 1.601e+02 2.161e+02, threshold=2.852e+02, percent-clipped=0.0
2024-09-23 05:08:09,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=184006.66666666666, ans=0.0
2024-09-23 05:08:16,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=184006.66666666666, ans=0.1
2024-09-23 05:08:22,810 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.29 vs. limit=15.0
2024-09-23 05:08:27,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=184053.33333333334, ans=0.1
2024-09-23 05:08:53,365 INFO [train.py:1198] (1/4) Epoch 11, batch 500, loss[loss=0.2339, ctc_loss=0.1607, cr_loss=0.3661, over 17148.00 frames. ], tot_loss[loss=0.2479, ctc_loss=0.1716, cr_loss=0.3815, over 3085804.82 frames. ], batch size: 45, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 05:08:53,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=184146.66666666666, ans=0.1
2024-09-23 05:09:17,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=184193.33333333334, ans=0.125
2024-09-23 05:09:38,168 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.59 vs. limit=15.0
2024-09-23 05:10:05,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=184333.33333333334, ans=0.025
2024-09-23 05:10:22,037 INFO [train.py:1198] (1/4) Epoch 11, batch 550, loss[loss=0.2184, ctc_loss=0.1496, cr_loss=0.3438, over 17298.00 frames. ], tot_loss[loss=0.2479, ctc_loss=0.1715, cr_loss=0.382, over 3145310.59 frames. ], batch size: 46, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 05:10:34,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=184380.0, ans=0.0
2024-09-23 05:10:36,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=184426.66666666666, ans=0.07
2024-09-23 05:10:36,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=184426.66666666666, ans=0.125
2024-09-23 05:11:00,279 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.305e+02 1.444e+02 1.636e+02 3.823e+02, threshold=2.888e+02, percent-clipped=1.0
2024-09-23 05:11:06,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=184473.33333333334, ans=0.125
2024-09-23 05:11:07,215 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.49 vs. limit=15.0
2024-09-23 05:11:17,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=184520.0, ans=0.035
2024-09-23 05:11:31,400 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.02 vs. limit=15.0
2024-09-23 05:11:41,804 INFO [train.py:1198] (1/4) Epoch 11, batch 600, loss[loss=0.2592, ctc_loss=0.1771, cr_loss=0.4105, over 17097.00 frames. ], tot_loss[loss=0.2472, ctc_loss=0.171, cr_loss=0.3812, over 3198684.48 frames. ], batch size: 49, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 05:11:48,945 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.17 vs. limit=15.0
2024-09-23 05:12:01,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=184660.0, ans=0.0
2024-09-23 05:12:32,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=184753.33333333334, ans=0.125
2024-09-23 05:12:36,666 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.08 vs. limit=15.0
2024-09-23 05:12:46,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=6.49 vs. limit=15.0
2024-09-23 05:13:01,768 INFO [train.py:1198] (1/4) Epoch 11, batch 650, loss[loss=0.2498, ctc_loss=0.1697, cr_loss=0.4005, over 16785.00 frames. ], tot_loss[loss=0.2456, ctc_loss=0.1696, cr_loss=0.38, over 3239460.23 frames. ], batch size: 61, lr: 1.12e-02, grad_scale: 32.0
2024-09-23 05:13:08,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=184846.66666666666, ans=0.0
2024-09-23 05:13:21,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=184893.33333333334, ans=0.125
2024-09-23 05:13:37,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=184940.0, ans=0.2
2024-09-23 05:13:43,066 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.331e+02 1.445e+02 1.618e+02 2.572e+02, threshold=2.890e+02, percent-clipped=0.0
2024-09-23 05:14:27,583 INFO [train.py:1198] (1/4) Epoch 11, batch 700, loss[loss=0.245, ctc_loss=0.1734, cr_loss=0.3578, over 17222.00 frames. ], tot_loss[loss=0.2461, ctc_loss=0.17, cr_loss=0.3802, over 3275252.89 frames. ], batch size: 50, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:14:29,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.85 vs. limit=15.0
2024-09-23 05:14:39,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.whiten.whitening_limit, batch_count=185080.0, ans=12.0
2024-09-23 05:15:25,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=185220.0, ans=0.125
2024-09-23 05:15:41,959 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.45 vs. limit=15.0
2024-09-23 05:15:43,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=185266.66666666666, ans=0.125
2024-09-23 05:15:49,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=185266.66666666666, ans=0.125
2024-09-23 05:15:52,062 INFO [train.py:1198] (1/4) Epoch 11, batch 750, loss[loss=0.2143, ctc_loss=0.1431, cr_loss=0.3558, over 17085.00 frames. ], tot_loss[loss=0.2453, ctc_loss=0.1694, cr_loss=0.3793, over 3298229.16 frames. ], batch size: 43, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:16:05,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=185313.33333333334, ans=0.125
2024-09-23 05:16:16,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=185360.0, ans=0.0
2024-09-23 05:16:20,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=185360.0, ans=0.125
2024-09-23 05:16:20,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=185360.0, ans=0.1
2024-09-23 05:16:30,315 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.321e+02 1.443e+02 1.699e+02 2.904e+02, threshold=2.886e+02, percent-clipped=1.0
2024-09-23 05:16:30,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.21 vs. limit=12.0
2024-09-23 05:16:47,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.18 vs. limit=15.0
2024-09-23 05:16:49,008 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.82 vs. limit=10.0
2024-09-23 05:16:53,435 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.23 vs. limit=22.5
2024-09-23 05:16:56,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=185500.0, ans=0.125
2024-09-23 05:17:05,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_abs, batch_count=185500.0, ans=0.5
2024-09-23 05:17:05,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.min_positive, batch_count=185500.0, ans=0.025
2024-09-23 05:17:08,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=185500.0, ans=0.1
2024-09-23 05:17:11,521 INFO [train.py:1198] (1/4) Epoch 11, batch 800, loss[loss=0.2539, ctc_loss=0.1782, cr_loss=0.3787, over 16996.00 frames. ], tot_loss[loss=0.245, ctc_loss=0.1693, cr_loss=0.3781, over 3310543.75 frames. ], batch size: 51, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:17:43,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=185640.0, ans=10.0
2024-09-23 05:18:02,874 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.43 vs. limit=15.0
2024-09-23 05:18:29,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=185780.0, ans=0.0
2024-09-23 05:18:31,203 INFO [train.py:1198] (1/4) Epoch 11, batch 850, loss[loss=0.2194, ctc_loss=0.1485, cr_loss=0.3548, over 17302.00 frames. ], tot_loss[loss=0.2437, ctc_loss=0.1684, cr_loss=0.3764, over 3322021.38 frames. ], batch size: 46, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:18:51,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=185826.66666666666, ans=0.0
2024-09-23 05:19:04,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=185873.33333333334, ans=0.0
2024-09-23 05:19:12,273 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.355e+02 1.504e+02 1.794e+02 2.469e+02, threshold=3.008e+02, percent-clipped=0.0
2024-09-23 05:19:25,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=185920.0, ans=0.0
2024-09-23 05:19:59,415 INFO [train.py:1198] (1/4) Epoch 11, batch 900, loss[loss=0.2426, ctc_loss=0.1655, cr_loss=0.3855, over 17106.00 frames. ], tot_loss[loss=0.2452, ctc_loss=0.1695, cr_loss=0.3787, over 3331392.51 frames. ], batch size: 43, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:20:08,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=186013.33333333334, ans=0.125
2024-09-23 05:20:16,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=186060.0, ans=0.2
2024-09-23 05:20:21,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=186060.0, ans=0.0
2024-09-23 05:20:47,703 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=5.07 vs. limit=12.0
2024-09-23 05:21:19,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=186200.0, ans=0.125
2024-09-23 05:21:22,044 INFO [train.py:1198] (1/4) Epoch 11, batch 950, loss[loss=0.234, ctc_loss=0.1591, cr_loss=0.3743, over 17014.00 frames. ], tot_loss[loss=0.2459, ctc_loss=0.1701, cr_loss=0.3793, over 3336431.68 frames. ], batch size: 44, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:21:40,298 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 05:21:45,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=186293.33333333334, ans=0.125
2024-09-23 05:21:45,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=186293.33333333334, ans=0.125
2024-09-23 05:21:51,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=186293.33333333334, ans=0.0
2024-09-23 05:21:56,741 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=12.32 vs. limit=15.0
2024-09-23 05:22:00,860 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.261e+02 1.404e+02 1.572e+02 2.138e+02, threshold=2.809e+02, percent-clipped=0.0
2024-09-23 05:22:29,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=186433.33333333334, ans=0.125
2024-09-23 05:22:42,315 INFO [train.py:1198] (1/4) Epoch 11, batch 1000, loss[loss=0.2441, ctc_loss=0.1679, cr_loss=0.3809, over 17293.00 frames. ], tot_loss[loss=0.2455, ctc_loss=0.1697, cr_loss=0.3788, over 3336470.84 frames. ], batch size: 49, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:23:33,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=186620.0, ans=0.1
2024-09-23 05:23:44,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=186620.0, ans=0.125
2024-09-23 05:23:56,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=186666.66666666666, ans=0.1
2024-09-23 05:24:00,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=186666.66666666666, ans=0.1
2024-09-23 05:24:06,720 INFO [train.py:1198] (1/4) Epoch 11, batch 1050, loss[loss=0.2931, ctc_loss=0.2013, cr_loss=0.459, over 17028.00 frames. ], tot_loss[loss=0.2468, ctc_loss=0.1707, cr_loss=0.3805, over 3342384.49 frames. ], batch size: 56, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:24:13,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=186713.33333333334, ans=0.1
2024-09-23 05:24:32,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=186760.0, ans=0.125
2024-09-23 05:24:50,586 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.363e+02 1.550e+02 2.011e+02 3.304e+02, threshold=3.099e+02, percent-clipped=2.0
2024-09-23 05:25:02,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=186853.33333333334, ans=0.1
2024-09-23 05:25:13,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=186853.33333333334, ans=0.04949747468305833
2024-09-23 05:25:28,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=186900.0, ans=0.125
2024-09-23 05:25:34,180 INFO [train.py:1198] (1/4) Epoch 11, batch 1100, loss[loss=0.2132, ctc_loss=0.145, cr_loss=0.3408, over 17223.00 frames. ], tot_loss[loss=0.2451, ctc_loss=0.1694, cr_loss=0.3786, over 3351603.23 frames. ], batch size: 47, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:26:17,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=187040.0, ans=0.0
2024-09-23 05:26:20,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=187086.66666666666, ans=0.125
2024-09-23 05:26:53,985 INFO [train.py:1198] (1/4) Epoch 11, batch 1150, loss[loss=0.251, ctc_loss=0.1697, cr_loss=0.4064, over 17182.00 frames. ], tot_loss[loss=0.2466, ctc_loss=0.1705, cr_loss=0.3804, over 3350675.91 frames. ], batch size: 45, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:26:55,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=187180.0, ans=0.0
2024-09-23 05:27:21,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=187226.66666666666, ans=0.2
2024-09-23 05:27:32,067 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.352e+02 1.474e+02 1.715e+02 2.168e+02, threshold=2.948e+02, percent-clipped=0.0
2024-09-23 05:27:33,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=187273.33333333334, ans=0.1
2024-09-23 05:27:44,081 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.51 vs. limit=12.0
2024-09-23 05:27:53,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer_na.min_abs, batch_count=187320.0, ans=0.02
2024-09-23 05:27:54,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=187320.0, ans=10.0
2024-09-23 05:27:57,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=187366.66666666666, ans=0.125
2024-09-23 05:28:11,048 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=14.42 vs. limit=15.0
2024-09-23 05:28:13,188 INFO [train.py:1198] (1/4) Epoch 11, batch 1200, loss[loss=0.2378, ctc_loss=0.164, cr_loss=0.3686, over 17064.00 frames. ], tot_loss[loss=0.246, ctc_loss=0.1701, cr_loss=0.3792, over 3354503.65 frames. ], batch size: 46, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:28:26,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=187413.33333333334, ans=0.0
2024-09-23 05:28:43,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=187460.0, ans=0.0
2024-09-23 05:29:16,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=187553.33333333334, ans=0.125
2024-09-23 05:29:28,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=187600.0, ans=0.1
2024-09-23 05:29:41,427 INFO [train.py:1198] (1/4) Epoch 11, batch 1250, loss[loss=0.268, ctc_loss=0.191, cr_loss=0.3851, over 16872.00 frames. ], tot_loss[loss=0.2454, ctc_loss=0.1697, cr_loss=0.3785, over 3359514.65 frames. ], batch size: 58, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:29:44,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=187646.66666666666, ans=0.0
2024-09-23 05:30:13,299 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=187693.33333333334, ans=0.125
2024-09-23 05:30:22,691 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.347e+02 1.488e+02 1.647e+02 3.078e+02, threshold=2.976e+02, percent-clipped=1.0
2024-09-23 05:30:51,795 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.80 vs. limit=15.0
2024-09-23 05:30:59,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=187833.33333333334, ans=0.125
2024-09-23 05:31:04,205 INFO [train.py:1198] (1/4) Epoch 11, batch 1300, loss[loss=0.2255, ctc_loss=0.1534, cr_loss=0.3605, over 17302.00 frames. ], tot_loss[loss=0.2459, ctc_loss=0.17, cr_loss=0.3793, over 3367081.18 frames. ], batch size: 49, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:31:09,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=187880.0, ans=0.125
2024-09-23 05:31:26,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=187926.66666666666, ans=0.125
2024-09-23 05:31:29,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=187926.66666666666, ans=0.1
2024-09-23 05:32:00,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=188020.0, ans=0.0
2024-09-23 05:32:06,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=188066.66666666666, ans=0.025
2024-09-23 05:32:23,726 INFO [train.py:1198] (1/4) Epoch 11, batch 1350, loss[loss=0.267, ctc_loss=0.1924, cr_loss=0.3735, over 15757.00 frames. ], tot_loss[loss=0.2465, ctc_loss=0.1706, cr_loss=0.3799, over 3363279.33 frames. ], batch size: 74, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:32:35,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=188113.33333333334, ans=0.0
2024-09-23 05:33:02,037 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.248e+02 1.402e+02 1.590e+02 2.779e+02, threshold=2.805e+02, percent-clipped=0.0
2024-09-23 05:33:03,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_abs, batch_count=188206.66666666666, ans=0.5
2024-09-23 05:33:07,275 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 05:33:27,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=188300.0, ans=0.0
2024-09-23 05:33:29,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=188300.0, ans=0.0
2024-09-23 05:33:44,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=188346.66666666666, ans=0.125
2024-09-23 05:33:45,864 INFO [train.py:1198] (1/4) Epoch 11, batch 1400, loss[loss=0.2327, ctc_loss=0.159, cr_loss=0.3688, over 17114.00 frames. ], tot_loss[loss=0.2468, ctc_loss=0.1707, cr_loss=0.3803, over 3360783.02 frames. ], batch size: 40, lr: 1.11e-02, grad_scale: 32.0
2024-09-23 05:33:57,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=188346.66666666666, ans=0.0
2024-09-23 05:34:05,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=188393.33333333334, ans=0.125
2024-09-23 05:35:13,557 INFO [train.py:1198] (1/4) Epoch 11, batch 1450, loss[loss=0.2749, ctc_loss=0.189, cr_loss=0.4297, over 17347.00 frames. ], tot_loss[loss=0.2476, ctc_loss=0.1713, cr_loss=0.3815, over 3360962.29 frames. ], batch size: 48, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:35:39,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=188626.66666666666, ans=0.0
2024-09-23 05:35:51,362 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.355e+02 1.506e+02 1.702e+02 2.506e+02, threshold=3.011e+02, percent-clipped=0.0
2024-09-23 05:36:10,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=188720.0, ans=0.0
2024-09-23 05:36:12,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.40 vs. limit=15.0
2024-09-23 05:36:25,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=188766.66666666666, ans=0.125
2024-09-23 05:36:32,872 INFO [train.py:1198] (1/4) Epoch 11, batch 1500, loss[loss=0.2735, ctc_loss=0.1943, cr_loss=0.396, over 17203.00 frames. ], tot_loss[loss=0.2471, ctc_loss=0.1708, cr_loss=0.3815, over 3366371.92 frames. ], batch size: 47, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:36:39,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=188813.33333333334, ans=0.125
2024-09-23 05:36:46,100 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=188813.33333333334, ans=0.125
2024-09-23 05:37:02,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=188860.0, ans=0.0
2024-09-23 05:37:12,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.06 vs. limit=15.0
2024-09-23 05:37:16,956 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.54 vs. limit=15.0
2024-09-23 05:37:35,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=189000.0, ans=0.2
2024-09-23 05:37:52,941 INFO [train.py:1198] (1/4) Epoch 11, batch 1550, loss[loss=0.277, ctc_loss=0.1976, cr_loss=0.3969, over 15925.00 frames. ], tot_loss[loss=0.2477, ctc_loss=0.1712, cr_loss=0.3826, over 3371593.05 frames. ], batch size: 74, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:37:58,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=189046.66666666666, ans=0.125
2024-09-23 05:38:00,322 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.18 vs. limit=15.0
2024-09-23 05:38:01,619 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.96 vs. limit=15.0
2024-09-23 05:38:04,907 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.74 vs. limit=15.0
2024-09-23 05:38:06,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=189046.66666666666, ans=0.0
2024-09-23 05:38:33,047 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.268e+02 1.378e+02 1.550e+02 2.066e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-23 05:39:07,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=189233.33333333334, ans=0.125
2024-09-23 05:39:15,064 INFO [train.py:1198] (1/4) Epoch 11, batch 1600, loss[loss=0.3053, ctc_loss=0.2264, cr_loss=0.3947, over 11410.00 frames. ], tot_loss[loss=0.2481, ctc_loss=0.1717, cr_loss=0.3821, over 3353622.15 frames. ], batch size: 123, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:39:22,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=189280.0, ans=0.0
2024-09-23 05:39:24,697 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.84 vs. limit=22.5
2024-09-23 05:39:34,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=189326.66666666666, ans=0.09899494936611666
2024-09-23 05:40:04,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=189373.33333333334, ans=0.2
2024-09-23 05:40:27,629 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=7.68 vs. limit=12.0
2024-09-23 05:40:28,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=189466.66666666666, ans=0.125
2024-09-23 05:40:37,483 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.52 vs. limit=22.5
2024-09-23 05:40:42,829 INFO [train.py:1198] (1/4) Epoch 11, batch 1650, loss[loss=0.1945, ctc_loss=0.1313, cr_loss=0.3157, over 16963.00 frames. ], tot_loss[loss=0.2485, ctc_loss=0.172, cr_loss=0.3826, over 3351724.01 frames. ], batch size: 42, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:41:22,848 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.323e+02 1.468e+02 1.701e+02 2.632e+02, threshold=2.937e+02, percent-clipped=0.0
2024-09-23 05:41:38,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=189653.33333333334, ans=0.125
2024-09-23 05:41:48,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=189700.0, ans=0.025
2024-09-23 05:42:02,591 INFO [train.py:1198] (1/4) Epoch 11, batch 1700, loss[loss=0.2544, ctc_loss=0.179, cr_loss=0.377, over 17351.00 frames. ], tot_loss[loss=0.2471, ctc_loss=0.171, cr_loss=0.3806, over 3348662.01 frames. ], batch size: 48, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:42:09,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=189746.66666666666, ans=0.0
2024-09-23 05:42:52,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=189886.66666666666, ans=0.2
2024-09-23 05:42:55,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=189886.66666666666, ans=0.0
2024-09-23 05:43:11,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=189933.33333333334, ans=0.125
2024-09-23 05:43:22,420 INFO [train.py:1198] (1/4) Epoch 11, batch 1750, loss[loss=0.2474, ctc_loss=0.1701, cr_loss=0.3864, over 17222.00 frames. ], tot_loss[loss=0.2471, ctc_loss=0.1709, cr_loss=0.3807, over 3359231.05 frames. ], batch size: 50, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:43:41,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=190026.66666666666, ans=0.2
2024-09-23 05:43:43,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=190026.66666666666, ans=0.125
2024-09-23 05:44:04,904 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.343e+02 1.475e+02 1.671e+02 2.216e+02, threshold=2.950e+02, percent-clipped=0.0
2024-09-23 05:44:20,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=190120.0, ans=0.0
2024-09-23 05:44:21,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=190120.0, ans=0.1
2024-09-23 05:44:52,206 INFO [train.py:1198] (1/4) Epoch 11, batch 1800, loss[loss=0.2508, ctc_loss=0.1777, cr_loss=0.3655, over 17313.00 frames. ], tot_loss[loss=0.2466, ctc_loss=0.1706, cr_loss=0.3801, over 3358366.61 frames. ], batch size: 49, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:44:52,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=190213.33333333334, ans=0.1
2024-09-23 05:45:21,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=190260.0, ans=0.125
2024-09-23 05:45:27,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=190306.66666666666, ans=0.0
2024-09-23 05:45:30,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=190306.66666666666, ans=0.09899494936611666
2024-09-23 05:45:32,995 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.89 vs. limit=15.0
2024-09-23 05:45:37,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=190306.66666666666, ans=0.0
2024-09-23 05:45:40,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=190353.33333333334, ans=0.125
2024-09-23 05:45:45,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=190353.33333333334, ans=0.125
2024-09-23 05:45:55,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.89 vs. limit=15.0
2024-09-23 05:46:12,735 INFO [train.py:1198] (1/4) Epoch 11, batch 1850, loss[loss=0.2095, ctc_loss=0.1415, cr_loss=0.34, over 17006.00 frames. ], tot_loss[loss=0.2458, ctc_loss=0.1699, cr_loss=0.3794, over 3362962.98 frames. ], batch size: 44, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:46:35,595 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.35 vs. limit=15.0
2024-09-23 05:46:43,378 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 05:46:52,382 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.336e+02 1.526e+02 1.737e+02 2.806e+02, threshold=3.051e+02, percent-clipped=0.0
2024-09-23 05:47:03,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=190586.66666666666, ans=0.0
2024-09-23 05:47:07,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=190586.66666666666, ans=0.125
2024-09-23 05:47:13,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=190586.66666666666, ans=0.125
2024-09-23 05:47:16,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=190633.33333333334, ans=0.125
2024-09-23 05:47:17,091 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.74 vs. limit=12.0
2024-09-23 05:47:32,658 INFO [train.py:1198] (1/4) Epoch 11, batch 1900, loss[loss=0.2725, ctc_loss=0.1936, cr_loss=0.3943, over 16893.00 frames. ], tot_loss[loss=0.2453, ctc_loss=0.1695, cr_loss=0.3794, over 3366485.15 frames. ], batch size: 58, lr: 1.10e-02, grad_scale: 16.0
2024-09-23 05:47:36,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=190680.0, ans=0.125
2024-09-23 05:48:55,013 INFO [train.py:1198] (1/4) Epoch 11, batch 1950, loss[loss=0.2897, ctc_loss=0.2057, cr_loss=0.4202, over 15158.00 frames. ], tot_loss[loss=0.2464, ctc_loss=0.1702, cr_loss=0.3809, over 3370049.76 frames. ], batch size: 89, lr: 1.10e-02, grad_scale: 16.0
2024-09-23 05:48:58,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=190913.33333333334, ans=0.125
2024-09-23 05:49:26,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=190960.0, ans=0.125
2024-09-23 05:49:37,889 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=2.79 vs. limit=15.0
2024-09-23 05:49:41,889 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.367e+02 1.554e+02 1.740e+02 3.825e+02, threshold=3.108e+02, percent-clipped=1.0
2024-09-23 05:49:50,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=191053.33333333334, ans=0.125
2024-09-23 05:50:00,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=191053.33333333334, ans=0.2
2024-09-23 05:50:01,418 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.24 vs. limit=12.0
2024-09-23 05:50:05,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=191100.0, ans=0.125
2024-09-23 05:50:11,146 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=8.34 vs. limit=15.0
2024-09-23 05:50:22,739 INFO [train.py:1198] (1/4) Epoch 11, batch 2000, loss[loss=0.2679, ctc_loss=0.1884, cr_loss=0.3975, over 16772.00 frames. ], tot_loss[loss=0.2464, ctc_loss=0.1702, cr_loss=0.381, over 3371865.64 frames. ], batch size: 61, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:50:32,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=191146.66666666666, ans=0.125
2024-09-23 05:50:39,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=6.29 vs. limit=15.0
2024-09-23 05:51:29,831 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=10.59 vs. limit=15.0
2024-09-23 05:51:42,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=191380.0, ans=0.125
2024-09-23 05:51:43,321 INFO [train.py:1198] (1/4) Epoch 11, batch 2050, loss[loss=0.2256, ctc_loss=0.1546, cr_loss=0.3547, over 16967.00 frames. ], tot_loss[loss=0.2465, ctc_loss=0.1703, cr_loss=0.3812, over 3372367.97 frames. ], batch size: 42, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:51:45,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=191380.0, ans=0.025
2024-09-23 05:51:55,358 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.59 vs. limit=22.5
2024-09-23 05:52:15,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=191473.33333333334, ans=0.0
2024-09-23 05:52:23,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=191473.33333333334, ans=0.0
2024-09-23 05:52:24,542 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.258e+02 1.361e+02 1.544e+02 2.924e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-23 05:52:47,870 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.23 vs. limit=22.5
2024-09-23 05:52:53,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=191566.66666666666, ans=0.025
2024-09-23 05:53:03,035 INFO [train.py:1198] (1/4) Epoch 11, batch 2100, loss[loss=0.2106, ctc_loss=0.1423, cr_loss=0.3416, over 17142.00 frames. ], tot_loss[loss=0.2452, ctc_loss=0.1694, cr_loss=0.379, over 3366823.70 frames. ], batch size: 45, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:53:13,480 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.79 vs. limit=22.5
2024-09-23 05:53:14,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=191613.33333333334, ans=0.0
2024-09-23 05:53:19,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=191660.0, ans=0.2
2024-09-23 05:53:58,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=191753.33333333334, ans=0.125
2024-09-23 05:54:30,862 INFO [train.py:1198] (1/4) Epoch 11, batch 2150, loss[loss=0.2004, ctc_loss=0.1351, cr_loss=0.3263, over 16658.00 frames. ], tot_loss[loss=0.2446, ctc_loss=0.1689, cr_loss=0.3787, over 3373279.20 frames. ], batch size: 37, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:54:36,734 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.23 vs. limit=15.0
2024-09-23 05:54:40,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=191846.66666666666, ans=0.2
2024-09-23 05:54:54,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=191893.33333333334, ans=0.0
2024-09-23 05:55:14,825 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.341e+02 1.514e+02 1.705e+02 2.483e+02, threshold=3.028e+02, percent-clipped=0.0
2024-09-23 05:55:53,281 INFO [train.py:1198] (1/4) Epoch 11, batch 2200, loss[loss=0.261, ctc_loss=0.1811, cr_loss=0.3994, over 17353.00 frames. ], tot_loss[loss=0.2452, ctc_loss=0.1693, cr_loss=0.3793, over 3379231.50 frames. ], batch size: 48, lr: 1.10e-02, grad_scale: 32.0
2024-09-23 05:56:11,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=192126.66666666666, ans=0.125
2024-09-23 05:56:12,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=192126.66666666666, ans=0.1
2024-09-23 05:57:13,773 INFO [train.py:1198] (1/4) Epoch 11, batch 2250, loss[loss=0.204, ctc_loss=0.1374, cr_loss=0.3328, over 17205.00 frames. ], tot_loss[loss=0.2439, ctc_loss=0.1683, cr_loss=0.3782, over 3381595.40 frames. ], batch size: 41, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 05:57:32,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.20 vs. limit=22.5
2024-09-23 05:57:55,416 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.304e+02 1.378e+02 1.493e+02 3.334e+02, threshold=2.757e+02, percent-clipped=1.0
2024-09-23 05:58:18,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=192500.0, ans=0.125
2024-09-23 05:58:36,212 INFO [train.py:1198] (1/4) Epoch 11, batch 2300, loss[loss=0.2599, ctc_loss=0.1816, cr_loss=0.3916, over 17018.00 frames. ], tot_loss[loss=0.2429, ctc_loss=0.1675, cr_loss=0.377, over 3380118.92 frames. ], batch size: 56, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 05:58:53,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=192593.33333333334, ans=0.1
2024-09-23 05:59:04,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=192593.33333333334, ans=0.125
2024-09-23 05:59:09,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=192640.0, ans=0.125
2024-09-23 05:59:35,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=192686.66666666666, ans=0.0
2024-09-23 05:59:42,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=192686.66666666666, ans=0.125
2024-09-23 06:00:03,694 INFO [train.py:1198] (1/4) Epoch 11, batch 2350, loss[loss=0.2585, ctc_loss=0.1812, cr_loss=0.3866, over 17304.00 frames. ], tot_loss[loss=0.2446, ctc_loss=0.1688, cr_loss=0.3788, over 3363230.26 frames. ], batch size: 49, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:00:05,796 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.83 vs. limit=10.0
2024-09-23 06:00:19,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=192826.66666666666, ans=0.2
2024-09-23 06:00:22,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=192826.66666666666, ans=0.1
2024-09-23 06:00:29,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=192826.66666666666, ans=0.0
2024-09-23 06:00:43,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=192873.33333333334, ans=0.0
2024-09-23 06:00:44,552 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.266e+02 1.351e+02 1.522e+02 2.296e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-23 06:00:56,488 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.63 vs. limit=15.0
2024-09-23 06:00:59,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=192920.0, ans=0.0
2024-09-23 06:00:59,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=192920.0, ans=0.125
2024-09-23 06:01:02,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=192920.0, ans=0.125
2024-09-23 06:01:23,281 INFO [train.py:1198] (1/4) Epoch 11, batch 2400, loss[loss=0.2145, ctc_loss=0.1485, cr_loss=0.3296, over 17263.00 frames. ], tot_loss[loss=0.2452, ctc_loss=0.1694, cr_loss=0.3791, over 3362988.52 frames. ], batch size: 42, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:01:36,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=193013.33333333334, ans=0.0
2024-09-23 06:01:39,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=193060.0, ans=0.0
2024-09-23 06:01:54,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=193106.66666666666, ans=0.0
2024-09-23 06:02:00,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=193106.66666666666, ans=0.2
2024-09-23 06:02:00,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=193106.66666666666, ans=0.0
2024-09-23 06:02:00,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=193106.66666666666, ans=0.125
2024-09-23 06:02:02,693 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=19.97 vs. limit=22.5
2024-09-23 06:02:13,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=193153.33333333334, ans=0.2
2024-09-23 06:02:18,254 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.78 vs. limit=22.5
2024-09-23 06:02:43,184 INFO [train.py:1198] (1/4) Epoch 11, batch 2450, loss[loss=0.2853, ctc_loss=0.1976, cr_loss=0.4383, over 17088.00 frames. ], tot_loss[loss=0.2448, ctc_loss=0.169, cr_loss=0.3788, over 3366440.07 frames. ], batch size: 49, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:03:24,778 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.301e+02 1.390e+02 1.544e+02 1.973e+02, threshold=2.781e+02, percent-clipped=0.0
2024-09-23 06:03:26,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer_na.min_abs, batch_count=193340.0, ans=0.02
2024-09-23 06:03:37,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=193386.66666666666, ans=0.125
2024-09-23 06:04:07,697 INFO [train.py:1198] (1/4) Epoch 11, batch 2500, loss[loss=0.2553, ctc_loss=0.1772, cr_loss=0.3901, over 17314.00 frames. ], tot_loss[loss=0.2451, ctc_loss=0.1692, cr_loss=0.3795, over 3364233.24 frames. ], batch size: 49, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:04:33,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=193526.66666666666, ans=0.0
2024-09-23 06:04:38,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=193526.66666666666, ans=0.0
2024-09-23 06:04:53,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=193573.33333333334, ans=0.0
2024-09-23 06:05:06,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten.whitening_limit, batch_count=193620.0, ans=15.0
2024-09-23 06:05:10,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=193620.0, ans=0.0
2024-09-23 06:05:16,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=193666.66666666666, ans=0.125
2024-09-23 06:05:32,500 INFO [train.py:1198] (1/4) Epoch 11, batch 2550, loss[loss=0.2388, ctc_loss=0.1638, cr_loss=0.3747, over 17233.00 frames. ], tot_loss[loss=0.2456, ctc_loss=0.1697, cr_loss=0.3795, over 3354494.19 frames. ], batch size: 55, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:05:36,725 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.91 vs. limit=15.0
2024-09-23 06:05:39,751 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.20 vs. limit=15.0
2024-09-23 06:05:48,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=193760.0, ans=0.2
2024-09-23 06:06:06,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=193806.66666666666, ans=0.125
2024-09-23 06:06:06,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=193806.66666666666, ans=0.1
2024-09-23 06:06:13,861 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.308e+02 1.472e+02 1.682e+02 2.409e+02, threshold=2.944e+02, percent-clipped=0.0
2024-09-23 06:06:26,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=193853.33333333334, ans=0.07
2024-09-23 06:06:31,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=193853.33333333334, ans=0.0
2024-09-23 06:06:42,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=193900.0, ans=0.0
2024-09-23 06:06:51,767 INFO [train.py:1198] (1/4) Epoch 11, batch 2600, loss[loss=0.2163, ctc_loss=0.146, cr_loss=0.3516, over 17274.00 frames. ], tot_loss[loss=0.2447, ctc_loss=0.169, cr_loss=0.3782, over 3353050.12 frames. ], batch size: 44, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:06:56,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=193946.66666666666, ans=0.125
2024-09-23 06:07:01,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=193946.66666666666, ans=0.1
2024-09-23 06:07:03,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=3.99 vs. limit=15.0
2024-09-23 06:07:06,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=193993.33333333334, ans=0.1
2024-09-23 06:07:10,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=193993.33333333334, ans=0.1
2024-09-23 06:07:11,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=193993.33333333334, ans=0.5
2024-09-23 06:07:28,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=194040.0, ans=0.0
2024-09-23 06:08:00,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=194133.33333333334, ans=0.1
2024-09-23 06:08:11,671 INFO [train.py:1198] (1/4) Epoch 11, batch 2650, loss[loss=0.2024, ctc_loss=0.1344, cr_loss=0.34, over 16271.00 frames. ], tot_loss[loss=0.2441, ctc_loss=0.1687, cr_loss=0.3769, over 3352553.61 frames. ], batch size: 36, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:08:29,581 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.88 vs. limit=15.0
2024-09-23 06:08:33,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=194226.66666666666, ans=0.1
2024-09-23 06:08:38,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=194226.66666666666, ans=0.125
2024-09-23 06:08:40,198 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.11 vs. limit=10.0
2024-09-23 06:08:48,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.94 vs. limit=15.0
2024-09-23 06:08:56,016 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.17 vs. limit=15.0
2024-09-23 06:08:57,081 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.341e+02 1.529e+02 1.824e+02 3.038e+02, threshold=3.058e+02, percent-clipped=1.0
2024-09-23 06:08:57,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=194273.33333333334, ans=0.125
2024-09-23 06:08:58,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=194273.33333333334, ans=0.1
2024-09-23 06:09:05,193 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.97 vs. limit=15.0
2024-09-23 06:09:25,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=194366.66666666666, ans=0.125
2024-09-23 06:09:41,680 INFO [train.py:1198] (1/4) Epoch 11, batch 2700, loss[loss=0.2633, ctc_loss=0.1825, cr_loss=0.4037, over 17040.00 frames. ], tot_loss[loss=0.2446, ctc_loss=0.1691, cr_loss=0.3776, over 3352010.66 frames. ], batch size: 51, lr: 1.09e-02, grad_scale: 16.0
2024-09-23 06:10:07,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=194460.0, ans=0.0
2024-09-23 06:10:12,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=194506.66666666666, ans=0.04949747468305833
2024-09-23 06:10:17,164 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 06:10:19,005 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.77 vs. limit=15.0
2024-09-23 06:10:23,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=194506.66666666666, ans=0.125
2024-09-23 06:10:25,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=194506.66666666666, ans=0.025
2024-09-23 06:10:36,539 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.84 vs. limit=22.5
2024-09-23 06:10:37,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=194553.33333333334, ans=0.05
2024-09-23 06:11:01,390 INFO [train.py:1198] (1/4) Epoch 11, batch 2750, loss[loss=0.2482, ctc_loss=0.1686, cr_loss=0.3981, over 17045.00 frames. ], tot_loss[loss=0.2438, ctc_loss=0.1685, cr_loss=0.3762, over 3352789.94 frames. ], batch size: 52, lr: 1.09e-02, grad_scale: 16.0
2024-09-23 06:11:04,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=194646.66666666666, ans=0.125
2024-09-23 06:11:28,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=194693.33333333334, ans=0.125
2024-09-23 06:11:36,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=194740.0, ans=0.125
2024-09-23 06:11:36,509 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 06:11:36,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=194740.0, ans=0.2
2024-09-23 06:11:42,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=194740.0, ans=0.025
2024-09-23 06:11:44,296 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.275e+02 1.375e+02 1.599e+02 2.337e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-23 06:12:21,171 INFO [train.py:1198] (1/4) Epoch 11, batch 2800, loss[loss=0.2238, ctc_loss=0.1522, cr_loss=0.358, over 16969.00 frames. ], tot_loss[loss=0.2445, ctc_loss=0.169, cr_loss=0.3773, over 3354945.79 frames. ], batch size: 42, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:13:09,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=195020.0, ans=0.125
2024-09-23 06:13:42,581 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=2.61 vs. limit=15.0
2024-09-23 06:13:44,868 INFO [train.py:1198] (1/4) Epoch 11, batch 2850, loss[loss=0.2821, ctc_loss=0.1964, cr_loss=0.4287, over 17023.00 frames. ], tot_loss[loss=0.2445, ctc_loss=0.169, cr_loss=0.3774, over 3358694.23 frames. ], batch size: 53, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:14:15,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=195160.0, ans=0.125
2024-09-23 06:14:35,500 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.307e+02 1.410e+02 1.605e+02 2.111e+02, threshold=2.819e+02, percent-clipped=0.0
2024-09-23 06:14:40,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=195253.33333333334, ans=0.0
2024-09-23 06:14:53,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_abs, batch_count=195253.33333333334, ans=0.5
2024-09-23 06:15:11,916 INFO [train.py:1198] (1/4) Epoch 11, batch 2900, loss[loss=0.2569, ctc_loss=0.1786, cr_loss=0.3913, over 17014.00 frames. ], tot_loss[loss=0.2448, ctc_loss=0.1692, cr_loss=0.3775, over 3350846.47 frames. ], batch size: 56, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:15:29,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=195393.33333333334, ans=0.2
2024-09-23 06:15:46,216 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=8.46 vs. limit=15.0
2024-09-23 06:16:03,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=195486.66666666666, ans=0.0
2024-09-23 06:16:20,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=195533.33333333334, ans=0.125
2024-09-23 06:16:30,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=195580.0, ans=0.1
2024-09-23 06:16:30,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=195580.0, ans=0.2
2024-09-23 06:16:31,835 INFO [train.py:1198] (1/4) Epoch 11, batch 2950, loss[loss=0.2237, ctc_loss=0.154, cr_loss=0.3481, over 17224.00 frames. ], tot_loss[loss=0.2458, ctc_loss=0.1699, cr_loss=0.3792, over 3350356.62 frames. ], batch size: 47, lr: 1.09e-02, grad_scale: 32.0
2024-09-23 06:16:32,650 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.97 vs. limit=10.0
2024-09-23 06:16:36,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=195580.0, ans=0.125
2024-09-23 06:16:41,742 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 06:17:14,835 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.272e+02 1.364e+02 1.479e+02 2.031e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-23 06:17:18,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=195720.0, ans=0.125
2024-09-23 06:17:38,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=195766.66666666666, ans=0.125
2024-09-23 06:17:50,960 INFO [train.py:1198] (1/4) Epoch 11, batch 3000, loss[loss=0.2411, ctc_loss=0.166, cr_loss=0.3753, over 17309.00 frames. ], tot_loss[loss=0.2455, ctc_loss=0.1697, cr_loss=0.3791, over 3355223.71 frames. ], batch size: 46, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:17:50,960 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 06:18:06,136 INFO [train.py:1230] (1/4) Epoch 11, validation: loss=0.04835, ctc_loss=0.04835, cr_loss=7.412e-15, over 944034.00 frames. 
2024-09-23 06:18:06,137 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 06:18:08,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=195813.33333333334, ans=0.1
2024-09-23 06:18:14,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=195813.33333333334, ans=0.125
2024-09-23 06:18:34,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=195860.0, ans=0.125
2024-09-23 06:19:02,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=195953.33333333334, ans=0.125
2024-09-23 06:19:03,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=195953.33333333334, ans=0.125
2024-09-23 06:19:27,162 INFO [train.py:1198] (1/4) Epoch 11, batch 3050, loss[loss=0.2752, ctc_loss=0.1923, cr_loss=0.4144, over 16470.00 frames. ], tot_loss[loss=0.2454, ctc_loss=0.1697, cr_loss=0.3788, over 3355605.59 frames. ], batch size: 66, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:19:29,586 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.10 vs. limit=15.0
2024-09-23 06:19:32,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=196046.66666666666, ans=0.125
2024-09-23 06:20:01,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=196140.0, ans=0.2
2024-09-23 06:20:13,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=196140.0, ans=0.125
2024-09-23 06:20:14,517 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.336e+02 1.544e+02 1.822e+02 2.726e+02, threshold=3.088e+02, percent-clipped=0.0
2024-09-23 06:20:17,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=196186.66666666666, ans=0.125
2024-09-23 06:20:39,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=196233.33333333334, ans=0.125
2024-09-23 06:20:48,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.37 vs. limit=15.0
2024-09-23 06:20:52,700 INFO [train.py:1198] (1/4) Epoch 11, batch 3100, loss[loss=0.2415, ctc_loss=0.1666, cr_loss=0.3741, over 17090.00 frames. ], tot_loss[loss=0.2443, ctc_loss=0.1687, cr_loss=0.3779, over 3364664.78 frames. ], batch size: 49, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:21:14,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=196326.66666666666, ans=0.125
2024-09-23 06:21:15,355 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.20 vs. limit=15.0
2024-09-23 06:21:40,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2.whitening_limit, batch_count=196420.0, ans=15.0
2024-09-23 06:22:05,663 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=18.33 vs. limit=22.5
2024-09-23 06:22:11,264 INFO [train.py:1198] (1/4) Epoch 11, batch 3150, loss[loss=0.2939, ctc_loss=0.2097, cr_loss=0.4213, over 14834.00 frames. ], tot_loss[loss=0.2452, ctc_loss=0.1693, cr_loss=0.3793, over 3361305.11 frames. ], batch size: 89, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:22:39,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=196560.0, ans=0.2
2024-09-23 06:22:47,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=196606.66666666666, ans=0.0
2024-09-23 06:22:53,533 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.336e+02 1.469e+02 1.670e+02 2.912e+02, threshold=2.938e+02, percent-clipped=0.0
2024-09-23 06:23:29,551 INFO [train.py:1198] (1/4) Epoch 11, batch 3200, loss[loss=0.2555, ctc_loss=0.1752, cr_loss=0.4017, over 17236.00 frames. ], tot_loss[loss=0.2455, ctc_loss=0.1696, cr_loss=0.3792, over 3354031.18 frames. ], batch size: 55, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:24:24,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=196886.66666666666, ans=0.125
2024-09-23 06:24:24,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=196886.66666666666, ans=0.0
2024-09-23 06:24:35,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=196933.33333333334, ans=0.0
2024-09-23 06:24:47,353 INFO [train.py:1198] (1/4) Epoch 11, batch 3250, loss[loss=0.2589, ctc_loss=0.1778, cr_loss=0.4056, over 17157.00 frames. ], tot_loss[loss=0.2459, ctc_loss=0.1701, cr_loss=0.3787, over 3336409.49 frames. ], batch size: 45, lr: 1.08e-02, grad_scale: 16.0
2024-09-23 06:25:04,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=197026.66666666666, ans=0.125
2024-09-23 06:25:15,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=197026.66666666666, ans=0.2
2024-09-23 06:25:16,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=197073.33333333334, ans=0.0
2024-09-23 06:25:18,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=197073.33333333334, ans=0.125
2024-09-23 06:25:30,818 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.294e+02 1.384e+02 1.561e+02 5.237e+02, threshold=2.769e+02, percent-clipped=1.0
2024-09-23 06:25:32,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=197120.0, ans=0.2
2024-09-23 06:25:47,348 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=19.38 vs. limit=22.5
2024-09-23 06:25:58,380 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.89 vs. limit=6.0
2024-09-23 06:26:05,132 INFO [train.py:1198] (1/4) Epoch 11, batch 3300, loss[loss=0.2712, ctc_loss=0.1889, cr_loss=0.4115, over 17030.00 frames. ], tot_loss[loss=0.2445, ctc_loss=0.169, cr_loss=0.3772, over 3335033.34 frames. ], batch size: 56, lr: 1.08e-02, grad_scale: 16.0
2024-09-23 06:26:54,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=197353.33333333334, ans=0.0
2024-09-23 06:27:21,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=197446.66666666666, ans=0.0
2024-09-23 06:27:22,860 INFO [train.py:1198] (1/4) Epoch 11, batch 3350, loss[loss=0.2126, ctc_loss=0.1423, cr_loss=0.3517, over 17099.00 frames. ], tot_loss[loss=0.2444, ctc_loss=0.1689, cr_loss=0.3776, over 3344693.81 frames. ], batch size: 40, lr: 1.08e-02, grad_scale: 16.0
2024-09-23 06:27:27,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=197446.66666666666, ans=0.0
2024-09-23 06:27:29,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=197446.66666666666, ans=0.125
2024-09-23 06:27:31,502 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.27 vs. limit=12.0
2024-09-23 06:27:51,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=197493.33333333334, ans=0.0
2024-09-23 06:27:52,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=197540.0, ans=0.0
2024-09-23 06:28:06,586 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.310e+02 1.428e+02 1.594e+02 2.854e+02, threshold=2.856e+02, percent-clipped=1.0
2024-09-23 06:28:14,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=197586.66666666666, ans=0.1
2024-09-23 06:28:36,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=197633.33333333334, ans=0.125
2024-09-23 06:28:39,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=197680.0, ans=0.2
2024-09-23 06:28:41,094 INFO [train.py:1198] (1/4) Epoch 11, batch 3400, loss[loss=0.3054, ctc_loss=0.216, cr_loss=0.447, over 15009.00 frames. ], tot_loss[loss=0.2439, ctc_loss=0.1684, cr_loss=0.3774, over 3354138.31 frames. ], batch size: 90, lr: 1.08e-02, grad_scale: 16.0
2024-09-23 06:28:58,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=197726.66666666666, ans=0.2
2024-09-23 06:28:59,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=197726.66666666666, ans=0.1
2024-09-23 06:29:44,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=197866.66666666666, ans=0.125
2024-09-23 06:30:04,325 INFO [train.py:1198] (1/4) Epoch 11, batch 3450, loss[loss=0.2308, ctc_loss=0.1597, cr_loss=0.3554, over 17095.00 frames. ], tot_loss[loss=0.2431, ctc_loss=0.1677, cr_loss=0.3769, over 3362787.20 frames. ], batch size: 43, lr: 1.08e-02, grad_scale: 16.0
2024-09-23 06:30:06,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=197913.33333333334, ans=0.2
2024-09-23 06:30:09,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=197913.33333333334, ans=0.125
2024-09-23 06:30:20,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=197960.0, ans=0.05
2024-09-23 06:30:44,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.05 vs. limit=15.0
2024-09-23 06:30:49,692 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.318e+02 1.398e+02 1.528e+02 2.343e+02, threshold=2.795e+02, percent-clipped=0.0
2024-09-23 06:31:01,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=198053.33333333334, ans=0.0
2024-09-23 06:31:04,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=198053.33333333334, ans=0.09899494936611666
2024-09-23 06:31:20,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=198100.0, ans=0.2
2024-09-23 06:31:26,438 INFO [train.py:1198] (1/4) Epoch 11, batch 3500, loss[loss=0.2739, ctc_loss=0.1902, cr_loss=0.4185, over 15887.00 frames. ], tot_loss[loss=0.2433, ctc_loss=0.1677, cr_loss=0.378, over 3365215.86 frames. ], batch size: 74, lr: 1.08e-02, grad_scale: 16.0
2024-09-23 06:31:55,673 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.77 vs. limit=5.0
2024-09-23 06:32:33,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.max_positive, batch_count=198333.33333333334, ans=0.95
2024-09-23 06:32:41,847 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.91 vs. limit=15.0
2024-09-23 06:32:44,007 INFO [train.py:1198] (1/4) Epoch 11, batch 3550, loss[loss=0.227, ctc_loss=0.154, cr_loss=0.3651, over 17166.00 frames. ], tot_loss[loss=0.2442, ctc_loss=0.1683, cr_loss=0.3794, over 3374545.49 frames. ], batch size: 45, lr: 1.08e-02, grad_scale: 16.0
2024-09-23 06:32:47,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=198380.0, ans=0.0
2024-09-23 06:33:04,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=198426.66666666666, ans=0.1
2024-09-23 06:33:06,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=198426.66666666666, ans=0.125
2024-09-23 06:33:21,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=198473.33333333334, ans=0.1
2024-09-23 06:33:24,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=198473.33333333334, ans=0.125
2024-09-23 06:33:27,727 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.313e+02 1.389e+02 1.597e+02 2.419e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-23 06:33:31,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=198520.0, ans=0.5
2024-09-23 06:34:02,340 INFO [train.py:1198] (1/4) Epoch 11, batch 3600, loss[loss=0.2603, ctc_loss=0.18, cr_loss=0.4016, over 17300.00 frames. ], tot_loss[loss=0.2437, ctc_loss=0.1679, cr_loss=0.3788, over 3380119.35 frames. ], batch size: 46, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:34:42,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=198706.66666666666, ans=0.125
2024-09-23 06:35:12,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=198800.0, ans=0.125
2024-09-23 06:35:19,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=198846.66666666666, ans=0.125
2024-09-23 06:35:21,045 INFO [train.py:1198] (1/4) Epoch 11, batch 3650, loss[loss=0.2839, ctc_loss=0.1998, cr_loss=0.4207, over 15000.00 frames. ], tot_loss[loss=0.2455, ctc_loss=0.1694, cr_loss=0.3805, over 3364313.17 frames. ], batch size: 89, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:35:24,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=198846.66666666666, ans=0.07
2024-09-23 06:35:29,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=198846.66666666666, ans=0.125
2024-09-23 06:36:04,372 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.052e+02 1.307e+02 1.406e+02 1.514e+02 2.450e+02, threshold=2.812e+02, percent-clipped=0.0
2024-09-23 06:36:04,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=198940.0, ans=0.1
2024-09-23 06:36:04,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=198940.0, ans=0.07
2024-09-23 06:36:26,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.41 vs. limit=15.0
2024-09-23 06:36:39,291 INFO [train.py:1198] (1/4) Epoch 11, batch 3700, loss[loss=0.2757, ctc_loss=0.1925, cr_loss=0.4158, over 17017.00 frames. ], tot_loss[loss=0.2449, ctc_loss=0.1691, cr_loss=0.3792, over 3362333.81 frames. ], batch size: 51, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:36:39,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=199080.0, ans=0.0
2024-09-23 06:36:45,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=199080.0, ans=0.125
2024-09-23 06:37:26,679 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=15.05 vs. limit=22.5
2024-09-23 06:37:57,659 INFO [train.py:1198] (1/4) Epoch 11, batch 3750, loss[loss=0.2698, ctc_loss=0.1864, cr_loss=0.417, over 16902.00 frames. ], tot_loss[loss=0.2442, ctc_loss=0.1686, cr_loss=0.3781, over 3361897.54 frames. ], batch size: 58, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:38:28,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=199406.66666666666, ans=0.125
2024-09-23 06:38:32,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.97 vs. limit=6.0
2024-09-23 06:38:41,139 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.417e+02 1.573e+02 1.870e+02 3.069e+02, threshold=3.146e+02, percent-clipped=3.0
2024-09-23 06:39:16,468 INFO [train.py:1198] (1/4) Epoch 11, batch 3800, loss[loss=0.2506, ctc_loss=0.1757, cr_loss=0.3744, over 17028.00 frames. ], tot_loss[loss=0.2449, ctc_loss=0.1692, cr_loss=0.3785, over 3350044.66 frames. ], batch size: 52, lr: 1.08e-02, grad_scale: 32.0
2024-09-23 06:39:29,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=199546.66666666666, ans=0.125
2024-09-23 06:39:39,710 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.53 vs. limit=22.5
2024-09-23 06:40:34,878 INFO [train.py:1198] (1/4) Epoch 11, batch 3850, loss[loss=0.2578, ctc_loss=0.1766, cr_loss=0.4058, over 16913.00 frames. ], tot_loss[loss=0.251, ctc_loss=0.1744, cr_loss=0.3828, over 3261827.91 frames. ], batch size: 58, lr: 1.07e-02, grad_scale: 16.0
2024-09-23 06:40:38,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=199780.0, ans=0.04949747468305833
2024-09-23 06:40:51,083 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.34 vs. limit=12.0
2024-09-23 06:40:56,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=199826.66666666666, ans=0.0
2024-09-23 06:41:03,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=199873.33333333334, ans=0.2
2024-09-23 06:41:05,950 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.25 vs. limit=22.5
2024-09-23 06:41:18,611 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.500e+02 1.672e+02 1.857e+02 2.511e+02, threshold=3.343e+02, percent-clipped=0.0
2024-09-23 06:41:29,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=199920.0, ans=0.025
2024-09-23 06:42:36,534 INFO [train.py:1198] (1/4) Epoch 12, batch 0, loss[loss=0.2114, ctc_loss=0.1446, cr_loss=0.3337, over 17118.00 frames. ], tot_loss[loss=0.2114, ctc_loss=0.1446, cr_loss=0.3337, over 17118.00 frames. ], batch size: 40, lr: 1.03e-02, grad_scale: 32.0
2024-09-23 06:42:36,535 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 06:42:43,774 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.4.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([3.0619, 3.8935, 3.8296, 3.7974], device='cuda:1')
2024-09-23 06:42:52,084 INFO [train.py:1230] (1/4) Epoch 12, validation: loss=0.0478, ctc_loss=0.0478, cr_loss=7.52e-15, over 944034.00 frames. 
2024-09-23 06:42:52,085 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 06:43:10,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=200041.33333333334, ans=0.125
2024-09-23 06:43:22,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=200088.0, ans=0.1
2024-09-23 06:43:40,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=200134.66666666666, ans=0.125
2024-09-23 06:43:46,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=200134.66666666666, ans=0.125
2024-09-23 06:44:06,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=200181.33333333334, ans=0.125
2024-09-23 06:44:10,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=200228.0, ans=0.125
2024-09-23 06:44:11,654 INFO [train.py:1198] (1/4) Epoch 12, batch 50, loss[loss=0.1748, ctc_loss=0.1162, cr_loss=0.293, over 16746.00 frames. ], tot_loss[loss=0.2443, ctc_loss=0.1685, cr_loss=0.3788, over 750788.13 frames. ], batch size: 37, lr: 1.03e-02, grad_scale: 32.0
2024-09-23 06:44:43,230 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.76 vs. limit=15.0
2024-09-23 06:45:03,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=200321.33333333334, ans=10.0
2024-09-23 06:45:08,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=200368.0, ans=0.0
2024-09-23 06:45:12,534 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.318e+02 1.438e+02 1.597e+02 2.419e+02, threshold=2.876e+02, percent-clipped=0.0
2024-09-23 06:45:27,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=200414.66666666666, ans=0.125
2024-09-23 06:45:40,750 INFO [train.py:1198] (1/4) Epoch 12, batch 100, loss[loss=0.2527, ctc_loss=0.176, cr_loss=0.3838, over 17218.00 frames. ], tot_loss[loss=0.2445, ctc_loss=0.1688, cr_loss=0.3784, over 1311059.19 frames. ], batch size: 50, lr: 1.03e-02, grad_scale: 16.0
2024-09-23 06:46:59,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=200694.66666666666, ans=0.0
2024-09-23 06:47:00,676 INFO [train.py:1198] (1/4) Epoch 12, batch 150, loss[loss=0.2352, ctc_loss=0.1613, cr_loss=0.3695, over 17016.00 frames. ], tot_loss[loss=0.2455, ctc_loss=0.1696, cr_loss=0.3799, over 1759723.66 frames. ], batch size: 44, lr: 1.03e-02, grad_scale: 16.0
2024-09-23 06:47:04,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=200694.66666666666, ans=0.2
2024-09-23 06:47:32,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=200788.0, ans=0.025
2024-09-23 06:47:40,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=200788.0, ans=0.1
2024-09-23 06:47:54,507 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.306e+02 1.461e+02 1.660e+02 2.321e+02, threshold=2.923e+02, percent-clipped=0.0
2024-09-23 06:48:19,935 INFO [train.py:1198] (1/4) Epoch 12, batch 200, loss[loss=0.3147, ctc_loss=0.2313, cr_loss=0.4169, over 11344.00 frames. ], tot_loss[loss=0.2456, ctc_loss=0.1695, cr_loss=0.3806, over 2112010.70 frames. ], batch size: 123, lr: 1.03e-02, grad_scale: 16.0
2024-09-23 06:48:32,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=200928.0, ans=0.2
2024-09-23 06:48:48,033 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=7.16 vs. limit=15.0
2024-09-23 06:49:09,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=201068.0, ans=0.2
2024-09-23 06:49:09,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=201068.0, ans=0.125
2024-09-23 06:49:20,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=201068.0, ans=0.125
2024-09-23 06:49:43,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=201114.66666666666, ans=0.1
2024-09-23 06:49:47,801 INFO [train.py:1198] (1/4) Epoch 12, batch 250, loss[loss=0.2425, ctc_loss=0.1657, cr_loss=0.3839, over 17030.00 frames. ], tot_loss[loss=0.2447, ctc_loss=0.1687, cr_loss=0.3802, over 2395147.80 frames. ], batch size: 44, lr: 1.03e-02, grad_scale: 16.0
2024-09-23 06:49:48,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.min_positive, batch_count=201161.33333333334, ans=0.025
2024-09-23 06:49:59,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=201161.33333333334, ans=0.125
2024-09-23 06:50:01,555 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=5.35 vs. limit=6.0
2024-09-23 06:50:43,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=201301.33333333334, ans=0.025
2024-09-23 06:50:45,029 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.293e+02 1.367e+02 1.506e+02 2.268e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-23 06:50:46,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=201301.33333333334, ans=0.1
2024-09-23 06:50:50,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=201301.33333333334, ans=0.0
2024-09-23 06:50:58,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=201348.0, ans=0.125
2024-09-23 06:51:10,810 INFO [train.py:1198] (1/4) Epoch 12, batch 300, loss[loss=0.2705, ctc_loss=0.1874, cr_loss=0.4156, over 16736.00 frames. ], tot_loss[loss=0.2433, ctc_loss=0.1677, cr_loss=0.3781, over 2610358.65 frames. ], batch size: 61, lr: 1.03e-02, grad_scale: 16.0
2024-09-23 06:51:50,373 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=7.27 vs. limit=15.0
2024-09-23 06:52:00,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer_na.min_abs, batch_count=201534.66666666666, ans=0.02
2024-09-23 06:52:02,638 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.25 vs. limit=15.0
2024-09-23 06:52:30,549 INFO [train.py:1198] (1/4) Epoch 12, batch 350, loss[loss=0.2427, ctc_loss=0.1639, cr_loss=0.3942, over 16637.00 frames. ], tot_loss[loss=0.244, ctc_loss=0.1682, cr_loss=0.3791, over 2769880.90 frames. ], batch size: 37, lr: 1.02e-02, grad_scale: 16.0
2024-09-23 06:52:52,181 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.17 vs. limit=15.0
2024-09-23 06:52:58,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=201674.66666666666, ans=0.0
2024-09-23 06:53:08,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=201721.33333333334, ans=0.125
2024-09-23 06:53:13,564 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.45 vs. limit=6.0
2024-09-23 06:53:20,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=201768.0, ans=0.125
2024-09-23 06:53:21,259 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.79 vs. limit=10.0
2024-09-23 06:53:25,561 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.271e+02 1.388e+02 1.541e+02 2.258e+02, threshold=2.777e+02, percent-clipped=0.0
2024-09-23 06:53:27,518 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 06:53:39,119 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.16 vs. limit=15.0
2024-09-23 06:53:51,107 INFO [train.py:1198] (1/4) Epoch 12, batch 400, loss[loss=0.2552, ctc_loss=0.1799, cr_loss=0.3766, over 17303.00 frames. ], tot_loss[loss=0.2442, ctc_loss=0.1682, cr_loss=0.38, over 2903195.13 frames. ], batch size: 49, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 06:54:20,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=201908.0, ans=0.0
2024-09-23 06:54:36,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=201954.66666666666, ans=0.2
2024-09-23 06:55:14,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=202048.0, ans=0.125
2024-09-23 06:55:18,842 INFO [train.py:1198] (1/4) Epoch 12, batch 450, loss[loss=0.2205, ctc_loss=0.1463, cr_loss=0.371, over 17187.00 frames. ], tot_loss[loss=0.2445, ctc_loss=0.1684, cr_loss=0.3804, over 3013387.00 frames. ], batch size: 41, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 06:55:19,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.23 vs. limit=15.0
2024-09-23 06:55:40,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=202141.33333333334, ans=0.1
2024-09-23 06:56:08,446 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.05 vs. limit=10.0
2024-09-23 06:56:15,455 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.330e+02 1.519e+02 1.735e+02 3.092e+02, threshold=3.038e+02, percent-clipped=2.0
2024-09-23 06:56:22,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=202234.66666666666, ans=0.2
2024-09-23 06:56:25,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=202281.33333333334, ans=0.125
2024-09-23 06:56:33,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=9.16 vs. limit=12.0
2024-09-23 06:56:40,845 INFO [train.py:1198] (1/4) Epoch 12, batch 500, loss[loss=0.2759, ctc_loss=0.1942, cr_loss=0.4089, over 17018.00 frames. ], tot_loss[loss=0.2441, ctc_loss=0.1682, cr_loss=0.3794, over 3089034.11 frames. ], batch size: 52, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 06:57:24,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=202421.33333333334, ans=0.0
2024-09-23 06:57:43,809 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.78 vs. limit=15.0
2024-09-23 06:58:00,476 INFO [train.py:1198] (1/4) Epoch 12, batch 550, loss[loss=0.24, ctc_loss=0.1647, cr_loss=0.3763, over 17308.00 frames. ], tot_loss[loss=0.2425, ctc_loss=0.167, cr_loss=0.3775, over 3159715.53 frames. ], batch size: 49, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 06:58:10,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=202561.33333333334, ans=0.125
2024-09-23 06:58:16,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=202608.0, ans=0.125
2024-09-23 06:58:19,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=202608.0, ans=0.0
2024-09-23 06:58:38,769 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 06:58:49,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=202701.33333333334, ans=0.0
2024-09-23 06:58:54,272 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.321e+02 1.421e+02 1.522e+02 2.085e+02, threshold=2.842e+02, percent-clipped=0.0
2024-09-23 06:58:54,550 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 06:59:02,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=202748.0, ans=0.0
2024-09-23 06:59:10,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=202748.0, ans=0.0
2024-09-23 06:59:10,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=202748.0, ans=0.2
2024-09-23 06:59:22,365 INFO [train.py:1198] (1/4) Epoch 12, batch 600, loss[loss=0.2713, ctc_loss=0.1923, cr_loss=0.3947, over 15104.00 frames. ], tot_loss[loss=0.2409, ctc_loss=0.1659, cr_loss=0.3755, over 3207362.42 frames. ], batch size: 90, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 06:59:22,901 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.47 vs. limit=15.0
2024-09-23 06:59:44,684 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.67 vs. limit=15.0
2024-09-23 07:00:03,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=202888.0, ans=0.125
2024-09-23 07:00:03,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=202888.0, ans=0.0
2024-09-23 07:00:19,532 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.49 vs. limit=15.0
2024-09-23 07:00:21,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=202934.66666666666, ans=0.0
2024-09-23 07:00:30,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=202981.33333333334, ans=0.125
2024-09-23 07:00:35,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=202981.33333333334, ans=0.125
2024-09-23 07:00:47,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=202981.33333333334, ans=0.125
2024-09-23 07:00:50,043 INFO [train.py:1198] (1/4) Epoch 12, batch 650, loss[loss=0.2292, ctc_loss=0.1559, cr_loss=0.3665, over 16942.00 frames. ], tot_loss[loss=0.2428, ctc_loss=0.1674, cr_loss=0.3771, over 3233137.03 frames. ], batch size: 42, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:01:22,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=203121.33333333334, ans=0.125
2024-09-23 07:01:44,224 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.271e+02 1.387e+02 1.567e+02 2.285e+02, threshold=2.773e+02, percent-clipped=0.0
2024-09-23 07:01:50,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=203168.0, ans=0.2
2024-09-23 07:02:09,751 INFO [train.py:1198] (1/4) Epoch 12, batch 700, loss[loss=0.2209, ctc_loss=0.1514, cr_loss=0.3472, over 17095.00 frames. ], tot_loss[loss=0.2426, ctc_loss=0.1672, cr_loss=0.3769, over 3255702.57 frames. ], batch size: 43, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:02:11,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=203261.33333333334, ans=0.125
2024-09-23 07:02:35,906 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:02:58,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=203401.33333333334, ans=0.125
2024-09-23 07:02:59,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=203401.33333333334, ans=0.125
2024-09-23 07:03:29,845 INFO [train.py:1198] (1/4) Epoch 12, batch 750, loss[loss=0.2561, ctc_loss=0.1787, cr_loss=0.3871, over 17035.00 frames. ], tot_loss[loss=0.2426, ctc_loss=0.1671, cr_loss=0.3775, over 3287528.57 frames. ], batch size: 56, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:03:30,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=203494.66666666666, ans=0.125
2024-09-23 07:03:31,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=203494.66666666666, ans=0.125
2024-09-23 07:03:36,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=203494.66666666666, ans=0.0
2024-09-23 07:03:49,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=203541.33333333334, ans=0.2
2024-09-23 07:04:14,980 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.12 vs. limit=15.0
2024-09-23 07:04:19,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=203634.66666666666, ans=0.025
2024-09-23 07:04:29,347 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.298e+02 1.398e+02 1.549e+02 2.740e+02, threshold=2.796e+02, percent-clipped=0.0
2024-09-23 07:04:46,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=203681.33333333334, ans=0.125
2024-09-23 07:04:57,396 INFO [train.py:1198] (1/4) Epoch 12, batch 800, loss[loss=0.1963, ctc_loss=0.1341, cr_loss=0.3106, over 17105.00 frames. ], tot_loss[loss=0.2433, ctc_loss=0.1676, cr_loss=0.3789, over 3304145.31 frames. ], batch size: 40, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:06:03,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=203914.66666666666, ans=0.1
2024-09-23 07:06:16,250 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=10.71 vs. limit=15.0
2024-09-23 07:06:18,738 INFO [train.py:1198] (1/4) Epoch 12, batch 850, loss[loss=0.2631, ctc_loss=0.1819, cr_loss=0.406, over 16164.00 frames. ], tot_loss[loss=0.2442, ctc_loss=0.1683, cr_loss=0.3797, over 3322790.41 frames. ], batch size: 74, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:06:36,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=204008.0, ans=0.125
2024-09-23 07:06:36,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=204008.0, ans=0.125
2024-09-23 07:07:05,601 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.86 vs. limit=6.0
2024-09-23 07:07:12,568 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.288e+02 1.402e+02 1.556e+02 2.231e+02, threshold=2.804e+02, percent-clipped=0.0
2024-09-23 07:07:38,074 INFO [train.py:1198] (1/4) Epoch 12, batch 900, loss[loss=0.2096, ctc_loss=0.1424, cr_loss=0.3361, over 17084.00 frames. ], tot_loss[loss=0.2443, ctc_loss=0.1682, cr_loss=0.3802, over 3335579.48 frames. ], batch size: 43, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:07:55,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=204241.33333333334, ans=0.1
2024-09-23 07:08:19,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=204288.0, ans=0.125
2024-09-23 07:08:21,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=204288.0, ans=0.0
2024-09-23 07:08:32,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=204334.66666666666, ans=0.1
2024-09-23 07:08:33,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=204334.66666666666, ans=0.125
2024-09-23 07:08:46,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=204381.33333333334, ans=0.125
2024-09-23 07:08:57,660 INFO [train.py:1198] (1/4) Epoch 12, batch 950, loss[loss=0.2672, ctc_loss=0.185, cr_loss=0.4106, over 17132.00 frames. ], tot_loss[loss=0.2441, ctc_loss=0.1682, cr_loss=0.3798, over 3331838.23 frames. ], batch size: 48, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:09:05,224 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.49 vs. limit=15.0
2024-09-23 07:09:19,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=204474.66666666666, ans=0.125
2024-09-23 07:09:28,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=204474.66666666666, ans=0.125
2024-09-23 07:09:37,777 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.39 vs. limit=15.0
2024-09-23 07:09:42,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=204521.33333333334, ans=0.0
2024-09-23 07:09:44,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=204521.33333333334, ans=22.5
2024-09-23 07:09:45,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=204521.33333333334, ans=0.0
2024-09-23 07:09:49,063 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.05 vs. limit=6.0
2024-09-23 07:09:50,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=204521.33333333334, ans=0.125
2024-09-23 07:09:52,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.15 vs. limit=12.0
2024-09-23 07:09:56,854 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:09:59,531 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.295e+02 1.392e+02 1.548e+02 2.202e+02, threshold=2.785e+02, percent-clipped=0.0
2024-09-23 07:10:01,739 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.22 vs. limit=15.0
2024-09-23 07:10:03,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=204568.0, ans=0.0
2024-09-23 07:10:04,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=204568.0, ans=0.125
2024-09-23 07:10:20,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=204614.66666666666, ans=0.2
2024-09-23 07:10:25,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=204614.66666666666, ans=0.0
2024-09-23 07:10:28,003 INFO [train.py:1198] (1/4) Epoch 12, batch 1000, loss[loss=0.2237, ctc_loss=0.1524, cr_loss=0.3567, over 17078.00 frames. ], tot_loss[loss=0.2425, ctc_loss=0.167, cr_loss=0.3776, over 3341328.25 frames. ], batch size: 40, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:10:28,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=204661.33333333334, ans=0.125
2024-09-23 07:10:29,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=204661.33333333334, ans=0.125
2024-09-23 07:10:31,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=204661.33333333334, ans=0.0
2024-09-23 07:10:59,850 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=204754.66666666666, ans=0.2
2024-09-23 07:11:01,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=204754.66666666666, ans=0.2
2024-09-23 07:11:14,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=204801.33333333334, ans=0.1
2024-09-23 07:11:47,281 INFO [train.py:1198] (1/4) Epoch 12, batch 1050, loss[loss=0.2143, ctc_loss=0.1456, cr_loss=0.3433, over 17183.00 frames. ], tot_loss[loss=0.2431, ctc_loss=0.1674, cr_loss=0.3785, over 3341357.32 frames. ], batch size: 41, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:11:54,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=204894.66666666666, ans=0.0
2024-09-23 07:12:08,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=204941.33333333334, ans=0.0
2024-09-23 07:12:15,094 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.68 vs. limit=15.0
2024-09-23 07:12:28,149 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.69 vs. limit=15.0
2024-09-23 07:12:41,478 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.299e+02 1.476e+02 1.734e+02 2.912e+02, threshold=2.951e+02, percent-clipped=1.0
2024-09-23 07:12:41,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=205034.66666666666, ans=0.125
2024-09-23 07:13:07,027 INFO [train.py:1198] (1/4) Epoch 12, batch 1100, loss[loss=0.2436, ctc_loss=0.1683, cr_loss=0.3762, over 17143.00 frames. ], tot_loss[loss=0.2432, ctc_loss=0.1675, cr_loss=0.3787, over 3344177.34 frames. ], batch size: 48, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:14:36,514 INFO [train.py:1198] (1/4) Epoch 12, batch 1150, loss[loss=0.2586, ctc_loss=0.1797, cr_loss=0.3945, over 17015.00 frames. ], tot_loss[loss=0.2426, ctc_loss=0.1669, cr_loss=0.3786, over 3351134.98 frames. ], batch size: 53, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:14:49,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=205361.33333333334, ans=0.1
2024-09-23 07:15:03,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=205408.0, ans=0.2
2024-09-23 07:15:08,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=205454.66666666666, ans=0.0
2024-09-23 07:15:11,399 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:15:14,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=205454.66666666666, ans=0.125
2024-09-23 07:15:32,629 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.321e+02 1.590e+02 1.805e+02 2.687e+02, threshold=3.179e+02, percent-clipped=0.0
2024-09-23 07:15:39,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=205501.33333333334, ans=0.125
2024-09-23 07:15:45,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=205548.0, ans=0.125
2024-09-23 07:15:55,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=205548.0, ans=0.0
2024-09-23 07:15:58,032 INFO [train.py:1198] (1/4) Epoch 12, batch 1200, loss[loss=0.1902, ctc_loss=0.128, cr_loss=0.3112, over 17061.00 frames. ], tot_loss[loss=0.2423, ctc_loss=0.1666, cr_loss=0.3786, over 3357342.40 frames. ], batch size: 39, lr: 1.02e-02, grad_scale: 32.0
2024-09-23 07:16:27,003 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:17:02,125 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=8.77 vs. limit=15.0
2024-09-23 07:17:17,301 INFO [train.py:1198] (1/4) Epoch 12, batch 1250, loss[loss=0.2336, ctc_loss=0.161, cr_loss=0.3629, over 17292.00 frames. ], tot_loss[loss=0.243, ctc_loss=0.1673, cr_loss=0.3786, over 3348769.49 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:17:47,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=205921.33333333334, ans=0.09899494936611666
2024-09-23 07:17:51,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=205921.33333333334, ans=0.125
2024-09-23 07:17:55,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=205921.33333333334, ans=0.0
2024-09-23 07:18:12,962 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.281e+02 1.384e+02 1.564e+02 2.899e+02, threshold=2.767e+02, percent-clipped=0.0
2024-09-23 07:18:29,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=206014.66666666666, ans=0.125
2024-09-23 07:18:32,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=206014.66666666666, ans=0.125
2024-09-23 07:18:36,853 INFO [train.py:1198] (1/4) Epoch 12, batch 1300, loss[loss=0.2727, ctc_loss=0.1891, cr_loss=0.4176, over 15888.00 frames. ], tot_loss[loss=0.242, ctc_loss=0.1666, cr_loss=0.377, over 3360576.61 frames. ], batch size: 74, lr: 1.01e-02, grad_scale: 16.0
2024-09-23 07:18:39,116 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.47 vs. limit=15.0
2024-09-23 07:18:43,721 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.94 vs. limit=22.5
2024-09-23 07:19:09,567 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.07 vs. limit=10.0
2024-09-23 07:19:16,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=206154.66666666666, ans=0.0
2024-09-23 07:19:29,454 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.59 vs. limit=15.0
2024-09-23 07:19:49,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=206248.0, ans=0.125
2024-09-23 07:19:54,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=206248.0, ans=0.125
2024-09-23 07:19:56,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=206248.0, ans=0.125
2024-09-23 07:20:03,898 INFO [train.py:1198] (1/4) Epoch 12, batch 1350, loss[loss=0.2696, ctc_loss=0.1879, cr_loss=0.4084, over 17307.00 frames. ], tot_loss[loss=0.2429, ctc_loss=0.1673, cr_loss=0.3782, over 3363448.45 frames. ], batch size: 51, lr: 1.01e-02, grad_scale: 16.0
2024-09-23 07:20:19,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=206294.66666666666, ans=0.0
2024-09-23 07:20:35,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=206341.33333333334, ans=0.125
2024-09-23 07:20:41,966 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=15.34 vs. limit=22.5
2024-09-23 07:20:45,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=206388.0, ans=0.2
2024-09-23 07:21:02,411 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.350e+02 1.483e+02 1.734e+02 2.832e+02, threshold=2.966e+02, percent-clipped=2.0
2024-09-23 07:21:23,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=206481.33333333334, ans=0.125
2024-09-23 07:21:26,076 INFO [train.py:1198] (1/4) Epoch 12, batch 1400, loss[loss=0.2302, ctc_loss=0.1569, cr_loss=0.3664, over 17002.00 frames. ], tot_loss[loss=0.242, ctc_loss=0.1665, cr_loss=0.3773, over 3367548.33 frames. ], batch size: 51, lr: 1.01e-02, grad_scale: 16.0
2024-09-23 07:21:40,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=206574.66666666666, ans=0.025
2024-09-23 07:21:47,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=206574.66666666666, ans=0.0
2024-09-23 07:22:01,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=206621.33333333334, ans=0.125
2024-09-23 07:22:02,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten.whitening_limit, batch_count=206621.33333333334, ans=15.0
2024-09-23 07:22:14,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=206668.0, ans=0.2
2024-09-23 07:22:23,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=206668.0, ans=0.125
2024-09-23 07:22:24,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=206668.0, ans=0.025
2024-09-23 07:22:33,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=206714.66666666666, ans=0.0
2024-09-23 07:22:46,290 INFO [train.py:1198] (1/4) Epoch 12, batch 1450, loss[loss=0.2053, ctc_loss=0.1407, cr_loss=0.3229, over 16948.00 frames. ], tot_loss[loss=0.2424, ctc_loss=0.1669, cr_loss=0.3775, over 3362828.91 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 16.0
2024-09-23 07:23:12,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=206808.0, ans=0.125
2024-09-23 07:23:19,954 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:23:31,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=206854.66666666666, ans=0.05
2024-09-23 07:23:39,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=206901.33333333334, ans=0.1
2024-09-23 07:23:42,178 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.321e+02 1.435e+02 1.538e+02 2.142e+02, threshold=2.870e+02, percent-clipped=0.0
2024-09-23 07:24:10,783 INFO [train.py:1198] (1/4) Epoch 12, batch 1500, loss[loss=0.2501, ctc_loss=0.1735, cr_loss=0.3826, over 17285.00 frames. ], tot_loss[loss=0.2422, ctc_loss=0.1667, cr_loss=0.3773, over 3362175.91 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 16.0
2024-09-23 07:24:12,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=206994.66666666666, ans=0.125
2024-09-23 07:24:26,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=206994.66666666666, ans=0.2
2024-09-23 07:24:49,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=207088.0, ans=0.2
2024-09-23 07:24:50,821 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.72 vs. limit=15.0
2024-09-23 07:25:24,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=207181.33333333334, ans=0.125
2024-09-23 07:25:35,346 INFO [train.py:1198] (1/4) Epoch 12, batch 1550, loss[loss=0.2447, ctc_loss=0.1686, cr_loss=0.3802, over 17224.00 frames. ], tot_loss[loss=0.2418, ctc_loss=0.1664, cr_loss=0.3773, over 3368598.57 frames. ], batch size: 50, lr: 1.01e-02, grad_scale: 16.0
2024-09-23 07:25:35,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=207228.0, ans=0.04949747468305833
2024-09-23 07:25:56,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=207274.66666666666, ans=0.0
2024-09-23 07:25:57,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=207274.66666666666, ans=0.0
2024-09-23 07:25:58,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=207274.66666666666, ans=0.09899494936611666
2024-09-23 07:26:31,711 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.334e+02 1.473e+02 1.633e+02 2.267e+02, threshold=2.947e+02, percent-clipped=0.0
2024-09-23 07:26:32,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=207368.0, ans=0.0
2024-09-23 07:26:46,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=207414.66666666666, ans=0.125
2024-09-23 07:26:50,599 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=6.85 vs. limit=15.0
2024-09-23 07:26:55,792 INFO [train.py:1198] (1/4) Epoch 12, batch 1600, loss[loss=0.2096, ctc_loss=0.141, cr_loss=0.3432, over 16671.00 frames. ], tot_loss[loss=0.2429, ctc_loss=0.1672, cr_loss=0.3785, over 3348258.00 frames. ], batch size: 37, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:27:30,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=11.39 vs. limit=15.0
2024-09-23 07:28:00,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=207648.0, ans=0.125
2024-09-23 07:28:15,592 INFO [train.py:1198] (1/4) Epoch 12, batch 1650, loss[loss=0.2008, ctc_loss=0.1356, cr_loss=0.3259, over 16993.00 frames. ], tot_loss[loss=0.2417, ctc_loss=0.1663, cr_loss=0.3768, over 3350782.38 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:28:17,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=207694.66666666666, ans=0.125
2024-09-23 07:28:44,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=207741.33333333334, ans=0.5
2024-09-23 07:28:46,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=207788.0, ans=0.1
2024-09-23 07:28:58,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=207788.0, ans=0.125
2024-09-23 07:29:13,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=207834.66666666666, ans=0.125
2024-09-23 07:29:16,235 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.260e+02 1.353e+02 1.462e+02 2.073e+02, threshold=2.705e+02, percent-clipped=0.0
2024-09-23 07:29:42,766 INFO [train.py:1198] (1/4) Epoch 12, batch 1700, loss[loss=0.3141, ctc_loss=0.2296, cr_loss=0.4227, over 11706.00 frames. ], tot_loss[loss=0.2407, ctc_loss=0.1655, cr_loss=0.376, over 3353195.95 frames. ], batch size: 123, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:30:02,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=207974.66666666666, ans=0.0
2024-09-23 07:30:19,439 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:30:47,137 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.21 vs. limit=15.0
2024-09-23 07:31:04,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=208161.33333333334, ans=0.04949747468305833
2024-09-23 07:31:06,203 INFO [train.py:1198] (1/4) Epoch 12, batch 1750, loss[loss=0.2244, ctc_loss=0.156, cr_loss=0.3422, over 17166.00 frames. ], tot_loss[loss=0.2412, ctc_loss=0.1659, cr_loss=0.3763, over 3340008.48 frames. ], batch size: 41, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:31:57,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=208301.33333333334, ans=0.2
2024-09-23 07:32:02,198 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.329e+02 1.442e+02 1.637e+02 2.396e+02, threshold=2.884e+02, percent-clipped=0.0
2024-09-23 07:32:16,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=208348.0, ans=0.1
2024-09-23 07:32:25,983 INFO [train.py:1198] (1/4) Epoch 12, batch 1800, loss[loss=0.2406, ctc_loss=0.1657, cr_loss=0.3744, over 17326.00 frames. ], tot_loss[loss=0.2413, ctc_loss=0.1658, cr_loss=0.3772, over 3348740.71 frames. ], batch size: 51, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:32:26,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=208394.66666666666, ans=0.125
2024-09-23 07:32:46,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=208441.33333333334, ans=0.1
2024-09-23 07:33:23,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=208534.66666666666, ans=0.05
2024-09-23 07:33:41,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.out_whiten.whitening_limit, batch_count=208581.33333333334, ans=8.0
2024-09-23 07:33:45,651 INFO [train.py:1198] (1/4) Epoch 12, batch 1850, loss[loss=0.2097, ctc_loss=0.1424, cr_loss=0.3369, over 17259.00 frames. ], tot_loss[loss=0.2404, ctc_loss=0.1651, cr_loss=0.3766, over 3355843.95 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:34:48,401 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=7.28 vs. limit=15.0
2024-09-23 07:34:48,756 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.290e+02 1.399e+02 1.526e+02 2.337e+02, threshold=2.797e+02, percent-clipped=0.0
2024-09-23 07:34:55,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=208814.66666666666, ans=0.125
2024-09-23 07:34:55,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=208814.66666666666, ans=0.125
2024-09-23 07:34:58,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=208814.66666666666, ans=0.125
2024-09-23 07:35:15,124 INFO [train.py:1198] (1/4) Epoch 12, batch 1900, loss[loss=0.2109, ctc_loss=0.1425, cr_loss=0.3422, over 17268.00 frames. ], tot_loss[loss=0.2397, ctc_loss=0.1646, cr_loss=0.3756, over 3363780.62 frames. ], batch size: 44, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:35:15,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=208861.33333333334, ans=15.0
2024-09-23 07:35:24,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=208861.33333333334, ans=0.125
2024-09-23 07:35:29,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=208908.0, ans=0.125
2024-09-23 07:35:36,212 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.45 vs. limit=15.0
2024-09-23 07:35:56,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=208954.66666666666, ans=0.125
2024-09-23 07:35:59,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=208954.66666666666, ans=0.125
2024-09-23 07:36:01,467 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.00 vs. limit=15.0
2024-09-23 07:36:16,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.63 vs. limit=15.0
2024-09-23 07:36:19,397 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.28 vs. limit=22.5
2024-09-23 07:36:25,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=209048.0, ans=0.2
2024-09-23 07:36:31,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.05 vs. limit=10.0
2024-09-23 07:36:34,860 INFO [train.py:1198] (1/4) Epoch 12, batch 1950, loss[loss=0.205, ctc_loss=0.1416, cr_loss=0.3174, over 17213.00 frames. ], tot_loss[loss=0.2413, ctc_loss=0.1659, cr_loss=0.377, over 3360459.46 frames. ], batch size: 41, lr: 1.01e-02, grad_scale: 16.0
2024-09-23 07:36:54,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=209141.33333333334, ans=0.125
2024-09-23 07:37:08,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=209188.0, ans=0.125
2024-09-23 07:37:31,804 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.280e+02 1.393e+02 1.533e+02 3.563e+02, threshold=2.786e+02, percent-clipped=1.0
2024-09-23 07:37:32,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=209234.66666666666, ans=0.125
2024-09-23 07:37:35,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=209234.66666666666, ans=0.125
2024-09-23 07:37:54,115 INFO [train.py:1198] (1/4) Epoch 12, batch 2000, loss[loss=0.2356, ctc_loss=0.1586, cr_loss=0.3851, over 17178.00 frames. ], tot_loss[loss=0.2418, ctc_loss=0.1663, cr_loss=0.3776, over 3360125.81 frames. ], batch size: 45, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:37:57,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=209328.0, ans=0.125
2024-09-23 07:38:02,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=209328.0, ans=0.1
2024-09-23 07:38:04,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=209328.0, ans=0.04949747468305833
2024-09-23 07:38:30,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=209421.33333333334, ans=0.0
2024-09-23 07:38:44,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=209468.0, ans=0.0
2024-09-23 07:38:44,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.75 vs. limit=15.0
2024-09-23 07:39:03,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=209514.66666666666, ans=0.125
2024-09-23 07:39:09,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=209514.66666666666, ans=0.125
2024-09-23 07:39:21,410 INFO [train.py:1198] (1/4) Epoch 12, batch 2050, loss[loss=0.2504, ctc_loss=0.1726, cr_loss=0.3891, over 17297.00 frames. ], tot_loss[loss=0.2422, ctc_loss=0.1667, cr_loss=0.3778, over 3351927.51 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:40:05,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=209654.66666666666, ans=0.5
2024-09-23 07:40:10,758 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.03 vs. limit=15.0
2024-09-23 07:40:21,041 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.313e+02 1.485e+02 1.661e+02 2.450e+02, threshold=2.969e+02, percent-clipped=0.0
2024-09-23 07:40:24,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=209701.33333333334, ans=0.125
2024-09-23 07:40:29,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=209748.0, ans=0.2
2024-09-23 07:40:39,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=209748.0, ans=0.0
2024-09-23 07:40:43,461 INFO [train.py:1198] (1/4) Epoch 12, batch 2100, loss[loss=0.214, ctc_loss=0.1434, cr_loss=0.3528, over 17155.00 frames. ], tot_loss[loss=0.2421, ctc_loss=0.1664, cr_loss=0.3786, over 3358226.46 frames. ], batch size: 45, lr: 1.01e-02, grad_scale: 32.0
2024-09-23 07:40:45,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=209794.66666666666, ans=0.025
2024-09-23 07:40:48,788 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.89 vs. limit=15.0
2024-09-23 07:40:58,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=209841.33333333334, ans=0.0
2024-09-23 07:41:05,377 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=9.29 vs. limit=22.5
2024-09-23 07:41:25,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=209888.0, ans=0.0
2024-09-23 07:41:32,028 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=17.69 vs. limit=22.5
2024-09-23 07:41:43,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten.whitening_limit, batch_count=209934.66666666666, ans=15.0
2024-09-23 07:42:03,574 INFO [train.py:1198] (1/4) Epoch 12, batch 2150, loss[loss=0.2486, ctc_loss=0.1696, cr_loss=0.3948, over 17024.00 frames. ], tot_loss[loss=0.2428, ctc_loss=0.1671, cr_loss=0.3786, over 3351798.84 frames. ], batch size: 52, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:42:03,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=210028.0, ans=0.5
2024-09-23 07:42:05,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=210028.0, ans=0.07
2024-09-23 07:42:07,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.16 vs. limit=15.0
2024-09-23 07:42:15,811 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=7.03 vs. limit=15.0
2024-09-23 07:42:33,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=210121.33333333334, ans=0.125
2024-09-23 07:42:51,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=210168.0, ans=0.0
2024-09-23 07:43:00,522 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.318e+02 1.400e+02 1.574e+02 2.180e+02, threshold=2.800e+02, percent-clipped=0.0
2024-09-23 07:43:10,799 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.97 vs. limit=22.5
2024-09-23 07:43:22,704 INFO [train.py:1198] (1/4) Epoch 12, batch 2200, loss[loss=0.2316, ctc_loss=0.1556, cr_loss=0.3799, over 17217.00 frames. ], tot_loss[loss=0.2417, ctc_loss=0.1662, cr_loss=0.3776, over 3362299.99 frames. ], batch size: 41, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:43:28,364 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.16 vs. limit=15.0
2024-09-23 07:43:31,852 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.04 vs. limit=10.0
2024-09-23 07:43:46,702 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.77 vs. limit=22.5
2024-09-23 07:43:56,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=210308.0, ans=0.1
2024-09-23 07:44:07,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=210354.66666666666, ans=0.0
2024-09-23 07:44:16,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=210401.33333333334, ans=0.0
2024-09-23 07:44:18,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=210401.33333333334, ans=0.125
2024-09-23 07:44:39,522 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.93 vs. limit=6.0
2024-09-23 07:44:40,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=210448.0, ans=0.0
2024-09-23 07:44:50,074 INFO [train.py:1198] (1/4) Epoch 12, batch 2250, loss[loss=0.2376, ctc_loss=0.1613, cr_loss=0.3814, over 17131.00 frames. ], tot_loss[loss=0.2393, ctc_loss=0.1644, cr_loss=0.3743, over 3361737.50 frames. ], batch size: 48, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:45:46,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=210634.66666666666, ans=0.025
2024-09-23 07:45:49,555 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.306e+02 1.486e+02 1.708e+02 2.318e+02, threshold=2.971e+02, percent-clipped=0.0
2024-09-23 07:45:50,782 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.17 vs. limit=10.0
2024-09-23 07:45:53,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=210634.66666666666, ans=0.125
2024-09-23 07:45:56,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=210681.33333333334, ans=0.125
2024-09-23 07:46:01,500 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.24 vs. limit=15.0
2024-09-23 07:46:04,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=210681.33333333334, ans=0.2
2024-09-23 07:46:07,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=210681.33333333334, ans=0.125
2024-09-23 07:46:11,907 INFO [train.py:1198] (1/4) Epoch 12, batch 2300, loss[loss=0.2473, ctc_loss=0.1701, cr_loss=0.3859, over 17221.00 frames. ], tot_loss[loss=0.2413, ctc_loss=0.1659, cr_loss=0.3773, over 3352197.50 frames. ], batch size: 55, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:46:13,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=210728.0, ans=0.125
2024-09-23 07:46:16,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=210728.0, ans=0.0
2024-09-23 07:46:58,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=210868.0, ans=0.025
2024-09-23 07:46:59,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=9.16 vs. limit=22.5
2024-09-23 07:47:06,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=210868.0, ans=0.125
2024-09-23 07:47:13,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=210868.0, ans=0.125
2024-09-23 07:47:18,413 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=19.77 vs. limit=22.5
2024-09-23 07:47:27,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=210914.66666666666, ans=0.04949747468305833
2024-09-23 07:47:30,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=210961.33333333334, ans=0.125
2024-09-23 07:47:32,370 INFO [train.py:1198] (1/4) Epoch 12, batch 2350, loss[loss=0.2691, ctc_loss=0.1866, cr_loss=0.4128, over 17199.00 frames. ], tot_loss[loss=0.2403, ctc_loss=0.1651, cr_loss=0.3763, over 3352689.86 frames. ], batch size: 50, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:47:42,726 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.20 vs. limit=12.0
2024-09-23 07:47:53,615 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:48:19,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=211101.33333333334, ans=0.125
2024-09-23 07:48:23,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=211101.33333333334, ans=0.125
2024-09-23 07:48:29,120 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.329e+02 1.457e+02 1.570e+02 2.040e+02, threshold=2.915e+02, percent-clipped=0.0
2024-09-23 07:48:56,505 INFO [train.py:1198] (1/4) Epoch 12, batch 2400, loss[loss=0.1849, ctc_loss=0.1249, cr_loss=0.3, over 16710.00 frames. ], tot_loss[loss=0.2399, ctc_loss=0.1646, cr_loss=0.3762, over 3357553.78 frames. ], batch size: 37, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:49:07,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.85 vs. limit=10.0
2024-09-23 07:50:21,841 INFO [train.py:1198] (1/4) Epoch 12, batch 2450, loss[loss=0.2571, ctc_loss=0.1821, cr_loss=0.3752, over 15059.00 frames. ], tot_loss[loss=0.2399, ctc_loss=0.1647, cr_loss=0.3758, over 3337817.81 frames. ], batch size: 89, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:51:19,220 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.359e+02 1.540e+02 1.815e+02 2.529e+02, threshold=3.080e+02, percent-clipped=0.0
2024-09-23 07:51:41,364 INFO [train.py:1198] (1/4) Epoch 12, batch 2500, loss[loss=0.2232, ctc_loss=0.1489, cr_loss=0.3718, over 17358.00 frames. ], tot_loss[loss=0.239, ctc_loss=0.164, cr_loss=0.3749, over 3348515.30 frames. ], batch size: 48, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:51:54,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=211661.33333333334, ans=0.025
2024-09-23 07:52:13,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=211754.66666666666, ans=0.07
2024-09-23 07:52:20,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=211754.66666666666, ans=0.0
2024-09-23 07:52:46,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=211848.0, ans=0.125
2024-09-23 07:53:00,936 INFO [train.py:1198] (1/4) Epoch 12, batch 2550, loss[loss=0.2465, ctc_loss=0.1672, cr_loss=0.3966, over 17227.00 frames. ], tot_loss[loss=0.2393, ctc_loss=0.1643, cr_loss=0.3754, over 3356459.70 frames. ], batch size: 50, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:53:05,163 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.67 vs. limit=6.0
2024-09-23 07:53:29,167 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.26 vs. limit=12.0
2024-09-23 07:53:41,009 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.58 vs. limit=22.5
2024-09-23 07:54:00,929 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.326e+02 1.484e+02 1.777e+02 2.748e+02, threshold=2.968e+02, percent-clipped=0.0
2024-09-23 07:54:13,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=212081.33333333334, ans=0.07
2024-09-23 07:54:25,810 INFO [train.py:1198] (1/4) Epoch 12, batch 2600, loss[loss=0.1911, ctc_loss=0.1279, cr_loss=0.3159, over 17111.00 frames. ], tot_loss[loss=0.2396, ctc_loss=0.1645, cr_loss=0.3756, over 3350519.02 frames. ], batch size: 40, lr: 1.00e-02, grad_scale: 32.0
2024-09-23 07:54:52,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=212174.66666666666, ans=0.2
2024-09-23 07:54:52,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=212174.66666666666, ans=0.1
2024-09-23 07:55:06,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=212221.33333333334, ans=0.0
2024-09-23 07:55:08,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=212221.33333333334, ans=0.07
2024-09-23 07:55:47,927 INFO [train.py:1198] (1/4) Epoch 12, batch 2650, loss[loss=0.2305, ctc_loss=0.1579, cr_loss=0.3627, over 17286.00 frames. ], tot_loss[loss=0.2388, ctc_loss=0.1639, cr_loss=0.3746, over 3361409.27 frames. ], batch size: 46, lr: 9.99e-03, grad_scale: 32.0
2024-09-23 07:55:57,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=212361.33333333334, ans=0.05
2024-09-23 07:56:15,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=212408.0, ans=0.0
2024-09-23 07:56:15,554 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 07:56:36,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=212501.33333333334, ans=0.05
2024-09-23 07:56:45,680 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.240e+02 1.338e+02 1.470e+02 2.352e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-23 07:56:52,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=212548.0, ans=0.125
2024-09-23 07:57:08,182 INFO [train.py:1198] (1/4) Epoch 12, batch 2700, loss[loss=0.2082, ctc_loss=0.1397, cr_loss=0.3426, over 17203.00 frames. ], tot_loss[loss=0.238, ctc_loss=0.1632, cr_loss=0.3741, over 3360636.06 frames. ], batch size: 41, lr: 9.99e-03, grad_scale: 32.0
2024-09-23 07:57:18,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2.whitening_limit, batch_count=212594.66666666666, ans=15.0
2024-09-23 07:57:43,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=212688.0, ans=0.125
2024-09-23 07:58:04,933 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.55 vs. limit=15.0
2024-09-23 07:58:12,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=212781.33333333334, ans=0.0
2024-09-23 07:58:12,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=212781.33333333334, ans=0.125
2024-09-23 07:58:12,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=212781.33333333334, ans=0.0
2024-09-23 07:58:13,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=212781.33333333334, ans=0.025
2024-09-23 07:58:23,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=212781.33333333334, ans=0.125
2024-09-23 07:58:28,274 INFO [train.py:1198] (1/4) Epoch 12, batch 2750, loss[loss=0.2578, ctc_loss=0.1794, cr_loss=0.3918, over 16895.00 frames. ], tot_loss[loss=0.2379, ctc_loss=0.1631, cr_loss=0.3736, over 3364497.45 frames. ], batch size: 58, lr: 9.98e-03, grad_scale: 32.0
2024-09-23 07:58:46,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=212828.0, ans=0.025
2024-09-23 07:59:33,712 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.344e+02 1.475e+02 1.726e+02 2.611e+02, threshold=2.950e+02, percent-clipped=0.0
2024-09-23 07:59:58,918 INFO [train.py:1198] (1/4) Epoch 12, batch 2800, loss[loss=0.2831, ctc_loss=0.2034, cr_loss=0.3984, over 12064.00 frames. ], tot_loss[loss=0.2384, ctc_loss=0.1636, cr_loss=0.3741, over 3362377.76 frames. ], batch size: 125, lr: 9.98e-03, grad_scale: 32.0
2024-09-23 08:00:04,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=213061.33333333334, ans=0.125
2024-09-23 08:00:17,448 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.99 vs. limit=10.0
2024-09-23 08:00:20,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=213108.0, ans=0.025
2024-09-23 08:00:37,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=213154.66666666666, ans=0.125
2024-09-23 08:00:38,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=213154.66666666666, ans=0.0
2024-09-23 08:00:43,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=213154.66666666666, ans=0.125
2024-09-23 08:01:18,197 INFO [train.py:1198] (1/4) Epoch 12, batch 2850, loss[loss=0.2755, ctc_loss=0.1893, cr_loss=0.4309, over 16869.00 frames. ], tot_loss[loss=0.2395, ctc_loss=0.1645, cr_loss=0.3753, over 3363446.76 frames. ], batch size: 58, lr: 9.97e-03, grad_scale: 32.0
2024-09-23 08:01:28,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=213294.66666666666, ans=0.1
2024-09-23 08:01:30,055 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.34 vs. limit=15.0
2024-09-23 08:01:57,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=213388.0, ans=0.2
2024-09-23 08:02:15,761 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.286e+02 1.451e+02 1.716e+02 2.634e+02, threshold=2.902e+02, percent-clipped=0.0
2024-09-23 08:02:32,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=213481.33333333334, ans=0.125
2024-09-23 08:02:38,157 INFO [train.py:1198] (1/4) Epoch 12, batch 2900, loss[loss=0.2426, ctc_loss=0.1682, cr_loss=0.3718, over 17316.00 frames. ], tot_loss[loss=0.2384, ctc_loss=0.1636, cr_loss=0.3743, over 3369600.65 frames. ], batch size: 49, lr: 9.97e-03, grad_scale: 32.0
2024-09-23 08:02:44,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=213528.0, ans=0.125
2024-09-23 08:02:52,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=213574.66666666666, ans=0.0
2024-09-23 08:02:55,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=213574.66666666666, ans=0.2
2024-09-23 08:03:18,636 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.52 vs. limit=15.0
2024-09-23 08:03:29,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=213668.0, ans=0.125
2024-09-23 08:03:32,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=213668.0, ans=0.2
2024-09-23 08:03:54,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=213714.66666666666, ans=0.1
2024-09-23 08:04:02,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=213714.66666666666, ans=0.0
2024-09-23 08:04:04,995 INFO [train.py:1198] (1/4) Epoch 12, batch 2950, loss[loss=0.2023, ctc_loss=0.1346, cr_loss=0.3383, over 17262.00 frames. ], tot_loss[loss=0.2389, ctc_loss=0.164, cr_loss=0.3747, over 3368912.23 frames. ], batch size: 42, lr: 9.96e-03, grad_scale: 32.0
2024-09-23 08:04:33,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=213808.0, ans=0.125
2024-09-23 08:04:55,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=213901.33333333334, ans=0.125
2024-09-23 08:05:04,862 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.287e+02 1.399e+02 1.568e+02 2.905e+02, threshold=2.798e+02, percent-clipped=1.0
2024-09-23 08:05:07,628 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.66 vs. limit=6.0
2024-09-23 08:05:14,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=213948.0, ans=0.1
2024-09-23 08:05:16,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=213948.0, ans=0.0
2024-09-23 08:05:26,775 INFO [train.py:1198] (1/4) Epoch 12, batch 3000, loss[loss=0.2428, ctc_loss=0.166, cr_loss=0.3838, over 17209.00 frames. ], tot_loss[loss=0.2389, ctc_loss=0.1639, cr_loss=0.3748, over 3364089.67 frames. ], batch size: 47, lr: 9.96e-03, grad_scale: 32.0
2024-09-23 08:05:26,775 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 08:05:34,245 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.5.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([2.8272, 4.1708, 3.8140, 4.6390], device='cuda:1')
2024-09-23 08:05:42,575 INFO [train.py:1230] (1/4) Epoch 12, validation: loss=0.04588, ctc_loss=0.04588, cr_loss=7.526e-15, over 944034.00 frames. 
2024-09-23 08:05:42,575 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 08:05:53,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=213994.66666666666, ans=0.2
2024-09-23 08:05:54,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.36 vs. limit=15.0
2024-09-23 08:06:07,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=214041.33333333334, ans=0.2
2024-09-23 08:06:18,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=214088.0, ans=0.025
2024-09-23 08:06:42,434 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.87 vs. limit=15.0
2024-09-23 08:07:00,680 INFO [train.py:1198] (1/4) Epoch 12, batch 3050, loss[loss=0.2328, ctc_loss=0.1607, cr_loss=0.3603, over 17141.00 frames. ], tot_loss[loss=0.2395, ctc_loss=0.1643, cr_loss=0.3757, over 3359578.96 frames. ], batch size: 48, lr: 9.95e-03, grad_scale: 32.0
2024-09-23 08:07:22,886 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=15.18 vs. limit=15.0
2024-09-23 08:07:31,026 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.33 vs. limit=22.5
2024-09-23 08:07:51,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=214368.0, ans=0.125
2024-09-23 08:07:56,197 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.305e+02 1.426e+02 1.616e+02 2.340e+02, threshold=2.852e+02, percent-clipped=0.0
2024-09-23 08:08:17,698 INFO [train.py:1198] (1/4) Epoch 12, batch 3100, loss[loss=0.2599, ctc_loss=0.1805, cr_loss=0.3968, over 17103.00 frames. ], tot_loss[loss=0.2408, ctc_loss=0.1655, cr_loss=0.3765, over 3348735.85 frames. ], batch size: 49, lr: 9.94e-03, grad_scale: 32.0
2024-09-23 08:09:10,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=214601.33333333334, ans=0.125
2024-09-23 08:09:12,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=214601.33333333334, ans=0.1
2024-09-23 08:09:13,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=214601.33333333334, ans=0.125
2024-09-23 08:09:30,145 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 08:09:36,171 INFO [train.py:1198] (1/4) Epoch 12, batch 3150, loss[loss=0.2462, ctc_loss=0.1689, cr_loss=0.3866, over 17217.00 frames. ], tot_loss[loss=0.2392, ctc_loss=0.1641, cr_loss=0.3756, over 3359077.47 frames. ], batch size: 50, lr: 9.94e-03, grad_scale: 32.0
2024-09-23 08:09:36,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=214694.66666666666, ans=0.125
2024-09-23 08:09:55,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=214741.33333333334, ans=0.125
2024-09-23 08:10:15,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=214788.0, ans=0.125
2024-09-23 08:10:32,318 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.379e+02 1.486e+02 1.636e+02 2.338e+02, threshold=2.971e+02, percent-clipped=0.0
2024-09-23 08:10:38,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=214881.33333333334, ans=0.125
2024-09-23 08:10:54,108 INFO [train.py:1198] (1/4) Epoch 12, batch 3200, loss[loss=0.3043, ctc_loss=0.2187, cr_loss=0.428, over 15266.00 frames. ], tot_loss[loss=0.24, ctc_loss=0.1648, cr_loss=0.3762, over 3362585.99 frames. ], batch size: 89, lr: 9.93e-03, grad_scale: 32.0
2024-09-23 08:11:12,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=214974.66666666666, ans=0.125
2024-09-23 08:12:08,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=215114.66666666666, ans=0.2
2024-09-23 08:12:16,041 INFO [train.py:1198] (1/4) Epoch 12, batch 3250, loss[loss=0.1842, ctc_loss=0.125, cr_loss=0.2962, over 16281.00 frames. ], tot_loss[loss=0.2406, ctc_loss=0.1651, cr_loss=0.3773, over 3359888.96 frames. ], batch size: 36, lr: 9.93e-03, grad_scale: 16.0
2024-09-23 08:12:16,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=215161.33333333334, ans=0.0
2024-09-23 08:12:17,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=215161.33333333334, ans=0.125
2024-09-23 08:12:28,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=215161.33333333334, ans=0.07
2024-09-23 08:12:53,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=215254.66666666666, ans=0.125
2024-09-23 08:12:56,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=215254.66666666666, ans=0.2
2024-09-23 08:13:13,440 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.295e+02 1.414e+02 1.544e+02 2.184e+02, threshold=2.828e+02, percent-clipped=0.0
2024-09-23 08:13:35,944 INFO [train.py:1198] (1/4) Epoch 12, batch 3300, loss[loss=0.2386, ctc_loss=0.1622, cr_loss=0.3819, over 17298.00 frames. ], tot_loss[loss=0.2402, ctc_loss=0.1648, cr_loss=0.3767, over 3358065.62 frames. ], batch size: 49, lr: 9.92e-03, grad_scale: 16.0
2024-09-23 08:13:39,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=215394.66666666666, ans=0.2
2024-09-23 08:13:54,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=215441.33333333334, ans=0.015
2024-09-23 08:13:56,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=215441.33333333334, ans=0.0
2024-09-23 08:13:57,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.max_abs, batch_count=215441.33333333334, ans=10.0
2024-09-23 08:14:12,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=215488.0, ans=0.125
2024-09-23 08:14:27,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=215534.66666666666, ans=0.125
2024-09-23 08:14:29,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=215534.66666666666, ans=0.125
2024-09-23 08:14:37,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=215534.66666666666, ans=0.2
2024-09-23 08:14:48,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=215581.33333333334, ans=0.125
2024-09-23 08:14:56,380 INFO [train.py:1198] (1/4) Epoch 12, batch 3350, loss[loss=0.2223, ctc_loss=0.1484, cr_loss=0.3698, over 17098.00 frames. ], tot_loss[loss=0.2402, ctc_loss=0.165, cr_loss=0.3764, over 3358582.72 frames. ], batch size: 43, lr: 9.92e-03, grad_scale: 16.0
2024-09-23 08:15:10,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=215674.66666666666, ans=0.2
2024-09-23 08:15:44,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=215768.0, ans=0.125
2024-09-23 08:15:49,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=215768.0, ans=0.1
2024-09-23 08:15:54,077 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.291e+02 1.444e+02 1.665e+02 2.877e+02, threshold=2.888e+02, percent-clipped=1.0
2024-09-23 08:15:55,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=215768.0, ans=0.125
2024-09-23 08:16:06,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=215814.66666666666, ans=0.2
2024-09-23 08:16:14,339 INFO [train.py:1198] (1/4) Epoch 12, batch 3400, loss[loss=0.217, ctc_loss=0.1499, cr_loss=0.3356, over 17206.00 frames. ], tot_loss[loss=0.2396, ctc_loss=0.1645, cr_loss=0.3756, over 3372987.46 frames. ], batch size: 41, lr: 9.91e-03, grad_scale: 16.0
2024-09-23 08:16:22,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=215861.33333333334, ans=0.1
2024-09-23 08:16:42,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=215908.0, ans=0.125
2024-09-23 08:17:12,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=216001.33333333334, ans=0.125
2024-09-23 08:17:19,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=216048.0, ans=0.125
2024-09-23 08:17:27,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=216048.0, ans=0.125
2024-09-23 08:17:32,348 INFO [train.py:1198] (1/4) Epoch 12, batch 3450, loss[loss=0.1942, ctc_loss=0.1311, cr_loss=0.3158, over 17122.00 frames. ], tot_loss[loss=0.2389, ctc_loss=0.1639, cr_loss=0.3749, over 3369466.28 frames. ], batch size: 40, lr: 9.91e-03, grad_scale: 16.0
2024-09-23 08:17:32,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=216094.66666666666, ans=0.0
2024-09-23 08:17:50,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=216141.33333333334, ans=0.0
2024-09-23 08:17:50,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.24 vs. limit=22.5
2024-09-23 08:18:30,276 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.280e+02 1.416e+02 1.630e+02 3.213e+02, threshold=2.832e+02, percent-clipped=1.0
2024-09-23 08:18:50,513 INFO [train.py:1198] (1/4) Epoch 12, batch 3500, loss[loss=0.2446, ctc_loss=0.1653, cr_loss=0.3968, over 17059.00 frames. ], tot_loss[loss=0.2389, ctc_loss=0.1639, cr_loss=0.3753, over 3367201.88 frames. ], batch size: 46, lr: 9.90e-03, grad_scale: 16.0
2024-09-23 08:18:53,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=7.56 vs. limit=15.0
2024-09-23 08:19:09,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=216374.66666666666, ans=0.1
2024-09-23 08:19:12,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=216374.66666666666, ans=0.125
2024-09-23 08:19:39,484 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.01 vs. limit=6.0
2024-09-23 08:19:53,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=216514.66666666666, ans=0.0
2024-09-23 08:20:03,093 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=15.27 vs. limit=15.0
2024-09-23 08:20:08,728 INFO [train.py:1198] (1/4) Epoch 12, batch 3550, loss[loss=0.2074, ctc_loss=0.1404, cr_loss=0.335, over 17096.00 frames. ], tot_loss[loss=0.2387, ctc_loss=0.1637, cr_loss=0.3751, over 3373140.58 frames. ], batch size: 40, lr: 9.90e-03, grad_scale: 16.0
2024-09-23 08:20:16,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=216561.33333333334, ans=0.125
2024-09-23 08:20:53,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=216701.33333333334, ans=0.04949747468305833
2024-09-23 08:20:55,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=216701.33333333334, ans=0.0
2024-09-23 08:21:05,845 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.266e+02 1.416e+02 1.619e+02 2.341e+02, threshold=2.832e+02, percent-clipped=0.0
2024-09-23 08:21:28,066 INFO [train.py:1198] (1/4) Epoch 12, batch 3600, loss[loss=0.225, ctc_loss=0.154, cr_loss=0.3552, over 17213.00 frames. ], tot_loss[loss=0.2391, ctc_loss=0.1641, cr_loss=0.375, over 3375852.15 frames. ], batch size: 55, lr: 9.89e-03, grad_scale: 32.0
2024-09-23 08:22:00,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=216888.0, ans=0.09899494936611666
2024-09-23 08:22:01,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=216888.0, ans=0.1
2024-09-23 08:22:01,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=216888.0, ans=0.2
2024-09-23 08:22:34,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=216981.33333333334, ans=0.2
2024-09-23 08:22:48,414 INFO [train.py:1198] (1/4) Epoch 12, batch 3650, loss[loss=0.2505, ctc_loss=0.1731, cr_loss=0.3867, over 16536.00 frames. ], tot_loss[loss=0.2399, ctc_loss=0.1648, cr_loss=0.3756, over 3376782.21 frames. ], batch size: 66, lr: 9.89e-03, grad_scale: 32.0
2024-09-23 08:23:37,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=217168.0, ans=0.1
2024-09-23 08:23:37,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=217168.0, ans=0.025
2024-09-23 08:23:46,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=217168.0, ans=0.125
2024-09-23 08:23:50,114 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.266e+02 1.360e+02 1.445e+02 2.351e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-23 08:23:54,927 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=15.08 vs. limit=22.5
2024-09-23 08:24:10,933 INFO [train.py:1198] (1/4) Epoch 12, batch 3700, loss[loss=0.202, ctc_loss=0.1322, cr_loss=0.3486, over 17097.00 frames. ], tot_loss[loss=0.2388, ctc_loss=0.1639, cr_loss=0.3746, over 3381808.26 frames. ], batch size: 40, lr: 9.88e-03, grad_scale: 32.0
2024-09-23 08:24:14,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=217261.33333333334, ans=0.05
2024-09-23 08:24:28,866 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.68 vs. limit=22.5
2024-09-23 08:24:42,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=217354.66666666666, ans=0.125
2024-09-23 08:24:56,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=217401.33333333334, ans=0.2
2024-09-23 08:25:29,359 INFO [train.py:1198] (1/4) Epoch 12, batch 3750, loss[loss=0.1907, ctc_loss=0.1273, cr_loss=0.3171, over 17020.00 frames. ], tot_loss[loss=0.2388, ctc_loss=0.1638, cr_loss=0.3748, over 3376389.31 frames. ], batch size: 39, lr: 9.88e-03, grad_scale: 32.0
2024-09-23 08:25:29,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=217494.66666666666, ans=0.125
2024-09-23 08:26:05,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=217588.0, ans=0.125
2024-09-23 08:26:06,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=217588.0, ans=0.125
2024-09-23 08:26:26,722 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.365e+02 1.477e+02 1.661e+02 2.472e+02, threshold=2.954e+02, percent-clipped=0.0
2024-09-23 08:26:36,880 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.78 vs. limit=22.5
2024-09-23 08:26:47,064 INFO [train.py:1198] (1/4) Epoch 12, batch 3800, loss[loss=0.2266, ctc_loss=0.1551, cr_loss=0.3579, over 17211.00 frames. ], tot_loss[loss=0.2403, ctc_loss=0.1653, cr_loss=0.3749, over 3333096.06 frames. ], batch size: 41, lr: 9.87e-03, grad_scale: 32.0
2024-09-23 08:26:47,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=217728.0, ans=0.125
2024-09-23 08:27:11,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=217774.66666666666, ans=0.2
2024-09-23 08:27:11,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=217774.66666666666, ans=0.1
2024-09-23 08:27:27,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=217821.33333333334, ans=0.125
2024-09-23 08:27:32,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten.whitening_limit, batch_count=217821.33333333334, ans=15.0
2024-09-23 08:27:42,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=217868.0, ans=0.125
2024-09-23 08:27:47,957 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.98 vs. limit=15.0
2024-09-23 08:27:53,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=217914.66666666666, ans=0.125
2024-09-23 08:28:04,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=217961.33333333334, ans=0.0
2024-09-23 08:28:04,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=217961.33333333334, ans=0.125
2024-09-23 08:28:05,900 INFO [train.py:1198] (1/4) Epoch 12, batch 3850, loss[loss=0.186, ctc_loss=0.1239, cr_loss=0.3103, over 17216.00 frames. ], tot_loss[loss=0.2433, ctc_loss=0.1678, cr_loss=0.3772, over 3281291.15 frames. ], batch size: 41, lr: 9.87e-03, grad_scale: 32.0
2024-09-23 08:28:30,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=218008.0, ans=0.025
2024-09-23 08:29:02,106 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.424e+02 1.630e+02 1.758e+02 2.384e+02, threshold=3.259e+02, percent-clipped=0.0
2024-09-23 08:29:12,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=218148.0, ans=0.0
2024-09-23 08:30:07,238 INFO [train.py:1198] (1/4) Epoch 13, batch 0, loss[loss=0.2973, ctc_loss=0.2065, cr_loss=0.4537, over 17013.00 frames. ], tot_loss[loss=0.2973, ctc_loss=0.2065, cr_loss=0.4537, over 17013.00 frames. ], batch size: 53, lr: 9.48e-03, grad_scale: 32.0
2024-09-23 08:30:07,239 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 08:30:15,310 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.5607, 4.0834, 3.8637, 4.0838], device='cuda:1')
2024-09-23 08:30:22,747 INFO [train.py:1230] (1/4) Epoch 13, validation: loss=0.04407, ctc_loss=0.04407, cr_loss=7.62e-15, over 944034.00 frames. 
2024-09-23 08:30:22,747 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 08:30:24,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=218176.0, ans=0.125
2024-09-23 08:30:24,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=218176.0, ans=0.125
2024-09-23 08:30:39,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=218222.66666666666, ans=0.125
2024-09-23 08:31:33,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=218362.66666666666, ans=0.125
2024-09-23 08:31:42,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.57 vs. limit=15.0
2024-09-23 08:31:43,033 INFO [train.py:1198] (1/4) Epoch 13, batch 50, loss[loss=0.2354, ctc_loss=0.1608, cr_loss=0.3729, over 17144.00 frames. ], tot_loss[loss=0.2356, ctc_loss=0.161, cr_loss=0.3731, over 762565.71 frames. ], batch size: 48, lr: 9.47e-03, grad_scale: 32.0
2024-09-23 08:32:05,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=218456.0, ans=0.125
2024-09-23 08:32:15,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=218456.0, ans=0.0
2024-09-23 08:32:29,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=218502.66666666666, ans=0.125
2024-09-23 08:32:35,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=218549.33333333334, ans=0.0
2024-09-23 08:32:45,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=218549.33333333334, ans=0.1
2024-09-23 08:32:50,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=218596.0, ans=0.125
2024-09-23 08:32:51,665 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.296e+02 1.397e+02 1.549e+02 2.228e+02, threshold=2.794e+02, percent-clipped=0.0
2024-09-23 08:32:56,244 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.56 vs. limit=10.0
2024-09-23 08:32:56,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=218596.0, ans=0.0
2024-09-23 08:33:08,627 INFO [train.py:1198] (1/4) Epoch 13, batch 100, loss[loss=0.2447, ctc_loss=0.1644, cr_loss=0.4011, over 17034.00 frames. ], tot_loss[loss=0.2396, ctc_loss=0.1642, cr_loss=0.3766, over 1338290.42 frames. ], batch size: 52, lr: 9.47e-03, grad_scale: 32.0
2024-09-23 08:33:09,778 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.20 vs. limit=15.0
2024-09-23 08:33:10,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=218642.66666666666, ans=0.125
2024-09-23 08:33:23,240 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 08:33:24,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=218689.33333333334, ans=0.125
2024-09-23 08:33:47,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=218736.0, ans=0.2
2024-09-23 08:33:49,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=218736.0, ans=0.125
2024-09-23 08:33:49,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=218736.0, ans=0.1
2024-09-23 08:33:52,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=218736.0, ans=0.0
2024-09-23 08:34:01,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=218782.66666666666, ans=0.0
2024-09-23 08:34:27,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=218876.0, ans=0.125
2024-09-23 08:34:28,452 INFO [train.py:1198] (1/4) Epoch 13, batch 150, loss[loss=0.2195, ctc_loss=0.1466, cr_loss=0.3646, over 17262.00 frames. ], tot_loss[loss=0.2393, ctc_loss=0.1641, cr_loss=0.3759, over 1770719.18 frames. ], batch size: 42, lr: 9.46e-03, grad_scale: 32.0
2024-09-23 08:34:36,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=218876.0, ans=0.125
2024-09-23 08:35:02,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=218969.33333333334, ans=0.1
2024-09-23 08:35:10,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=218969.33333333334, ans=0.025
2024-09-23 08:35:28,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=219016.0, ans=0.125
2024-09-23 08:35:30,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=219016.0, ans=0.125
2024-09-23 08:35:33,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=219062.66666666666, ans=0.125
2024-09-23 08:35:36,675 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.277e+02 1.397e+02 1.518e+02 2.217e+02, threshold=2.795e+02, percent-clipped=0.0
2024-09-23 08:35:51,083 INFO [train.py:1198] (1/4) Epoch 13, batch 200, loss[loss=0.236, ctc_loss=0.164, cr_loss=0.3603, over 17211.00 frames. ], tot_loss[loss=0.2382, ctc_loss=0.1633, cr_loss=0.3745, over 2133836.76 frames. ], batch size: 50, lr: 9.46e-03, grad_scale: 32.0
2024-09-23 08:36:58,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=219296.0, ans=0.125
2024-09-23 08:37:00,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=219296.0, ans=0.125
2024-09-23 08:37:03,474 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=219296.0, ans=0.025
2024-09-23 08:37:06,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=219296.0, ans=0.025
2024-09-23 08:37:11,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.min_positive, batch_count=219296.0, ans=0.025
2024-09-23 08:37:15,242 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.49 vs. limit=6.0
2024-09-23 08:37:16,183 INFO [train.py:1198] (1/4) Epoch 13, batch 250, loss[loss=0.2203, ctc_loss=0.1477, cr_loss=0.3631, over 17086.00 frames. ], tot_loss[loss=0.2382, ctc_loss=0.1633, cr_loss=0.3746, over 2411480.91 frames. ], batch size: 46, lr: 9.45e-03, grad_scale: 32.0
2024-09-23 08:37:32,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=219389.33333333334, ans=0.2
2024-09-23 08:38:24,170 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.243e+02 1.365e+02 1.573e+02 3.010e+02, threshold=2.729e+02, percent-clipped=2.0
2024-09-23 08:38:29,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=219529.33333333334, ans=0.125
2024-09-23 08:38:35,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=219529.33333333334, ans=0.125
2024-09-23 08:38:38,515 INFO [train.py:1198] (1/4) Epoch 13, batch 300, loss[loss=0.2487, ctc_loss=0.1723, cr_loss=0.3819, over 16469.00 frames. ], tot_loss[loss=0.2362, ctc_loss=0.1617, cr_loss=0.3725, over 2622064.14 frames. ], batch size: 66, lr: 9.45e-03, grad_scale: 32.0
2024-09-23 08:39:31,835 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.00 vs. limit=10.0
2024-09-23 08:39:53,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=219762.66666666666, ans=0.2
2024-09-23 08:39:55,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=219762.66666666666, ans=0.125
2024-09-23 08:39:58,339 INFO [train.py:1198] (1/4) Epoch 13, batch 350, loss[loss=0.2427, ctc_loss=0.1683, cr_loss=0.372, over 16693.00 frames. ], tot_loss[loss=0.2365, ctc_loss=0.1619, cr_loss=0.3731, over 2788347.41 frames. ], batch size: 61, lr: 9.44e-03, grad_scale: 32.0
2024-09-23 08:40:03,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=219809.33333333334, ans=0.125
2024-09-23 08:40:04,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=219809.33333333334, ans=0.0
2024-09-23 08:40:07,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=219809.33333333334, ans=0.125
2024-09-23 08:40:09,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=219809.33333333334, ans=0.125
2024-09-23 08:40:36,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=219902.66666666666, ans=0.125
2024-09-23 08:40:42,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=219902.66666666666, ans=0.1
2024-09-23 08:41:06,038 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.335e+02 1.492e+02 1.717e+02 2.357e+02, threshold=2.983e+02, percent-clipped=0.0
2024-09-23 08:41:20,270 INFO [train.py:1198] (1/4) Epoch 13, batch 400, loss[loss=0.2494, ctc_loss=0.1727, cr_loss=0.3833, over 17022.00 frames. ], tot_loss[loss=0.2365, ctc_loss=0.1619, cr_loss=0.3733, over 2915977.45 frames. ], batch size: 56, lr: 9.44e-03, grad_scale: 32.0
2024-09-23 08:42:45,838 INFO [train.py:1198] (1/4) Epoch 13, batch 450, loss[loss=0.2132, ctc_loss=0.1467, cr_loss=0.3327, over 17112.00 frames. ], tot_loss[loss=0.2353, ctc_loss=0.161, cr_loss=0.3718, over 3022422.20 frames. ], batch size: 40, lr: 9.43e-03, grad_scale: 32.0
2024-09-23 08:42:47,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=220276.0, ans=0.0
2024-09-23 08:43:14,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=220322.66666666666, ans=0.125
2024-09-23 08:43:19,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=220369.33333333334, ans=0.025
2024-09-23 08:43:53,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten.whitening_limit, batch_count=220462.66666666666, ans=22.5
2024-09-23 08:43:53,977 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.263e+02 1.351e+02 1.525e+02 2.528e+02, threshold=2.701e+02, percent-clipped=0.0
2024-09-23 08:43:54,699 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.84 vs. limit=6.0
2024-09-23 08:44:08,348 INFO [train.py:1198] (1/4) Epoch 13, batch 500, loss[loss=0.2273, ctc_loss=0.155, cr_loss=0.3613, over 17023.00 frames. ], tot_loss[loss=0.2344, ctc_loss=0.1604, cr_loss=0.3702, over 3103813.05 frames. ], batch size: 44, lr: 9.43e-03, grad_scale: 32.0
2024-09-23 08:45:06,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=220649.33333333334, ans=0.09899494936611666
2024-09-23 08:45:12,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=220696.0, ans=0.125
2024-09-23 08:45:17,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=220696.0, ans=0.2
2024-09-23 08:45:25,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=220696.0, ans=0.125
2024-09-23 08:45:31,233 INFO [train.py:1198] (1/4) Epoch 13, batch 550, loss[loss=0.2152, ctc_loss=0.1472, cr_loss=0.3398, over 17352.00 frames. ], tot_loss[loss=0.2338, ctc_loss=0.1599, cr_loss=0.3698, over 3154137.16 frames. ], batch size: 48, lr: 9.42e-03, grad_scale: 32.0
2024-09-23 08:45:45,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=220789.33333333334, ans=0.0
2024-09-23 08:45:52,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=220789.33333333334, ans=0.1
2024-09-23 08:46:00,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=220789.33333333334, ans=0.0
2024-09-23 08:46:20,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=220882.66666666666, ans=0.125
2024-09-23 08:46:36,681 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.262e+02 1.359e+02 1.486e+02 2.281e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-23 08:46:56,575 INFO [train.py:1198] (1/4) Epoch 13, batch 600, loss[loss=0.2223, ctc_loss=0.1478, cr_loss=0.3726, over 17059.00 frames. ], tot_loss[loss=0.2344, ctc_loss=0.1603, cr_loss=0.3705, over 3200791.36 frames. ], batch size: 39, lr: 9.42e-03, grad_scale: 32.0
2024-09-23 08:47:31,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=221069.33333333334, ans=0.125
2024-09-23 08:47:33,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=221069.33333333334, ans=0.125
2024-09-23 08:48:10,091 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.98 vs. limit=10.0
2024-09-23 08:48:18,960 INFO [train.py:1198] (1/4) Epoch 13, batch 650, loss[loss=0.2177, ctc_loss=0.15, cr_loss=0.3387, over 16965.00 frames. ], tot_loss[loss=0.235, ctc_loss=0.1607, cr_loss=0.3715, over 3232627.13 frames. ], batch size: 42, lr: 9.41e-03, grad_scale: 32.0
2024-09-23 08:48:27,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=221209.33333333334, ans=0.125
2024-09-23 08:48:30,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=221209.33333333334, ans=10.0
2024-09-23 08:48:44,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=221256.0, ans=0.0
2024-09-23 08:48:49,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=221302.66666666666, ans=0.125
2024-09-23 08:49:06,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=221349.33333333334, ans=0.125
2024-09-23 08:49:15,684 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=6.99 vs. limit=15.0
2024-09-23 08:49:23,893 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.284e+02 1.397e+02 1.593e+02 2.300e+02, threshold=2.794e+02, percent-clipped=0.0
2024-09-23 08:49:25,055 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.94 vs. limit=6.0
2024-09-23 08:49:27,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=221396.0, ans=0.125
2024-09-23 08:49:35,960 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.10 vs. limit=15.0
2024-09-23 08:49:38,325 INFO [train.py:1198] (1/4) Epoch 13, batch 700, loss[loss=0.2577, ctc_loss=0.1739, cr_loss=0.419, over 17151.00 frames. ], tot_loss[loss=0.2373, ctc_loss=0.1626, cr_loss=0.3734, over 3248609.21 frames. ], batch size: 48, lr: 9.41e-03, grad_scale: 32.0
2024-09-23 08:49:51,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=221442.66666666666, ans=0.125
2024-09-23 08:49:55,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=221489.33333333334, ans=0.0
2024-09-23 08:49:58,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=221489.33333333334, ans=0.125
2024-09-23 08:50:06,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=221489.33333333334, ans=0.0
2024-09-23 08:50:08,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=221536.0, ans=0.0
2024-09-23 08:50:34,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.max_positive, batch_count=221582.66666666666, ans=0.95
2024-09-23 08:50:42,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=221629.33333333334, ans=0.125
2024-09-23 08:50:52,636 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.99 vs. limit=15.0
2024-09-23 08:50:59,837 INFO [train.py:1198] (1/4) Epoch 13, batch 750, loss[loss=0.2559, ctc_loss=0.1758, cr_loss=0.4003, over 14846.00 frames. ], tot_loss[loss=0.238, ctc_loss=0.1631, cr_loss=0.3745, over 3264500.51 frames. ], batch size: 89, lr: 9.40e-03, grad_scale: 16.0
2024-09-23 08:51:02,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.95 vs. limit=10.0
2024-09-23 08:51:12,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=221676.0, ans=0.125
2024-09-23 08:51:53,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=221816.0, ans=0.0
2024-09-23 08:52:11,978 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.258e+02 1.365e+02 1.480e+02 2.047e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-23 08:52:15,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=221862.66666666666, ans=0.09899494936611666
2024-09-23 08:52:24,631 INFO [train.py:1198] (1/4) Epoch 13, batch 800, loss[loss=0.2225, ctc_loss=0.1511, cr_loss=0.3572, over 16999.00 frames. ], tot_loss[loss=0.2364, ctc_loss=0.1619, cr_loss=0.3725, over 3292156.39 frames. ], batch size: 51, lr: 9.40e-03, grad_scale: 32.0
2024-09-23 08:52:29,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=221909.33333333334, ans=0.1
2024-09-23 08:52:39,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=221956.0, ans=0.125
2024-09-23 08:53:15,839 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.13 vs. limit=15.0
2024-09-23 08:53:23,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=222049.33333333334, ans=0.025
2024-09-23 08:53:28,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=222049.33333333334, ans=0.2
2024-09-23 08:53:37,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=222096.0, ans=0.2
2024-09-23 08:53:47,085 INFO [train.py:1198] (1/4) Epoch 13, batch 850, loss[loss=0.2366, ctc_loss=0.1582, cr_loss=0.3921, over 16077.00 frames. ], tot_loss[loss=0.2369, ctc_loss=0.1622, cr_loss=0.3737, over 3311348.05 frames. ], batch size: 74, lr: 9.39e-03, grad_scale: 32.0
2024-09-23 08:53:50,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=222142.66666666666, ans=0.2
2024-09-23 08:53:58,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=222142.66666666666, ans=0.1
2024-09-23 08:54:11,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=222189.33333333334, ans=0.125
2024-09-23 08:54:14,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=222189.33333333334, ans=0.1
2024-09-23 08:54:37,070 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 08:54:54,220 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.309e+02 1.415e+02 1.608e+02 2.192e+02, threshold=2.830e+02, percent-clipped=0.0
2024-09-23 08:55:06,969 INFO [train.py:1198] (1/4) Epoch 13, batch 900, loss[loss=0.1995, ctc_loss=0.1346, cr_loss=0.3243, over 16960.00 frames. ], tot_loss[loss=0.2381, ctc_loss=0.1629, cr_loss=0.3759, over 3322436.32 frames. ], batch size: 42, lr: 9.39e-03, grad_scale: 32.0
2024-09-23 08:55:36,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=222422.66666666666, ans=0.125
2024-09-23 08:55:36,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=222422.66666666666, ans=0.2
2024-09-23 08:55:52,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=222469.33333333334, ans=0.0
2024-09-23 08:56:13,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=222562.66666666666, ans=0.1
2024-09-23 08:56:32,008 INFO [train.py:1198] (1/4) Epoch 13, batch 950, loss[loss=0.2415, ctc_loss=0.1695, cr_loss=0.3597, over 17259.00 frames. ], tot_loss[loss=0.2375, ctc_loss=0.1623, cr_loss=0.3756, over 3340584.07 frames. ], batch size: 44, lr: 9.38e-03, grad_scale: 32.0
2024-09-23 08:56:59,736 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.50 vs. limit=15.0
2024-09-23 08:57:15,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=222702.66666666666, ans=0.0
2024-09-23 08:57:25,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=222749.33333333334, ans=0.125
2024-09-23 08:57:30,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=222749.33333333334, ans=0.025
2024-09-23 08:57:41,764 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.336e+02 1.446e+02 1.624e+02 2.624e+02, threshold=2.892e+02, percent-clipped=0.0
2024-09-23 08:57:54,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=222796.0, ans=0.1
2024-09-23 08:57:57,067 INFO [train.py:1198] (1/4) Epoch 13, batch 1000, loss[loss=0.2584, ctc_loss=0.1773, cr_loss=0.4055, over 16880.00 frames. ], tot_loss[loss=0.2367, ctc_loss=0.1618, cr_loss=0.3745, over 3352162.81 frames. ], batch size: 58, lr: 9.38e-03, grad_scale: 32.0
2024-09-23 08:58:15,053 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=4.97 vs. limit=12.0
2024-09-23 08:58:44,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=222982.66666666666, ans=0.0
2024-09-23 08:58:51,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=222982.66666666666, ans=0.125
2024-09-23 08:59:16,890 INFO [train.py:1198] (1/4) Epoch 13, batch 1050, loss[loss=0.2302, ctc_loss=0.1561, cr_loss=0.3705, over 17312.00 frames. ], tot_loss[loss=0.238, ctc_loss=0.1628, cr_loss=0.3759, over 3353668.68 frames. ], batch size: 49, lr: 9.37e-03, grad_scale: 32.0
2024-09-23 08:59:52,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=223169.33333333334, ans=0.125
2024-09-23 09:00:03,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.71 vs. limit=15.0
2024-09-23 09:00:26,978 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.277e+02 1.397e+02 1.576e+02 3.836e+02, threshold=2.794e+02, percent-clipped=1.0
2024-09-23 09:00:39,436 INFO [train.py:1198] (1/4) Epoch 13, batch 1100, loss[loss=0.2689, ctc_loss=0.1863, cr_loss=0.4127, over 17035.00 frames. ], tot_loss[loss=0.2379, ctc_loss=0.1627, cr_loss=0.3762, over 3357621.26 frames. ], batch size: 51, lr: 9.37e-03, grad_scale: 32.0
2024-09-23 09:00:44,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=223309.33333333334, ans=0.2
2024-09-23 09:00:45,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.max_abs, batch_count=223309.33333333334, ans=10.0
2024-09-23 09:00:51,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=223309.33333333334, ans=0.125
2024-09-23 09:00:52,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer_na.min_abs, batch_count=223309.33333333334, ans=0.02
2024-09-23 09:01:22,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=223402.66666666666, ans=0.0
2024-09-23 09:01:25,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=223449.33333333334, ans=0.125
2024-09-23 09:01:25,820 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:01:42,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=223449.33333333334, ans=0.0
2024-09-23 09:02:01,297 INFO [train.py:1198] (1/4) Epoch 13, batch 1150, loss[loss=0.24, ctc_loss=0.1615, cr_loss=0.3923, over 17152.00 frames. ], tot_loss[loss=0.2373, ctc_loss=0.1622, cr_loss=0.3754, over 3360473.89 frames. ], batch size: 48, lr: 9.37e-03, grad_scale: 32.0
2024-09-23 09:02:14,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=223542.66666666666, ans=0.1
2024-09-23 09:02:30,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer_ff3.min_abs, batch_count=223589.33333333334, ans=0.2
2024-09-23 09:02:33,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=223636.0, ans=0.025
2024-09-23 09:02:41,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=223636.0, ans=0.1
2024-09-23 09:03:11,339 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.252e+02 1.365e+02 1.487e+02 2.591e+02, threshold=2.731e+02, percent-clipped=0.0
2024-09-23 09:03:13,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=223729.33333333334, ans=0.025
2024-09-23 09:03:23,912 INFO [train.py:1198] (1/4) Epoch 13, batch 1200, loss[loss=0.2016, ctc_loss=0.1353, cr_loss=0.3315, over 17091.00 frames. ], tot_loss[loss=0.237, ctc_loss=0.1619, cr_loss=0.3757, over 3361313.22 frames. ], batch size: 40, lr: 9.36e-03, grad_scale: 32.0
2024-09-23 09:03:28,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=223776.0, ans=0.1
2024-09-23 09:03:39,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=223822.66666666666, ans=0.2
2024-09-23 09:03:51,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=223822.66666666666, ans=0.125
2024-09-23 09:03:58,467 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.37 vs. limit=10.0
2024-09-23 09:04:02,732 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.21 vs. limit=22.5
2024-09-23 09:04:33,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=223962.66666666666, ans=0.2
2024-09-23 09:04:43,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=223962.66666666666, ans=0.2
2024-09-23 09:04:46,170 INFO [train.py:1198] (1/4) Epoch 13, batch 1250, loss[loss=0.2401, ctc_loss=0.1624, cr_loss=0.3884, over 17285.00 frames. ], tot_loss[loss=0.2354, ctc_loss=0.1608, cr_loss=0.3733, over 3359866.43 frames. ], batch size: 46, lr: 9.36e-03, grad_scale: 32.0
2024-09-23 09:05:27,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=224102.66666666666, ans=0.025
2024-09-23 09:05:55,484 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.315e+02 1.381e+02 1.503e+02 2.464e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-23 09:06:08,065 INFO [train.py:1198] (1/4) Epoch 13, batch 1300, loss[loss=0.2089, ctc_loss=0.1399, cr_loss=0.3449, over 16959.00 frames. ], tot_loss[loss=0.2362, ctc_loss=0.1614, cr_loss=0.3741, over 3355103.41 frames. ], batch size: 42, lr: 9.35e-03, grad_scale: 32.0
2024-09-23 09:06:35,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=224289.33333333334, ans=0.125
2024-09-23 09:06:44,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=224336.0, ans=0.125
2024-09-23 09:06:57,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=224336.0, ans=0.125
2024-09-23 09:07:01,357 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.69 vs. limit=22.5
2024-09-23 09:07:19,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=224429.33333333334, ans=0.0
2024-09-23 09:07:21,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=224429.33333333334, ans=0.0
2024-09-23 09:07:32,574 INFO [train.py:1198] (1/4) Epoch 13, batch 1350, loss[loss=0.2306, ctc_loss=0.1569, cr_loss=0.3688, over 17062.00 frames. ], tot_loss[loss=0.2368, ctc_loss=0.1619, cr_loss=0.3746, over 3359074.28 frames. ], batch size: 46, lr: 9.35e-03, grad_scale: 32.0
2024-09-23 09:07:36,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=224476.0, ans=0.5
2024-09-23 09:07:37,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=224476.0, ans=0.125
2024-09-23 09:07:43,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=224476.0, ans=0.04949747468305833
2024-09-23 09:07:54,984 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.26 vs. limit=15.0
2024-09-23 09:08:41,710 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.260e+02 1.380e+02 1.508e+02 2.232e+02, threshold=2.759e+02, percent-clipped=0.0
2024-09-23 09:08:46,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=224662.66666666666, ans=0.1
2024-09-23 09:08:54,700 INFO [train.py:1198] (1/4) Epoch 13, batch 1400, loss[loss=0.2318, ctc_loss=0.1534, cr_loss=0.3919, over 17094.00 frames. ], tot_loss[loss=0.2368, ctc_loss=0.1618, cr_loss=0.3749, over 3362713.79 frames. ], batch size: 49, lr: 9.34e-03, grad_scale: 32.0
2024-09-23 09:08:58,837 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.42 vs. limit=15.0
2024-09-23 09:09:00,126 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.19 vs. limit=6.0
2024-09-23 09:09:09,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.75 vs. limit=10.0
2024-09-23 09:09:36,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=224802.66666666666, ans=0.125
2024-09-23 09:09:44,840 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.64 vs. limit=22.5
2024-09-23 09:09:50,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=224849.33333333334, ans=0.025
2024-09-23 09:10:17,269 INFO [train.py:1198] (1/4) Epoch 13, batch 1450, loss[loss=0.2381, ctc_loss=0.1568, cr_loss=0.4068, over 17363.00 frames. ], tot_loss[loss=0.2365, ctc_loss=0.1616, cr_loss=0.3744, over 3367221.31 frames. ], batch size: 48, lr: 9.34e-03, grad_scale: 32.0
2024-09-23 09:10:20,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=224942.66666666666, ans=0.0
2024-09-23 09:10:48,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=225036.0, ans=0.0
2024-09-23 09:10:54,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=225036.0, ans=0.125
2024-09-23 09:11:03,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.79 vs. limit=22.5
2024-09-23 09:11:28,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=225129.33333333334, ans=0.0
2024-09-23 09:11:29,648 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.289e+02 1.416e+02 1.573e+02 2.089e+02, threshold=2.832e+02, percent-clipped=0.0
2024-09-23 09:11:39,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=225129.33333333334, ans=0.125
2024-09-23 09:11:42,311 INFO [train.py:1198] (1/4) Epoch 13, batch 1500, loss[loss=0.2349, ctc_loss=0.1624, cr_loss=0.3625, over 17262.00 frames. ], tot_loss[loss=0.236, ctc_loss=0.1613, cr_loss=0.3737, over 3364295.85 frames. ], batch size: 44, lr: 9.33e-03, grad_scale: 32.0
2024-09-23 09:12:29,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=225316.0, ans=0.0
2024-09-23 09:13:05,142 INFO [train.py:1198] (1/4) Epoch 13, batch 1550, loss[loss=0.2397, ctc_loss=0.1659, cr_loss=0.3688, over 17212.00 frames. ], tot_loss[loss=0.2377, ctc_loss=0.1626, cr_loss=0.3751, over 3362186.53 frames. ], batch size: 47, lr: 9.33e-03, grad_scale: 32.0
2024-09-23 09:13:29,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=225456.0, ans=0.1
2024-09-23 09:13:37,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=225502.66666666666, ans=0.125
2024-09-23 09:14:12,645 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.248e+02 1.360e+02 1.586e+02 2.535e+02, threshold=2.720e+02, percent-clipped=0.0
2024-09-23 09:14:13,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=225596.0, ans=0.2
2024-09-23 09:14:20,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=225596.0, ans=0.2
2024-09-23 09:14:25,419 INFO [train.py:1198] (1/4) Epoch 13, batch 1600, loss[loss=0.2492, ctc_loss=0.172, cr_loss=0.386, over 15057.00 frames. ], tot_loss[loss=0.2376, ctc_loss=0.1626, cr_loss=0.3747, over 3359601.61 frames. ], batch size: 89, lr: 9.32e-03, grad_scale: 32.0
2024-09-23 09:14:37,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.91 vs. limit=10.0
2024-09-23 09:14:49,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=225689.33333333334, ans=0.04949747468305833
2024-09-23 09:15:07,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=225736.0, ans=0.125
2024-09-23 09:15:37,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=225829.33333333334, ans=0.0
2024-09-23 09:15:48,383 INFO [train.py:1198] (1/4) Epoch 13, batch 1650, loss[loss=0.2311, ctc_loss=0.1548, cr_loss=0.3811, over 17275.00 frames. ], tot_loss[loss=0.2363, ctc_loss=0.1617, cr_loss=0.3732, over 3358823.09 frames. ], batch size: 44, lr: 9.32e-03, grad_scale: 32.0
2024-09-23 09:16:05,672 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.70 vs. limit=6.0
2024-09-23 09:16:19,109 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.17 vs. limit=10.0
2024-09-23 09:16:32,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=225969.33333333334, ans=0.125
2024-09-23 09:16:38,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=225969.33333333334, ans=0.125
2024-09-23 09:16:55,072 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.01 vs. limit=10.0
2024-09-23 09:17:00,540 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.262e+02 1.363e+02 1.503e+02 2.167e+02, threshold=2.727e+02, percent-clipped=0.0
2024-09-23 09:17:02,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=226062.66666666666, ans=0.1
2024-09-23 09:17:13,270 INFO [train.py:1198] (1/4) Epoch 13, batch 1700, loss[loss=0.2374, ctc_loss=0.1607, cr_loss=0.3835, over 17067.00 frames. ], tot_loss[loss=0.2358, ctc_loss=0.1612, cr_loss=0.3729, over 3361533.71 frames. ], batch size: 46, lr: 9.31e-03, grad_scale: 32.0
2024-09-23 09:17:14,119 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=5.92 vs. limit=15.0
2024-09-23 09:17:41,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=226156.0, ans=0.1
2024-09-23 09:17:43,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=226156.0, ans=0.125
2024-09-23 09:18:08,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=226249.33333333334, ans=0.125
2024-09-23 09:18:13,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=226249.33333333334, ans=0.025
2024-09-23 09:18:35,887 INFO [train.py:1198] (1/4) Epoch 13, batch 1750, loss[loss=0.2449, ctc_loss=0.1668, cr_loss=0.3908, over 17101.00 frames. ], tot_loss[loss=0.2357, ctc_loss=0.1612, cr_loss=0.3727, over 3355210.74 frames. ], batch size: 49, lr: 9.31e-03, grad_scale: 32.0
2024-09-23 09:18:45,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=226342.66666666666, ans=0.1
2024-09-23 09:18:46,079 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.84 vs. limit=6.0
2024-09-23 09:18:48,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=226342.66666666666, ans=0.125
2024-09-23 09:18:51,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=226389.33333333334, ans=0.0
2024-09-23 09:18:56,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=226389.33333333334, ans=0.125
2024-09-23 09:19:03,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=226389.33333333334, ans=0.125
2024-09-23 09:19:11,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.91 vs. limit=10.0
2024-09-23 09:19:25,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=226482.66666666666, ans=0.0
2024-09-23 09:19:42,591 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.255e+02 1.356e+02 1.518e+02 2.113e+02, threshold=2.712e+02, percent-clipped=0.0
2024-09-23 09:19:42,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=226529.33333333334, ans=0.0
2024-09-23 09:19:48,282 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.41 vs. limit=15.0
2024-09-23 09:19:55,546 INFO [train.py:1198] (1/4) Epoch 13, batch 1800, loss[loss=0.2086, ctc_loss=0.1414, cr_loss=0.3363, over 17325.00 frames. ], tot_loss[loss=0.236, ctc_loss=0.1614, cr_loss=0.3728, over 3359107.74 frames. ], batch size: 46, lr: 9.30e-03, grad_scale: 32.0
2024-09-23 09:19:59,480 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.39 vs. limit=15.0
2024-09-23 09:20:49,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=226716.0, ans=0.0
2024-09-23 09:21:06,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=226762.66666666666, ans=0.125
2024-09-23 09:21:22,738 INFO [train.py:1198] (1/4) Epoch 13, batch 1850, loss[loss=0.2295, ctc_loss=0.1568, cr_loss=0.3634, over 17153.00 frames. ], tot_loss[loss=0.2377, ctc_loss=0.1627, cr_loss=0.375, over 3358259.71 frames. ], batch size: 45, lr: 9.30e-03, grad_scale: 32.0
2024-09-23 09:21:46,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=226856.0, ans=0.125
2024-09-23 09:21:53,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=226902.66666666666, ans=0.125
2024-09-23 09:22:14,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=226949.33333333334, ans=0.0
2024-09-23 09:22:28,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=226996.0, ans=0.125
2024-09-23 09:22:29,433 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.309e+02 1.425e+02 1.783e+02 3.519e+02, threshold=2.851e+02, percent-clipped=1.0
2024-09-23 09:22:44,629 INFO [train.py:1198] (1/4) Epoch 13, batch 1900, loss[loss=0.2052, ctc_loss=0.1369, cr_loss=0.3416, over 17176.00 frames. ], tot_loss[loss=0.2383, ctc_loss=0.1631, cr_loss=0.3762, over 3359895.18 frames. ], batch size: 41, lr: 9.29e-03, grad_scale: 32.0
2024-09-23 09:23:04,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=227089.33333333334, ans=0.125
2024-09-23 09:23:04,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=227089.33333333334, ans=0.125
2024-09-23 09:23:18,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=227136.0, ans=0.0
2024-09-23 09:23:31,953 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.76 vs. limit=12.0
2024-09-23 09:23:55,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=227229.33333333334, ans=0.1
2024-09-23 09:24:04,047 INFO [train.py:1198] (1/4) Epoch 13, batch 1950, loss[loss=0.1936, ctc_loss=0.1291, cr_loss=0.3226, over 17039.00 frames. ], tot_loss[loss=0.2371, ctc_loss=0.1622, cr_loss=0.3744, over 3350789.94 frames. ], batch size: 39, lr: 9.29e-03, grad_scale: 32.0
2024-09-23 09:24:05,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=227276.0, ans=0.125
2024-09-23 09:24:07,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=227276.0, ans=0.125
2024-09-23 09:24:33,006 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:24:40,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.67 vs. limit=15.0
2024-09-23 09:24:50,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=227416.0, ans=0.125
2024-09-23 09:25:13,356 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.325e+02 1.427e+02 1.553e+02 2.368e+02, threshold=2.853e+02, percent-clipped=0.0
2024-09-23 09:25:19,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=227462.66666666666, ans=0.125
2024-09-23 09:25:24,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=227509.33333333334, ans=0.1
2024-09-23 09:25:25,898 INFO [train.py:1198] (1/4) Epoch 13, batch 2000, loss[loss=0.2286, ctc_loss=0.1509, cr_loss=0.3886, over 17199.00 frames. ], tot_loss[loss=0.2375, ctc_loss=0.1625, cr_loss=0.3753, over 3354745.77 frames. ], batch size: 41, lr: 9.29e-03, grad_scale: 32.0
2024-09-23 09:25:29,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=227509.33333333334, ans=0.025
2024-09-23 09:25:50,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=227556.0, ans=0.125
2024-09-23 09:26:44,607 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:26:51,000 INFO [train.py:1198] (1/4) Epoch 13, batch 2050, loss[loss=0.2544, ctc_loss=0.1764, cr_loss=0.3902, over 16998.00 frames. ], tot_loss[loss=0.2367, ctc_loss=0.1618, cr_loss=0.3748, over 3360917.47 frames. ], batch size: 51, lr: 9.28e-03, grad_scale: 16.0
2024-09-23 09:27:49,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=227882.66666666666, ans=0.2
2024-09-23 09:28:01,913 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.246e+02 1.339e+02 1.445e+02 2.585e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-23 09:28:13,095 INFO [train.py:1198] (1/4) Epoch 13, batch 2100, loss[loss=0.2306, ctc_loss=0.1565, cr_loss=0.3702, over 17013.00 frames. ], tot_loss[loss=0.2351, ctc_loss=0.1605, cr_loss=0.3731, over 3369173.03 frames. ], batch size: 44, lr: 9.28e-03, grad_scale: 16.0
2024-09-23 09:28:50,788 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=16.82 vs. limit=22.5
2024-09-23 09:28:54,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=228069.33333333334, ans=0.0
2024-09-23 09:29:02,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=228116.0, ans=0.125
2024-09-23 09:29:12,381 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:29:22,341 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.12 vs. limit=15.0
2024-09-23 09:29:32,533 INFO [train.py:1198] (1/4) Epoch 13, batch 2150, loss[loss=0.2719, ctc_loss=0.1895, cr_loss=0.412, over 16983.00 frames. ], tot_loss[loss=0.2368, ctc_loss=0.1618, cr_loss=0.3751, over 3366383.15 frames. ], batch size: 56, lr: 9.27e-03, grad_scale: 16.0
2024-09-23 09:29:35,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=228209.33333333334, ans=0.1
2024-09-23 09:29:58,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=228256.0, ans=0.5
2024-09-23 09:30:15,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=228302.66666666666, ans=0.1
2024-09-23 09:30:21,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=228349.33333333334, ans=0.05
2024-09-23 09:30:23,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=228349.33333333334, ans=0.1
2024-09-23 09:30:23,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=228349.33333333334, ans=0.2
2024-09-23 09:30:29,749 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.03 vs. limit=15.0
2024-09-23 09:30:43,654 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.254e+02 1.312e+02 1.458e+02 2.181e+02, threshold=2.624e+02, percent-clipped=0.0
2024-09-23 09:30:54,886 INFO [train.py:1198] (1/4) Epoch 13, batch 2200, loss[loss=0.1978, ctc_loss=0.1295, cr_loss=0.3418, over 17024.00 frames. ], tot_loss[loss=0.2375, ctc_loss=0.1623, cr_loss=0.3762, over 3371425.54 frames. ], batch size: 39, lr: 9.27e-03, grad_scale: 16.0
2024-09-23 09:30:55,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=228442.66666666666, ans=0.1
2024-09-23 09:31:32,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=228536.0, ans=0.0
2024-09-23 09:31:35,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=228536.0, ans=0.125
2024-09-23 09:32:19,399 INFO [train.py:1198] (1/4) Epoch 13, batch 2250, loss[loss=0.2025, ctc_loss=0.1339, cr_loss=0.3431, over 17184.00 frames. ], tot_loss[loss=0.2377, ctc_loss=0.1626, cr_loss=0.3757, over 3360970.46 frames. ], batch size: 41, lr: 9.26e-03, grad_scale: 16.0
2024-09-23 09:32:21,617 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.42 vs. limit=6.0
2024-09-23 09:32:22,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.73 vs. limit=22.5
2024-09-23 09:32:34,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=228676.0, ans=0.125
2024-09-23 09:33:00,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=228769.33333333334, ans=0.0
2024-09-23 09:33:19,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=228816.0, ans=0.125
2024-09-23 09:33:30,214 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.274e+02 1.381e+02 1.487e+02 1.972e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-23 09:33:41,416 INFO [train.py:1198] (1/4) Epoch 13, batch 2300, loss[loss=0.2674, ctc_loss=0.1828, cr_loss=0.4233, over 17006.00 frames. ], tot_loss[loss=0.2375, ctc_loss=0.1625, cr_loss=0.3755, over 3358150.56 frames. ], batch size: 53, lr: 9.26e-03, grad_scale: 16.0
2024-09-23 09:34:01,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=228956.0, ans=0.0
2024-09-23 09:34:01,539 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.09 vs. limit=15.0
2024-09-23 09:34:07,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=228956.0, ans=0.125
2024-09-23 09:34:17,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=229002.66666666666, ans=0.125
2024-09-23 09:34:18,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=229002.66666666666, ans=0.0
2024-09-23 09:34:32,033 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.07 vs. limit=10.0
2024-09-23 09:34:39,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=229049.33333333334, ans=0.125
2024-09-23 09:34:39,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=229049.33333333334, ans=0.0
2024-09-23 09:34:41,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=229049.33333333334, ans=0.125
2024-09-23 09:34:41,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=229049.33333333334, ans=0.2
2024-09-23 09:34:46,574 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.59 vs. limit=22.5
2024-09-23 09:35:01,961 INFO [train.py:1198] (1/4) Epoch 13, batch 2350, loss[loss=0.2557, ctc_loss=0.174, cr_loss=0.4083, over 17151.00 frames. ], tot_loss[loss=0.2365, ctc_loss=0.1617, cr_loss=0.3738, over 3351366.47 frames. ], batch size: 48, lr: 9.25e-03, grad_scale: 16.0
2024-09-23 09:35:05,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=229142.66666666666, ans=0.125
2024-09-23 09:35:21,336 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.96 vs. limit=10.0
2024-09-23 09:35:30,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=229189.33333333334, ans=0.09899494936611666
2024-09-23 09:35:35,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=229236.0, ans=0.125
2024-09-23 09:35:51,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=229282.66666666666, ans=0.125
2024-09-23 09:36:15,725 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.316e+02 1.425e+02 1.605e+02 2.479e+02, threshold=2.851e+02, percent-clipped=0.0
2024-09-23 09:36:18,635 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.52 vs. limit=8.0
2024-09-23 09:36:25,800 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:36:27,010 INFO [train.py:1198] (1/4) Epoch 13, batch 2400, loss[loss=0.252, ctc_loss=0.1748, cr_loss=0.3858, over 17050.00 frames. ], tot_loss[loss=0.2361, ctc_loss=0.1614, cr_loss=0.3733, over 3349092.23 frames. ], batch size: 52, lr: 9.25e-03, grad_scale: 32.0
2024-09-23 09:36:48,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.93 vs. limit=6.0
2024-09-23 09:36:54,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=229422.66666666666, ans=0.0
2024-09-23 09:37:05,791 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:37:07,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=229469.33333333334, ans=0.0
2024-09-23 09:37:10,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=229469.33333333334, ans=0.1
2024-09-23 09:37:10,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=229469.33333333334, ans=0.0
2024-09-23 09:37:30,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=229516.0, ans=0.2
2024-09-23 09:37:41,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=229562.66666666666, ans=0.0
2024-09-23 09:37:49,397 INFO [train.py:1198] (1/4) Epoch 13, batch 2450, loss[loss=0.2344, ctc_loss=0.1598, cr_loss=0.3732, over 17244.00 frames. ], tot_loss[loss=0.2365, ctc_loss=0.1616, cr_loss=0.3746, over 3348485.93 frames. ], batch size: 55, lr: 9.24e-03, grad_scale: 32.0
2024-09-23 09:37:51,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=229609.33333333334, ans=0.125
2024-09-23 09:37:56,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=229609.33333333334, ans=0.2
2024-09-23 09:38:18,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=229656.0, ans=0.2
2024-09-23 09:38:58,204 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.292e+02 1.402e+02 1.572e+02 2.224e+02, threshold=2.803e+02, percent-clipped=0.0
2024-09-23 09:39:09,494 INFO [train.py:1198] (1/4) Epoch 13, batch 2500, loss[loss=0.2135, ctc_loss=0.1451, cr_loss=0.342, over 16666.00 frames. ], tot_loss[loss=0.2361, ctc_loss=0.1612, cr_loss=0.3742, over 3348255.36 frames. ], batch size: 61, lr: 9.24e-03, grad_scale: 32.0
2024-09-23 09:39:24,789 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.81 vs. limit=22.5
2024-09-23 09:39:50,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=229936.0, ans=0.125
2024-09-23 09:40:00,316 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.07 vs. limit=10.0
2024-09-23 09:40:00,360 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.82 vs. limit=15.0
2024-09-23 09:40:13,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=229982.66666666666, ans=0.0
2024-09-23 09:40:22,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=230029.33333333334, ans=0.0
2024-09-23 09:40:32,033 INFO [train.py:1198] (1/4) Epoch 13, batch 2550, loss[loss=0.1921, ctc_loss=0.1277, cr_loss=0.322, over 17098.00 frames. ], tot_loss[loss=0.2362, ctc_loss=0.1614, cr_loss=0.3744, over 3350568.99 frames. ], batch size: 43, lr: 9.23e-03, grad_scale: 32.0
2024-09-23 09:40:42,440 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.20 vs. limit=15.0
2024-09-23 09:40:55,292 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.29 vs. limit=15.0
2024-09-23 09:41:08,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=230169.33333333334, ans=0.125
2024-09-23 09:41:11,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=230169.33333333334, ans=0.125
2024-09-23 09:41:34,097 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.11 vs. limit=15.0
2024-09-23 09:41:45,894 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.282e+02 1.430e+02 1.589e+02 2.134e+02, threshold=2.861e+02, percent-clipped=0.0
2024-09-23 09:41:57,176 INFO [train.py:1198] (1/4) Epoch 13, batch 2600, loss[loss=0.2667, ctc_loss=0.1865, cr_loss=0.4009, over 15129.00 frames. ], tot_loss[loss=0.2362, ctc_loss=0.1613, cr_loss=0.3743, over 3357655.32 frames. ], batch size: 89, lr: 9.23e-03, grad_scale: 32.0
2024-09-23 09:42:08,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=230309.33333333334, ans=0.0
2024-09-23 09:42:25,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=230356.0, ans=0.1
2024-09-23 09:42:35,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=230402.66666666666, ans=0.0
2024-09-23 09:43:06,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.02 vs. limit=15.0
2024-09-23 09:43:11,456 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.80 vs. limit=15.0
2024-09-23 09:43:20,133 INFO [train.py:1198] (1/4) Epoch 13, batch 2650, loss[loss=0.2619, ctc_loss=0.1779, cr_loss=0.42, over 16986.00 frames. ], tot_loss[loss=0.2351, ctc_loss=0.1605, cr_loss=0.373, over 3354248.90 frames. ], batch size: 53, lr: 9.23e-03, grad_scale: 32.0
2024-09-23 09:43:32,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=230542.66666666666, ans=0.2
2024-09-23 09:43:38,366 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.78 vs. limit=15.0
2024-09-23 09:43:39,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=230589.33333333334, ans=0.125
2024-09-23 09:43:46,838 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.79 vs. limit=15.0
2024-09-23 09:43:49,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=230589.33333333334, ans=0.125
2024-09-23 09:43:51,175 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.06 vs. limit=22.5
2024-09-23 09:44:02,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=230636.0, ans=0.0
2024-09-23 09:44:05,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=230636.0, ans=0.125
2024-09-23 09:44:08,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=230682.66666666666, ans=0.0
2024-09-23 09:44:12,183 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.48 vs. limit=22.5
2024-09-23 09:44:28,570 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.270e+02 1.366e+02 1.498e+02 2.280e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-23 09:44:33,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=230729.33333333334, ans=0.2
2024-09-23 09:44:39,786 INFO [train.py:1198] (1/4) Epoch 13, batch 2700, loss[loss=0.2212, ctc_loss=0.1489, cr_loss=0.3612, over 17082.00 frames. ], tot_loss[loss=0.2349, ctc_loss=0.1604, cr_loss=0.3726, over 3353741.16 frames. ], batch size: 39, lr: 9.22e-03, grad_scale: 32.0
2024-09-23 09:44:46,920 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.01 vs. limit=15.0
2024-09-23 09:44:56,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=230822.66666666666, ans=0.025
2024-09-23 09:45:19,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=230869.33333333334, ans=0.0
2024-09-23 09:45:24,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=230869.33333333334, ans=0.125
2024-09-23 09:46:02,466 INFO [train.py:1198] (1/4) Epoch 13, batch 2750, loss[loss=0.278, ctc_loss=0.192, cr_loss=0.4298, over 16938.00 frames. ], tot_loss[loss=0.2366, ctc_loss=0.1617, cr_loss=0.3743, over 3352505.86 frames. ], batch size: 58, lr: 9.22e-03, grad_scale: 32.0
2024-09-23 09:46:17,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.min_positive, batch_count=231009.33333333334, ans=0.025
2024-09-23 09:47:19,274 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.267e+02 1.386e+02 1.565e+02 1.913e+02, threshold=2.771e+02, percent-clipped=0.0
2024-09-23 09:47:30,586 INFO [train.py:1198] (1/4) Epoch 13, batch 2800, loss[loss=0.3224, ctc_loss=0.2356, cr_loss=0.4342, over 11943.00 frames. ], tot_loss[loss=0.2368, ctc_loss=0.1621, cr_loss=0.3737, over 3342613.19 frames. ], batch size: 124, lr: 9.21e-03, grad_scale: 32.0
2024-09-23 09:47:51,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=231289.33333333334, ans=0.0
2024-09-23 09:48:04,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=231336.0, ans=0.0
2024-09-23 09:48:13,560 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:48:50,268 INFO [train.py:1198] (1/4) Epoch 13, batch 2850, loss[loss=0.201, ctc_loss=0.1329, cr_loss=0.3407, over 17197.00 frames. ], tot_loss[loss=0.2369, ctc_loss=0.1621, cr_loss=0.3738, over 3336506.83 frames. ], batch size: 41, lr: 9.21e-03, grad_scale: 16.0
2024-09-23 09:49:07,985 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:49:08,350 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.23 vs. limit=15.0
2024-09-23 09:49:09,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=231522.66666666666, ans=0.1
2024-09-23 09:49:16,220 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=15.28 vs. limit=15.0
2024-09-23 09:49:19,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=231522.66666666666, ans=0.0
2024-09-23 09:50:00,333 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.288e+02 1.395e+02 1.565e+02 5.212e+02, threshold=2.790e+02, percent-clipped=1.0
2024-09-23 09:50:00,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=231662.66666666666, ans=0.125
2024-09-23 09:50:12,523 INFO [train.py:1198] (1/4) Epoch 13, batch 2900, loss[loss=0.254, ctc_loss=0.1706, cr_loss=0.4172, over 17213.00 frames. ], tot_loss[loss=0.2361, ctc_loss=0.1613, cr_loss=0.374, over 3349113.26 frames. ], batch size: 55, lr: 9.20e-03, grad_scale: 16.0
2024-09-23 09:50:40,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=231756.0, ans=0.0
2024-09-23 09:50:40,858 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.02 vs. limit=10.0
2024-09-23 09:51:35,581 INFO [train.py:1198] (1/4) Epoch 13, batch 2950, loss[loss=0.1937, ctc_loss=0.1295, cr_loss=0.3208, over 16671.00 frames. ], tot_loss[loss=0.235, ctc_loss=0.1604, cr_loss=0.3728, over 3358763.94 frames. ], batch size: 37, lr: 9.20e-03, grad_scale: 16.0
2024-09-23 09:51:40,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=231942.66666666666, ans=0.2
2024-09-23 09:51:53,664 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 09:52:47,838 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.241e+02 1.342e+02 1.458e+02 2.416e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-23 09:52:57,272 INFO [train.py:1198] (1/4) Epoch 13, batch 3000, loss[loss=0.2956, ctc_loss=0.2163, cr_loss=0.3967, over 11731.00 frames. ], tot_loss[loss=0.2346, ctc_loss=0.1601, cr_loss=0.3723, over 3358726.35 frames. ], batch size: 123, lr: 9.19e-03, grad_scale: 16.0
2024-09-23 09:52:57,272 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 09:53:12,985 INFO [train.py:1230] (1/4) Epoch 13, validation: loss=0.04424, ctc_loss=0.04424, cr_loss=7.269e-15, over 944034.00 frames. 
2024-09-23 09:53:12,986 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 09:53:18,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=232176.0, ans=0.025
2024-09-23 09:53:31,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=232222.66666666666, ans=0.0
2024-09-23 09:53:35,673 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=22.07 vs. limit=22.5
2024-09-23 09:53:36,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=232222.66666666666, ans=0.1
2024-09-23 09:54:03,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=232316.0, ans=0.125
2024-09-23 09:54:31,494 INFO [train.py:1198] (1/4) Epoch 13, batch 3050, loss[loss=0.2709, ctc_loss=0.1872, cr_loss=0.4183, over 17049.00 frames. ], tot_loss[loss=0.2346, ctc_loss=0.1602, cr_loss=0.3722, over 3364526.89 frames. ], batch size: 52, lr: 9.19e-03, grad_scale: 16.0
2024-09-23 09:54:47,931 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.44 vs. limit=22.5
2024-09-23 09:55:15,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=232502.66666666666, ans=0.125
2024-09-23 09:55:23,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=232549.33333333334, ans=0.125
2024-09-23 09:55:29,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=232549.33333333334, ans=0.125
2024-09-23 09:55:30,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=232549.33333333334, ans=0.0
2024-09-23 09:55:39,980 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.372e+02 1.503e+02 1.600e+02 2.658e+02, threshold=3.005e+02, percent-clipped=0.0
2024-09-23 09:55:49,567 INFO [train.py:1198] (1/4) Epoch 13, batch 3100, loss[loss=0.2717, ctc_loss=0.192, cr_loss=0.3989, over 17232.00 frames. ], tot_loss[loss=0.2335, ctc_loss=0.1592, cr_loss=0.3715, over 3370035.29 frames. ], batch size: 50, lr: 9.18e-03, grad_scale: 16.0
2024-09-23 09:55:59,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=232642.66666666666, ans=0.1
2024-09-23 09:56:10,640 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.07 vs. limit=15.0
2024-09-23 09:56:50,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=232782.66666666666, ans=0.0
2024-09-23 09:57:10,793 INFO [train.py:1198] (1/4) Epoch 13, batch 3150, loss[loss=0.2222, ctc_loss=0.1507, cr_loss=0.3573, over 17344.00 frames. ], tot_loss[loss=0.2348, ctc_loss=0.1602, cr_loss=0.3731, over 3375412.91 frames. ], batch size: 48, lr: 9.18e-03, grad_scale: 16.0
2024-09-23 09:57:44,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.74 vs. limit=15.0
2024-09-23 09:57:58,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.54 vs. limit=22.5
2024-09-23 09:57:59,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=233016.0, ans=0.1
2024-09-23 09:58:04,699 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.56 vs. limit=6.0
2024-09-23 09:58:19,650 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.308e+02 1.468e+02 1.678e+02 2.567e+02, threshold=2.937e+02, percent-clipped=0.0
2024-09-23 09:58:26,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=233062.66666666666, ans=0.125
2024-09-23 09:58:28,988 INFO [train.py:1198] (1/4) Epoch 13, batch 3200, loss[loss=0.2141, ctc_loss=0.1451, cr_loss=0.345, over 17361.00 frames. ], tot_loss[loss=0.2353, ctc_loss=0.1605, cr_loss=0.3738, over 3373024.46 frames. ], batch size: 48, lr: 9.18e-03, grad_scale: 32.0
2024-09-23 09:58:30,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=233109.33333333334, ans=0.1
2024-09-23 09:58:37,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=233109.33333333334, ans=0.0
2024-09-23 09:58:40,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=233109.33333333334, ans=0.0
2024-09-23 09:58:50,180 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.45 vs. limit=6.0
2024-09-23 09:59:05,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=233202.66666666666, ans=0.025
2024-09-23 09:59:14,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=233202.66666666666, ans=0.0
2024-09-23 09:59:19,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=233249.33333333334, ans=0.2
2024-09-23 09:59:47,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=233296.0, ans=0.2
2024-09-23 09:59:51,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=233342.66666666666, ans=0.0
2024-09-23 09:59:52,148 INFO [train.py:1198] (1/4) Epoch 13, batch 3250, loss[loss=0.2217, ctc_loss=0.1512, cr_loss=0.3527, over 17142.00 frames. ], tot_loss[loss=0.2364, ctc_loss=0.1615, cr_loss=0.3745, over 3365619.15 frames. ], batch size: 48, lr: 9.17e-03, grad_scale: 32.0
2024-09-23 10:00:05,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=233342.66666666666, ans=0.125
2024-09-23 10:00:09,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=233389.33333333334, ans=0.1
2024-09-23 10:00:36,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=233436.0, ans=0.125
2024-09-23 10:00:50,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=233482.66666666666, ans=0.125
2024-09-23 10:00:56,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=233529.33333333334, ans=0.125
2024-09-23 10:01:01,302 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.287e+02 1.416e+02 1.581e+02 2.137e+02, threshold=2.832e+02, percent-clipped=0.0
2024-09-23 10:01:02,045 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.12 vs. limit=6.0
2024-09-23 10:01:09,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=233576.0, ans=0.125
2024-09-23 10:01:10,547 INFO [train.py:1198] (1/4) Epoch 13, batch 3300, loss[loss=0.2501, ctc_loss=0.1692, cr_loss=0.4047, over 17019.00 frames. ], tot_loss[loss=0.2351, ctc_loss=0.1604, cr_loss=0.3733, over 3372638.97 frames. ], batch size: 53, lr: 9.17e-03, grad_scale: 32.0
2024-09-23 10:01:12,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=233576.0, ans=0.1
2024-09-23 10:01:16,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=233576.0, ans=0.1
2024-09-23 10:01:18,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=233576.0, ans=0.0
2024-09-23 10:01:23,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=233576.0, ans=0.1
2024-09-23 10:01:31,328 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=14.07 vs. limit=15.0
2024-09-23 10:01:35,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=233622.66666666666, ans=0.125
2024-09-23 10:02:05,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=233716.0, ans=0.1
2024-09-23 10:02:11,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=233716.0, ans=0.125
2024-09-23 10:02:30,430 INFO [train.py:1198] (1/4) Epoch 13, batch 3350, loss[loss=0.2306, ctc_loss=0.1564, cr_loss=0.3709, over 16904.00 frames. ], tot_loss[loss=0.2354, ctc_loss=0.1608, cr_loss=0.3729, over 3361678.17 frames. ], batch size: 58, lr: 9.16e-03, grad_scale: 32.0
2024-09-23 10:02:35,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=233809.33333333334, ans=0.125
2024-09-23 10:02:36,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=233809.33333333334, ans=0.04949747468305833
2024-09-23 10:02:47,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=233856.0, ans=0.0
2024-09-23 10:02:47,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=233856.0, ans=0.1
2024-09-23 10:02:49,788 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.66 vs. limit=15.0
2024-09-23 10:03:04,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=233902.66666666666, ans=0.125
2024-09-23 10:03:15,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=233949.33333333334, ans=0.1
2024-09-23 10:03:30,488 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.94 vs. limit=12.0
2024-09-23 10:03:34,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=233996.0, ans=0.025
2024-09-23 10:03:38,963 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.272e+02 1.377e+02 1.520e+02 2.187e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-23 10:03:46,198 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.16 vs. limit=15.0
2024-09-23 10:03:48,297 INFO [train.py:1198] (1/4) Epoch 13, batch 3400, loss[loss=0.2137, ctc_loss=0.1485, cr_loss=0.326, over 17238.00 frames. ], tot_loss[loss=0.235, ctc_loss=0.1604, cr_loss=0.3728, over 3350811.00 frames. ], batch size: 44, lr: 9.16e-03, grad_scale: 32.0
2024-09-23 10:03:49,059 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.39 vs. limit=15.0
2024-09-23 10:03:53,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=234042.66666666666, ans=0.0
2024-09-23 10:03:55,127 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.64 vs. limit=15.0
2024-09-23 10:03:59,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=234042.66666666666, ans=0.0
2024-09-23 10:03:59,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=234042.66666666666, ans=0.09899494936611666
2024-09-23 10:04:24,658 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.23 vs. limit=15.0
2024-09-23 10:04:35,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=234182.66666666666, ans=0.0
2024-09-23 10:04:37,059 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 10:04:38,620 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 10:04:44,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys.whitening_limit, batch_count=234182.66666666666, ans=6.0
2024-09-23 10:04:46,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=234182.66666666666, ans=0.125
2024-09-23 10:05:06,442 INFO [train.py:1198] (1/4) Epoch 13, batch 3450, loss[loss=0.2115, ctc_loss=0.1426, cr_loss=0.3448, over 17104.00 frames. ], tot_loss[loss=0.2348, ctc_loss=0.1603, cr_loss=0.3724, over 3350836.36 frames. ], batch size: 43, lr: 9.15e-03, grad_scale: 32.0
2024-09-23 10:05:17,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=234276.0, ans=0.0
2024-09-23 10:05:22,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=234322.66666666666, ans=0.125
2024-09-23 10:05:23,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=234322.66666666666, ans=0.1
2024-09-23 10:06:15,807 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.304e+02 1.412e+02 1.643e+02 2.368e+02, threshold=2.824e+02, percent-clipped=0.0
2024-09-23 10:06:22,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=234462.66666666666, ans=0.125
2024-09-23 10:06:25,215 INFO [train.py:1198] (1/4) Epoch 13, batch 3500, loss[loss=0.1864, ctc_loss=0.1219, cr_loss=0.3226, over 17205.00 frames. ], tot_loss[loss=0.235, ctc_loss=0.1605, cr_loss=0.3725, over 3351665.08 frames. ], batch size: 41, lr: 9.15e-03, grad_scale: 32.0
2024-09-23 10:07:21,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=234649.33333333334, ans=0.125
2024-09-23 10:07:45,894 INFO [train.py:1198] (1/4) Epoch 13, batch 3550, loss[loss=0.2623, ctc_loss=0.1795, cr_loss=0.4139, over 17026.00 frames. ], tot_loss[loss=0.2367, ctc_loss=0.1617, cr_loss=0.3752, over 3348865.87 frames. ], batch size: 56, lr: 9.14e-03, grad_scale: 32.0
2024-09-23 10:07:56,001 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.88 vs. limit=15.0
2024-09-23 10:08:09,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=234789.33333333334, ans=0.125
2024-09-23 10:08:39,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=234882.66666666666, ans=0.1
2024-09-23 10:09:00,083 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.323e+02 1.492e+02 1.691e+02 2.949e+02, threshold=2.984e+02, percent-clipped=2.0
2024-09-23 10:09:03,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=234929.33333333334, ans=0.0
2024-09-23 10:09:04,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=234929.33333333334, ans=0.125
2024-09-23 10:09:07,583 INFO [train.py:1198] (1/4) Epoch 13, batch 3600, loss[loss=0.2414, ctc_loss=0.1676, cr_loss=0.3691, over 16750.00 frames. ], tot_loss[loss=0.2355, ctc_loss=0.1607, cr_loss=0.3737, over 3361321.22 frames. ], batch size: 61, lr: 9.14e-03, grad_scale: 32.0
2024-09-23 10:09:45,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=235069.33333333334, ans=0.125
2024-09-23 10:10:04,020 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 10:10:04,658 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=8.69 vs. limit=15.0
2024-09-23 10:10:08,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=235162.66666666666, ans=0.125
2024-09-23 10:10:16,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=235162.66666666666, ans=0.2
2024-09-23 10:10:17,117 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.16 vs. limit=22.5
2024-09-23 10:10:25,895 INFO [train.py:1198] (1/4) Epoch 13, batch 3650, loss[loss=0.2179, ctc_loss=0.1433, cr_loss=0.3727, over 17307.00 frames. ], tot_loss[loss=0.2365, ctc_loss=0.1616, cr_loss=0.3749, over 3356184.64 frames. ], batch size: 51, lr: 9.14e-03, grad_scale: 32.0
2024-09-23 10:10:38,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=235209.33333333334, ans=0.2
2024-09-23 10:10:49,560 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 10:10:58,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=235302.66666666666, ans=0.125
2024-09-23 10:11:14,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=235349.33333333334, ans=0.015
2024-09-23 10:11:35,447 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.28 vs. limit=15.0
2024-09-23 10:11:37,917 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.270e+02 1.340e+02 1.472e+02 2.759e+02, threshold=2.681e+02, percent-clipped=0.0
2024-09-23 10:11:38,851 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.82 vs. limit=15.0
2024-09-23 10:11:39,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=235396.0, ans=0.125
2024-09-23 10:11:41,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=235396.0, ans=0.2
2024-09-23 10:11:45,789 INFO [train.py:1198] (1/4) Epoch 13, batch 3700, loss[loss=0.2724, ctc_loss=0.183, cr_loss=0.4469, over 17187.00 frames. ], tot_loss[loss=0.2366, ctc_loss=0.1616, cr_loss=0.3753, over 3361477.83 frames. ], batch size: 47, lr: 9.13e-03, grad_scale: 32.0
2024-09-23 10:11:46,947 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=6.76 vs. limit=15.0
2024-09-23 10:11:47,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=235442.66666666666, ans=0.125
2024-09-23 10:12:04,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=235489.33333333334, ans=0.0
2024-09-23 10:12:29,100 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.78 vs. limit=15.0
2024-09-23 10:12:29,121 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.76 vs. limit=15.0
2024-09-23 10:12:50,996 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.93 vs. limit=15.0
2024-09-23 10:13:01,767 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.95 vs. limit=15.0
2024-09-23 10:13:03,921 INFO [train.py:1198] (1/4) Epoch 13, batch 3750, loss[loss=0.2326, ctc_loss=0.1555, cr_loss=0.3853, over 17026.00 frames. ], tot_loss[loss=0.237, ctc_loss=0.1618, cr_loss=0.3756, over 3362514.63 frames. ], batch size: 44, lr: 9.13e-03, grad_scale: 32.0
2024-09-23 10:13:12,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=235676.0, ans=0.0
2024-09-23 10:13:18,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=235722.66666666666, ans=0.0
2024-09-23 10:14:02,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.31 vs. limit=10.0
2024-09-23 10:14:06,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=235862.66666666666, ans=0.04949747468305833
2024-09-23 10:14:14,235 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.350e+02 1.440e+02 1.635e+02 2.891e+02, threshold=2.880e+02, percent-clipped=1.0
2024-09-23 10:14:22,032 INFO [train.py:1198] (1/4) Epoch 13, batch 3800, loss[loss=0.2569, ctc_loss=0.1796, cr_loss=0.3869, over 16736.00 frames. ], tot_loss[loss=0.2387, ctc_loss=0.1633, cr_loss=0.3768, over 3336737.65 frames. ], batch size: 61, lr: 9.12e-03, grad_scale: 32.0
2024-09-23 10:14:32,007 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.39 vs. limit=15.0
2024-09-23 10:14:48,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=235956.0, ans=0.2
2024-09-23 10:14:51,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=236002.66666666666, ans=0.125
2024-09-23 10:14:51,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=236002.66666666666, ans=0.125
2024-09-23 10:15:05,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=236002.66666666666, ans=0.0
2024-09-23 10:15:10,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=236049.33333333334, ans=0.0
2024-09-23 10:15:29,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=236096.0, ans=0.0
2024-09-23 10:15:29,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=236096.0, ans=0.1
2024-09-23 10:15:30,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=236096.0, ans=0.0
2024-09-23 10:15:40,299 INFO [train.py:1198] (1/4) Epoch 13, batch 3850, loss[loss=0.2244, ctc_loss=0.1564, cr_loss=0.34, over 17334.00 frames. ], tot_loss[loss=0.241, ctc_loss=0.1654, cr_loss=0.3784, over 3287161.72 frames. ], batch size: 51, lr: 9.12e-03, grad_scale: 16.0
2024-09-23 10:15:51,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=236142.66666666666, ans=0.0
2024-09-23 10:16:06,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=236189.33333333334, ans=0.0
2024-09-23 10:16:32,895 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.40 vs. limit=22.5
2024-09-23 10:16:36,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=236282.66666666666, ans=0.125
2024-09-23 10:16:40,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=236329.33333333334, ans=0.125
2024-09-23 10:17:39,608 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.439e+02 1.592e+02 1.784e+02 2.502e+02, threshold=3.185e+02, percent-clipped=0.0
2024-09-23 10:17:39,632 INFO [train.py:1198] (1/4) Epoch 14, batch 0, loss[loss=0.2495, ctc_loss=0.1724, cr_loss=0.3857, over 17035.00 frames. ], tot_loss[loss=0.2495, ctc_loss=0.1724, cr_loss=0.3857, over 17035.00 frames. ], batch size: 52, lr: 8.78e-03, grad_scale: 32.0
2024-09-23 10:17:39,632 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 10:17:55,048 INFO [train.py:1230] (1/4) Epoch 14, validation: loss=0.04435, ctc_loss=0.04435, cr_loss=7.317e-15, over 944034.00 frames. 
2024-09-23 10:17:55,049 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 10:18:09,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=236357.33333333334, ans=0.125
2024-09-23 10:18:17,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=236404.0, ans=0.2
2024-09-23 10:18:19,605 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.26 vs. limit=22.5
2024-09-23 10:18:46,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.00 vs. limit=12.0
2024-09-23 10:19:02,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=236544.0, ans=0.125
2024-09-23 10:19:22,450 INFO [train.py:1198] (1/4) Epoch 14, batch 50, loss[loss=0.21, ctc_loss=0.1392, cr_loss=0.3538, over 16821.00 frames. ], tot_loss[loss=0.2291, ctc_loss=0.156, cr_loss=0.3654, over 764727.92 frames. ], batch size: 37, lr: 8.78e-03, grad_scale: 32.0
2024-09-23 10:19:32,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=236590.66666666666, ans=0.0
2024-09-23 10:19:37,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=236637.33333333334, ans=0.125
2024-09-23 10:19:44,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=236637.33333333334, ans=0.125
2024-09-23 10:20:05,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=236684.0, ans=0.1
2024-09-23 10:20:20,802 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.08 vs. limit=22.5
2024-09-23 10:20:30,541 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.07 vs. limit=15.0
2024-09-23 10:20:42,382 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.241e+02 1.435e+02 1.718e+02 2.310e+02, threshold=2.871e+02, percent-clipped=0.0
2024-09-23 10:20:42,407 INFO [train.py:1198] (1/4) Epoch 14, batch 100, loss[loss=0.2738, ctc_loss=0.1911, cr_loss=0.4135, over 16522.00 frames. ], tot_loss[loss=0.2364, ctc_loss=0.1614, cr_loss=0.3751, over 1337296.61 frames. ], batch size: 66, lr: 8.77e-03, grad_scale: 32.0
2024-09-23 10:20:52,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=236824.0, ans=0.0
2024-09-23 10:21:13,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=236917.33333333334, ans=0.0
2024-09-23 10:21:19,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=236917.33333333334, ans=0.125
2024-09-23 10:21:23,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=236917.33333333334, ans=0.2
2024-09-23 10:21:24,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=236917.33333333334, ans=0.125
2024-09-23 10:21:54,528 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.09 vs. limit=10.0
2024-09-23 10:22:03,253 INFO [train.py:1198] (1/4) Epoch 14, batch 150, loss[loss=0.2282, ctc_loss=0.1527, cr_loss=0.3774, over 17156.00 frames. ], tot_loss[loss=0.2346, ctc_loss=0.16, cr_loss=0.3731, over 1768590.85 frames. ], batch size: 48, lr: 8.77e-03, grad_scale: 32.0
2024-09-23 10:22:03,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=237057.33333333334, ans=0.125
2024-09-23 10:22:54,337 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.93 vs. limit=15.0
2024-09-23 10:23:28,629 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.307e+02 1.446e+02 1.689e+02 2.559e+02, threshold=2.892e+02, percent-clipped=0.0
2024-09-23 10:23:28,654 INFO [train.py:1198] (1/4) Epoch 14, batch 200, loss[loss=0.1831, ctc_loss=0.1212, cr_loss=0.3096, over 17199.00 frames. ], tot_loss[loss=0.2357, ctc_loss=0.1609, cr_loss=0.3738, over 2108515.74 frames. ], batch size: 41, lr: 8.76e-03, grad_scale: 32.0
2024-09-23 10:23:43,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=237290.66666666666, ans=0.0
2024-09-23 10:24:40,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=237477.33333333334, ans=0.0
2024-09-23 10:24:53,917 INFO [train.py:1198] (1/4) Epoch 14, batch 250, loss[loss=0.2753, ctc_loss=0.1905, cr_loss=0.4241, over 17033.00 frames. ], tot_loss[loss=0.237, ctc_loss=0.1618, cr_loss=0.3758, over 2384447.91 frames. ], batch size: 52, lr: 8.76e-03, grad_scale: 32.0
2024-09-23 10:24:54,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=237524.0, ans=0.09899494936611666
2024-09-23 10:25:21,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=237570.66666666666, ans=0.0
2024-09-23 10:25:26,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=237617.33333333334, ans=0.0
2024-09-23 10:25:59,665 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.07 vs. limit=15.0
2024-09-23 10:26:08,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=237710.66666666666, ans=0.125
2024-09-23 10:26:13,348 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.272e+02 1.438e+02 1.640e+02 2.630e+02, threshold=2.876e+02, percent-clipped=0.0
2024-09-23 10:26:13,372 INFO [train.py:1198] (1/4) Epoch 14, batch 300, loss[loss=0.272, ctc_loss=0.1929, cr_loss=0.3953, over 12073.00 frames. ], tot_loss[loss=0.2356, ctc_loss=0.1608, cr_loss=0.374, over 2593508.61 frames. ], batch size: 124, lr: 8.76e-03, grad_scale: 32.0
2024-09-23 10:26:38,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=237804.0, ans=0.0
2024-09-23 10:26:38,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=237804.0, ans=0.1
2024-09-23 10:26:48,978 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.47 vs. limit=10.0
2024-09-23 10:26:51,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=237850.66666666666, ans=0.0
2024-09-23 10:27:13,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=237897.33333333334, ans=0.125
2024-09-23 10:27:26,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=237944.0, ans=0.125
2024-09-23 10:27:32,701 INFO [train.py:1198] (1/4) Epoch 14, batch 350, loss[loss=0.2562, ctc_loss=0.1763, cr_loss=0.3995, over 16917.00 frames. ], tot_loss[loss=0.2366, ctc_loss=0.1614, cr_loss=0.376, over 2767481.84 frames. ], batch size: 58, lr: 8.75e-03, grad_scale: 32.0
2024-09-23 10:28:21,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer_ff3.min_abs, batch_count=238084.0, ans=0.2
2024-09-23 10:28:41,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=238130.66666666666, ans=0.125
2024-09-23 10:29:02,515 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.337e+02 1.482e+02 1.714e+02 2.531e+02, threshold=2.964e+02, percent-clipped=0.0
2024-09-23 10:29:02,539 INFO [train.py:1198] (1/4) Epoch 14, batch 400, loss[loss=0.2418, ctc_loss=0.1674, cr_loss=0.3722, over 17148.00 frames. ], tot_loss[loss=0.2366, ctc_loss=0.1614, cr_loss=0.3758, over 2894586.85 frames. ], batch size: 48, lr: 8.75e-03, grad_scale: 32.0
2024-09-23 10:30:21,630 INFO [train.py:1198] (1/4) Epoch 14, batch 450, loss[loss=0.28, ctc_loss=0.1934, cr_loss=0.4332, over 15099.00 frames. ], tot_loss[loss=0.2363, ctc_loss=0.1611, cr_loss=0.376, over 3004159.27 frames. ], batch size: 89, lr: 8.74e-03, grad_scale: 32.0
2024-09-23 10:30:39,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=238504.0, ans=0.125
2024-09-23 10:30:55,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=238550.66666666666, ans=0.0
2024-09-23 10:30:55,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=238550.66666666666, ans=0.125
2024-09-23 10:31:25,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=238644.0, ans=0.1
2024-09-23 10:31:34,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=238644.0, ans=0.5
2024-09-23 10:31:35,148 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.85 vs. limit=15.0
2024-09-23 10:31:41,085 INFO [train.py:1198] (1/4) Epoch 14, batch 500, loss[loss=0.2222, ctc_loss=0.1486, cr_loss=0.3682, over 17299.00 frames. ], tot_loss[loss=0.2365, ctc_loss=0.1613, cr_loss=0.3758, over 3079242.43 frames. ], batch size: 46, lr: 8.74e-03, grad_scale: 16.0
2024-09-23 10:31:42,719 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.208e+02 1.324e+02 1.495e+02 2.086e+02, threshold=2.649e+02, percent-clipped=0.0
2024-09-23 10:32:05,298 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 10:32:06,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=238737.33333333334, ans=0.5
2024-09-23 10:32:16,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=238784.0, ans=0.125
2024-09-23 10:32:24,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=238784.0, ans=0.125
2024-09-23 10:33:06,648 INFO [train.py:1198] (1/4) Epoch 14, batch 550, loss[loss=0.2535, ctc_loss=0.1702, cr_loss=0.4163, over 17312.00 frames. ], tot_loss[loss=0.2348, ctc_loss=0.1599, cr_loss=0.3741, over 3143263.53 frames. ], batch size: 49, lr: 8.74e-03, grad_scale: 16.0
2024-09-23 10:33:32,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=238970.66666666666, ans=0.125
2024-09-23 10:33:52,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=239017.33333333334, ans=10.0
2024-09-23 10:34:02,319 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.42 vs. limit=6.0
2024-09-23 10:34:03,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=239064.0, ans=0.125
2024-09-23 10:34:03,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=239064.0, ans=0.0
2024-09-23 10:34:08,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=239064.0, ans=0.1
2024-09-23 10:34:31,893 INFO [train.py:1198] (1/4) Epoch 14, batch 600, loss[loss=0.2382, ctc_loss=0.1624, cr_loss=0.379, over 17353.00 frames. ], tot_loss[loss=0.2344, ctc_loss=0.1596, cr_loss=0.3742, over 3197271.84 frames. ], batch size: 48, lr: 8.73e-03, grad_scale: 16.0
2024-09-23 10:34:32,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=239157.33333333334, ans=0.125
2024-09-23 10:34:33,454 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.266e+02 1.344e+02 1.475e+02 2.652e+02, threshold=2.689e+02, percent-clipped=1.0
2024-09-23 10:35:02,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=239250.66666666666, ans=0.125
2024-09-23 10:35:16,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=239250.66666666666, ans=0.0
2024-09-23 10:35:20,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.32 vs. limit=22.5
2024-09-23 10:35:29,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=239297.33333333334, ans=0.125
2024-09-23 10:35:35,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=239344.0, ans=0.1
2024-09-23 10:35:38,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=239344.0, ans=0.125
2024-09-23 10:35:51,348 INFO [train.py:1198] (1/4) Epoch 14, batch 650, loss[loss=0.2242, ctc_loss=0.1528, cr_loss=0.3569, over 17271.00 frames. ], tot_loss[loss=0.2348, ctc_loss=0.16, cr_loss=0.3739, over 3224127.69 frames. ], batch size: 44, lr: 8.73e-03, grad_scale: 16.0
2024-09-23 10:36:16,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=239437.33333333334, ans=0.125
2024-09-23 10:37:11,334 INFO [train.py:1198] (1/4) Epoch 14, batch 700, loss[loss=0.2548, ctc_loss=0.1719, cr_loss=0.4143, over 17035.00 frames. ], tot_loss[loss=0.2342, ctc_loss=0.1595, cr_loss=0.3732, over 3255933.54 frames. ], batch size: 52, lr: 8.72e-03, grad_scale: 16.0
2024-09-23 10:37:13,000 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.260e+02 1.373e+02 1.552e+02 2.322e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-23 10:38:24,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=239810.66666666666, ans=0.0
2024-09-23 10:38:39,754 INFO [train.py:1198] (1/4) Epoch 14, batch 750, loss[loss=0.2297, ctc_loss=0.1524, cr_loss=0.3868, over 17161.00 frames. ], tot_loss[loss=0.2328, ctc_loss=0.1584, cr_loss=0.3716, over 3278669.92 frames. ], batch size: 45, lr: 8.72e-03, grad_scale: 16.0
2024-09-23 10:38:52,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=239857.33333333334, ans=0.1
2024-09-23 10:39:08,755 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.93 vs. limit=6.0
2024-09-23 10:39:52,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=240044.0, ans=0.1
2024-09-23 10:39:56,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=240044.0, ans=0.0
2024-09-23 10:39:57,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=240044.0, ans=0.0
2024-09-23 10:40:02,417 INFO [train.py:1198] (1/4) Epoch 14, batch 800, loss[loss=0.2476, ctc_loss=0.17, cr_loss=0.388, over 16517.00 frames. ], tot_loss[loss=0.2326, ctc_loss=0.1583, cr_loss=0.3716, over 3305359.19 frames. ], batch size: 66, lr: 8.71e-03, grad_scale: 32.0
2024-09-23 10:40:03,951 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.283e+02 1.393e+02 1.518e+02 3.186e+02, threshold=2.786e+02, percent-clipped=2.0
2024-09-23 10:40:04,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=240090.66666666666, ans=0.0
2024-09-23 10:40:12,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=240090.66666666666, ans=0.0
2024-09-23 10:40:34,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.max_abs, batch_count=240184.0, ans=10.0
2024-09-23 10:40:34,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=240184.0, ans=0.125
2024-09-23 10:40:47,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=240184.0, ans=0.125
2024-09-23 10:41:11,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=240277.33333333334, ans=0.125
2024-09-23 10:41:22,389 INFO [train.py:1198] (1/4) Epoch 14, batch 850, loss[loss=0.2737, ctc_loss=0.1923, cr_loss=0.4071, over 16539.00 frames. ], tot_loss[loss=0.2329, ctc_loss=0.1586, cr_loss=0.3713, over 3300423.53 frames. ], batch size: 66, lr: 8.71e-03, grad_scale: 32.0
2024-09-23 10:41:22,666 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 10:41:37,078 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.86 vs. limit=15.0
2024-09-23 10:41:49,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=240370.66666666666, ans=0.125
2024-09-23 10:41:53,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=240417.33333333334, ans=0.125
2024-09-23 10:41:58,210 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.20 vs. limit=15.0
2024-09-23 10:42:06,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=240417.33333333334, ans=0.125
2024-09-23 10:42:44,012 INFO [train.py:1198] (1/4) Epoch 14, batch 900, loss[loss=0.2406, ctc_loss=0.1636, cr_loss=0.3848, over 17164.00 frames. ], tot_loss[loss=0.2315, ctc_loss=0.1575, cr_loss=0.3702, over 3322440.18 frames. ], batch size: 45, lr: 8.71e-03, grad_scale: 32.0
2024-09-23 10:42:48,314 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.265e+02 1.359e+02 1.500e+02 2.203e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-23 10:43:15,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=240604.0, ans=0.125
2024-09-23 10:43:20,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=240650.66666666666, ans=0.1
2024-09-23 10:43:20,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=240650.66666666666, ans=0.025
2024-09-23 10:43:33,069 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.64 vs. limit=15.0
2024-09-23 10:44:11,568 INFO [train.py:1198] (1/4) Epoch 14, batch 950, loss[loss=0.215, ctc_loss=0.1422, cr_loss=0.3642, over 17108.00 frames. ], tot_loss[loss=0.2315, ctc_loss=0.1576, cr_loss=0.3698, over 3323641.23 frames. ], batch size: 40, lr: 8.70e-03, grad_scale: 32.0
2024-09-23 10:44:27,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=240837.33333333334, ans=0.0
2024-09-23 10:44:49,100 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.38 vs. limit=15.0
2024-09-23 10:44:56,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=240884.0, ans=0.125
2024-09-23 10:45:01,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=240930.66666666666, ans=0.0
2024-09-23 10:45:06,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=240930.66666666666, ans=0.125
2024-09-23 10:45:17,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=240977.33333333334, ans=0.125
2024-09-23 10:45:25,188 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.52 vs. limit=22.5
2024-09-23 10:45:25,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=240977.33333333334, ans=0.0
2024-09-23 10:45:31,842 INFO [train.py:1198] (1/4) Epoch 14, batch 1000, loss[loss=0.2369, ctc_loss=0.1586, cr_loss=0.3917, over 17148.00 frames. ], tot_loss[loss=0.2327, ctc_loss=0.1584, cr_loss=0.3717, over 3333328.12 frames. ], batch size: 48, lr: 8.70e-03, grad_scale: 32.0
2024-09-23 10:45:33,325 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.317e+02 1.495e+02 1.766e+02 2.386e+02, threshold=2.990e+02, percent-clipped=0.0
2024-09-23 10:46:19,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=241164.0, ans=0.2
2024-09-23 10:46:31,943 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.14 vs. limit=10.0
2024-09-23 10:46:32,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=241164.0, ans=0.1
2024-09-23 10:46:42,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=241210.66666666666, ans=0.125
2024-09-23 10:46:44,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=241210.66666666666, ans=0.2
2024-09-23 10:46:44,364 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=10.83 vs. limit=15.0
2024-09-23 10:46:52,003 INFO [train.py:1198] (1/4) Epoch 14, batch 1050, loss[loss=0.2797, ctc_loss=0.196, cr_loss=0.4185, over 14584.00 frames. ], tot_loss[loss=0.2316, ctc_loss=0.1575, cr_loss=0.3705, over 3341989.34 frames. ], batch size: 88, lr: 8.69e-03, grad_scale: 32.0
2024-09-23 10:47:06,251 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.41 vs. limit=6.0
2024-09-23 10:47:07,665 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.93 vs. limit=12.0
2024-09-23 10:47:21,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=241304.0, ans=0.125
2024-09-23 10:48:08,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.86 vs. limit=15.0
2024-09-23 10:48:10,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=241444.0, ans=0.125
2024-09-23 10:48:15,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=241490.66666666666, ans=0.2
2024-09-23 10:48:16,987 INFO [train.py:1198] (1/4) Epoch 14, batch 1100, loss[loss=0.1945, ctc_loss=0.1317, cr_loss=0.3141, over 17206.00 frames. ], tot_loss[loss=0.232, ctc_loss=0.158, cr_loss=0.3703, over 3340606.00 frames. ], batch size: 41, lr: 8.69e-03, grad_scale: 32.0
2024-09-23 10:48:18,620 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.315e+02 1.444e+02 1.614e+02 2.728e+02, threshold=2.888e+02, percent-clipped=0.0
2024-09-23 10:48:29,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=241490.66666666666, ans=0.125
2024-09-23 10:48:35,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=241537.33333333334, ans=0.1
2024-09-23 10:48:42,316 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=13.51 vs. limit=22.5
2024-09-23 10:48:42,355 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=10.82 vs. limit=15.0
2024-09-23 10:48:54,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=241584.0, ans=0.1
2024-09-23 10:49:05,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=241584.0, ans=0.2
2024-09-23 10:49:14,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=241630.66666666666, ans=0.125
2024-09-23 10:49:21,520 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.47 vs. limit=12.0
2024-09-23 10:49:35,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=241677.33333333334, ans=0.2
2024-09-23 10:49:35,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=241677.33333333334, ans=0.0
2024-09-23 10:49:41,261 INFO [train.py:1198] (1/4) Epoch 14, batch 1150, loss[loss=0.2332, ctc_loss=0.1575, cr_loss=0.3787, over 17354.00 frames. ], tot_loss[loss=0.2326, ctc_loss=0.1586, cr_loss=0.3703, over 3345645.68 frames. ], batch size: 48, lr: 8.69e-03, grad_scale: 32.0
2024-09-23 10:49:45,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=241724.0, ans=0.125
2024-09-23 10:49:48,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer_ff2.min_abs, batch_count=241724.0, ans=0.1
2024-09-23 10:49:57,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=241770.66666666666, ans=0.125
2024-09-23 10:50:24,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=241817.33333333334, ans=0.125
2024-09-23 10:50:25,399 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=18.51 vs. limit=22.5
2024-09-23 10:50:35,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=241864.0, ans=0.0
2024-09-23 10:50:47,347 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.21 vs. limit=15.0
2024-09-23 10:50:56,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=241910.66666666666, ans=0.125
2024-09-23 10:51:01,211 INFO [train.py:1198] (1/4) Epoch 14, batch 1200, loss[loss=0.2473, ctc_loss=0.1667, cr_loss=0.4033, over 17019.00 frames. ], tot_loss[loss=0.2332, ctc_loss=0.1589, cr_loss=0.3716, over 3347558.20 frames. ], batch size: 56, lr: 8.68e-03, grad_scale: 32.0
2024-09-23 10:51:01,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=241957.33333333334, ans=0.0
2024-09-23 10:51:02,790 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.307e+02 1.418e+02 1.626e+02 2.907e+02, threshold=2.837e+02, percent-clipped=1.0
2024-09-23 10:51:33,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=242050.66666666666, ans=0.125
2024-09-23 10:51:34,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=242050.66666666666, ans=0.2
2024-09-23 10:51:51,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=242097.33333333334, ans=0.125
2024-09-23 10:52:13,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=242144.0, ans=0.125
2024-09-23 10:52:20,958 INFO [train.py:1198] (1/4) Epoch 14, batch 1250, loss[loss=0.2556, ctc_loss=0.1756, cr_loss=0.3999, over 15058.00 frames. ], tot_loss[loss=0.2326, ctc_loss=0.1584, cr_loss=0.3714, over 3350252.61 frames. ], batch size: 89, lr: 8.68e-03, grad_scale: 32.0
2024-09-23 10:52:40,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=242237.33333333334, ans=0.125
2024-09-23 10:53:10,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=242284.0, ans=0.125
2024-09-23 10:53:31,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=242377.33333333334, ans=0.125
2024-09-23 10:53:38,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=242377.33333333334, ans=0.0
2024-09-23 10:53:49,818 INFO [train.py:1198] (1/4) Epoch 14, batch 1300, loss[loss=0.2585, ctc_loss=0.1781, cr_loss=0.402, over 17040.00 frames. ], tot_loss[loss=0.2312, ctc_loss=0.1572, cr_loss=0.3704, over 3361611.75 frames. ], batch size: 52, lr: 8.67e-03, grad_scale: 32.0
2024-09-23 10:53:51,343 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.264e+02 1.376e+02 1.514e+02 2.274e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-23 10:53:56,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=242424.0, ans=0.125
2024-09-23 10:53:59,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=242424.0, ans=0.0
2024-09-23 10:54:18,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=242470.66666666666, ans=0.125
2024-09-23 10:54:46,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=242564.0, ans=0.125
2024-09-23 10:54:56,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=242610.66666666666, ans=0.025
2024-09-23 10:55:07,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=242610.66666666666, ans=0.125
2024-09-23 10:55:07,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.75 vs. limit=22.5
2024-09-23 10:55:10,058 INFO [train.py:1198] (1/4) Epoch 14, batch 1350, loss[loss=0.2809, ctc_loss=0.1904, cr_loss=0.4521, over 17042.00 frames. ], tot_loss[loss=0.2321, ctc_loss=0.1578, cr_loss=0.3714, over 3363459.62 frames. ], batch size: 52, lr: 8.67e-03, grad_scale: 32.0
2024-09-23 10:55:25,761 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.79 vs. limit=22.5
2024-09-23 10:55:30,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.97 vs. limit=15.0
2024-09-23 10:55:42,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=242750.66666666666, ans=0.2
2024-09-23 10:55:47,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=242750.66666666666, ans=0.04949747468305833
2024-09-23 10:56:24,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=242844.0, ans=0.0
2024-09-23 10:56:32,046 INFO [train.py:1198] (1/4) Epoch 14, batch 1400, loss[loss=0.2726, ctc_loss=0.1961, cr_loss=0.3822, over 12171.00 frames. ], tot_loss[loss=0.2321, ctc_loss=0.1579, cr_loss=0.3707, over 3366266.58 frames. ], batch size: 123, lr: 8.67e-03, grad_scale: 32.0
2024-09-23 10:56:33,632 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.305e+02 1.425e+02 1.607e+02 2.757e+02, threshold=2.850e+02, percent-clipped=1.0
2024-09-23 10:57:08,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=242984.0, ans=0.035
2024-09-23 10:57:56,917 INFO [train.py:1198] (1/4) Epoch 14, batch 1450, loss[loss=0.2702, ctc_loss=0.1892, cr_loss=0.4049, over 16980.00 frames. ], tot_loss[loss=0.2311, ctc_loss=0.1574, cr_loss=0.3687, over 3357617.57 frames. ], batch size: 56, lr: 8.66e-03, grad_scale: 16.0
2024-09-23 10:57:58,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=243124.0, ans=0.1
2024-09-23 10:58:03,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=243124.0, ans=0.0
2024-09-23 10:58:14,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=243170.66666666666, ans=0.07
2024-09-23 10:58:18,391 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=7.28 vs. limit=15.0
2024-09-23 10:58:22,442 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.50 vs. limit=15.0
2024-09-23 10:58:23,914 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.45 vs. limit=15.0
2024-09-23 10:58:29,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=243217.33333333334, ans=0.2
2024-09-23 10:58:45,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=243217.33333333334, ans=0.125
2024-09-23 10:58:48,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=243264.0, ans=0.125
2024-09-23 10:58:50,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=243264.0, ans=0.125
2024-09-23 10:59:15,984 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.30 vs. limit=12.0
2024-09-23 10:59:17,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.02 vs. limit=15.0
2024-09-23 10:59:18,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=243310.66666666666, ans=0.1
2024-09-23 10:59:21,663 INFO [train.py:1198] (1/4) Epoch 14, batch 1500, loss[loss=0.1985, ctc_loss=0.1302, cr_loss=0.3412, over 17114.00 frames. ], tot_loss[loss=0.233, ctc_loss=0.1587, cr_loss=0.3711, over 3348034.49 frames. ], batch size: 40, lr: 8.66e-03, grad_scale: 16.0
2024-09-23 10:59:24,853 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.258e+02 1.373e+02 1.539e+02 2.095e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-23 10:59:40,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=243404.0, ans=0.125
2024-09-23 10:59:47,238 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 10:59:48,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=243404.0, ans=0.2
2024-09-23 11:00:33,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=243544.0, ans=0.07
2024-09-23 11:00:41,713 INFO [train.py:1198] (1/4) Epoch 14, batch 1550, loss[loss=0.2715, ctc_loss=0.1855, cr_loss=0.4298, over 16555.00 frames. ], tot_loss[loss=0.2326, ctc_loss=0.1584, cr_loss=0.3709, over 3342797.20 frames. ], batch size: 66, lr: 8.65e-03, grad_scale: 16.0
2024-09-23 11:01:01,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=243637.33333333334, ans=0.2
2024-09-23 11:01:12,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=243684.0, ans=0.1
2024-09-23 11:01:16,460 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.44 vs. limit=22.5
2024-09-23 11:01:27,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=243684.0, ans=0.125
2024-09-23 11:01:32,559 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=14.07 vs. limit=15.0
2024-09-23 11:01:54,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=243777.33333333334, ans=0.0
2024-09-23 11:01:55,909 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.81 vs. limit=15.0
2024-09-23 11:02:01,665 INFO [train.py:1198] (1/4) Epoch 14, batch 1600, loss[loss=0.2473, ctc_loss=0.1711, cr_loss=0.381, over 17044.00 frames. ], tot_loss[loss=0.2324, ctc_loss=0.1583, cr_loss=0.3706, over 3344704.88 frames. ], batch size: 56, lr: 8.65e-03, grad_scale: 32.0
2024-09-23 11:02:04,729 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.281e+02 1.419e+02 1.549e+02 2.365e+02, threshold=2.838e+02, percent-clipped=0.0
2024-09-23 11:03:08,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=243964.0, ans=0.1
2024-09-23 11:03:17,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=244010.66666666666, ans=0.125
2024-09-23 11:03:30,841 INFO [train.py:1198] (1/4) Epoch 14, batch 1650, loss[loss=0.2534, ctc_loss=0.1714, cr_loss=0.4099, over 17034.00 frames. ], tot_loss[loss=0.234, ctc_loss=0.1595, cr_loss=0.3721, over 3341445.67 frames. ], batch size: 44, lr: 8.64e-03, grad_scale: 32.0
2024-09-23 11:03:31,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=244057.33333333334, ans=0.0
2024-09-23 11:03:42,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=244057.33333333334, ans=0.125
2024-09-23 11:03:43,704 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 11:03:59,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=244104.0, ans=0.0
2024-09-23 11:04:01,812 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.23 vs. limit=15.0
2024-09-23 11:04:20,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=244197.33333333334, ans=0.125
2024-09-23 11:04:34,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=244244.0, ans=0.04949747468305833
2024-09-23 11:04:50,834 INFO [train.py:1198] (1/4) Epoch 14, batch 1700, loss[loss=0.238, ctc_loss=0.1627, cr_loss=0.3764, over 17048.00 frames. ], tot_loss[loss=0.2335, ctc_loss=0.159, cr_loss=0.3722, over 3351136.75 frames. ], batch size: 52, lr: 8.64e-03, grad_scale: 32.0
2024-09-23 11:04:54,005 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.254e+02 1.382e+02 1.612e+02 3.536e+02, threshold=2.764e+02, percent-clipped=2.0
2024-09-23 11:04:55,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=244290.66666666666, ans=0.0
2024-09-23 11:05:05,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=244337.33333333334, ans=0.0
2024-09-23 11:05:24,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=244384.0, ans=0.05
2024-09-23 11:05:37,975 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.90 vs. limit=15.0
2024-09-23 11:05:44,547 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.98 vs. limit=6.0
2024-09-23 11:05:44,631 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.94 vs. limit=15.0
2024-09-23 11:06:09,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=244524.0, ans=0.035
2024-09-23 11:06:09,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=244524.0, ans=0.5
2024-09-23 11:06:10,497 INFO [train.py:1198] (1/4) Epoch 14, batch 1750, loss[loss=0.1884, ctc_loss=0.1245, cr_loss=0.3196, over 17087.00 frames. ], tot_loss[loss=0.2324, ctc_loss=0.1581, cr_loss=0.3712, over 3362237.53 frames. ], batch size: 43, lr: 8.64e-03, grad_scale: 32.0
2024-09-23 11:06:26,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=244570.66666666666, ans=0.125
2024-09-23 11:06:27,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=244570.66666666666, ans=0.0
2024-09-23 11:07:22,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=244710.66666666666, ans=0.025
2024-09-23 11:07:36,213 INFO [train.py:1198] (1/4) Epoch 14, batch 1800, loss[loss=0.206, ctc_loss=0.1379, cr_loss=0.3407, over 17256.00 frames. ], tot_loss[loss=0.2329, ctc_loss=0.1585, cr_loss=0.3717, over 3368813.80 frames. ], batch size: 42, lr: 8.63e-03, grad_scale: 32.0
2024-09-23 11:07:39,492 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.282e+02 1.372e+02 1.529e+02 2.252e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-23 11:08:00,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=244804.0, ans=0.1
2024-09-23 11:08:44,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=244944.0, ans=0.0
2024-09-23 11:08:48,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=244944.0, ans=0.125
2024-09-23 11:08:57,708 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.63 vs. limit=6.0
2024-09-23 11:09:01,689 INFO [train.py:1198] (1/4) Epoch 14, batch 1850, loss[loss=0.2338, ctc_loss=0.156, cr_loss=0.3889, over 17018.00 frames. ], tot_loss[loss=0.2333, ctc_loss=0.1589, cr_loss=0.3722, over 3368174.53 frames. ], batch size: 44, lr: 8.63e-03, grad_scale: 32.0
2024-09-23 11:09:08,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=244990.66666666666, ans=0.025
2024-09-23 11:09:16,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=245037.33333333334, ans=0.125
2024-09-23 11:09:33,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=245084.0, ans=0.125
2024-09-23 11:09:48,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=245130.66666666666, ans=0.0
2024-09-23 11:10:01,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=245130.66666666666, ans=0.125
2024-09-23 11:10:12,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=245177.33333333334, ans=0.125
2024-09-23 11:10:14,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=245177.33333333334, ans=0.0
2024-09-23 11:10:16,199 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.39 vs. limit=22.5
2024-09-23 11:10:21,848 INFO [train.py:1198] (1/4) Epoch 14, batch 1900, loss[loss=0.2744, ctc_loss=0.1872, cr_loss=0.4357, over 16750.00 frames. ], tot_loss[loss=0.2319, ctc_loss=0.1578, cr_loss=0.3706, over 3369533.51 frames. ], batch size: 61, lr: 8.62e-03, grad_scale: 32.0
2024-09-23 11:10:25,081 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.260e+02 1.374e+02 1.529e+02 3.130e+02, threshold=2.747e+02, percent-clipped=1.0
2024-09-23 11:10:51,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=245270.66666666666, ans=0.1
2024-09-23 11:11:27,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=245410.66666666666, ans=0.125
2024-09-23 11:11:38,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=245410.66666666666, ans=0.0
2024-09-23 11:11:41,462 INFO [train.py:1198] (1/4) Epoch 14, batch 1950, loss[loss=0.2298, ctc_loss=0.1565, cr_loss=0.3667, over 17216.00 frames. ], tot_loss[loss=0.2332, ctc_loss=0.1587, cr_loss=0.3724, over 3357945.07 frames. ], batch size: 50, lr: 8.62e-03, grad_scale: 32.0
2024-09-23 11:11:41,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=245457.33333333334, ans=0.1
2024-09-23 11:12:01,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=245504.0, ans=0.1
2024-09-23 11:12:04,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=245504.0, ans=0.125
2024-09-23 11:12:23,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=245550.66666666666, ans=0.025
2024-09-23 11:12:40,103 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=8.47 vs. limit=12.0
2024-09-23 11:12:49,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=245644.0, ans=0.05
2024-09-23 11:12:58,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=245644.0, ans=0.125
2024-09-23 11:13:09,060 INFO [train.py:1198] (1/4) Epoch 14, batch 2000, loss[loss=0.2439, ctc_loss=0.1671, cr_loss=0.3843, over 16779.00 frames. ], tot_loss[loss=0.2345, ctc_loss=0.1597, cr_loss=0.3739, over 3352183.79 frames. ], batch size: 61, lr: 8.62e-03, grad_scale: 32.0
2024-09-23 11:13:14,678 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.297e+02 1.399e+02 1.635e+02 2.518e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-23 11:13:16,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=245690.66666666666, ans=10.0
2024-09-23 11:14:19,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=245877.33333333334, ans=0.125
2024-09-23 11:14:28,279 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.03 vs. limit=22.5
2024-09-23 11:14:31,852 INFO [train.py:1198] (1/4) Epoch 14, batch 2050, loss[loss=0.259, ctc_loss=0.1791, cr_loss=0.3995, over 17304.00 frames. ], tot_loss[loss=0.2331, ctc_loss=0.1586, cr_loss=0.3721, over 3358736.59 frames. ], batch size: 51, lr: 8.61e-03, grad_scale: 32.0
2024-09-23 11:14:33,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=245924.0, ans=0.025
2024-09-23 11:15:52,134 INFO [train.py:1198] (1/4) Epoch 14, batch 2100, loss[loss=0.2365, ctc_loss=0.1608, cr_loss=0.3784, over 17233.00 frames. ], tot_loss[loss=0.2313, ctc_loss=0.1573, cr_loss=0.3703, over 3367268.05 frames. ], batch size: 50, lr: 8.61e-03, grad_scale: 32.0
2024-09-23 11:15:55,402 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.226e+02 1.315e+02 1.406e+02 3.033e+02, threshold=2.629e+02, percent-clipped=1.0
2024-09-23 11:16:13,831 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.06 vs. limit=15.0
2024-09-23 11:17:15,228 INFO [train.py:1198] (1/4) Epoch 14, batch 2150, loss[loss=0.2136, ctc_loss=0.1431, cr_loss=0.3524, over 17023.00 frames. ], tot_loss[loss=0.2314, ctc_loss=0.1574, cr_loss=0.3702, over 3357637.70 frames. ], batch size: 44, lr: 8.60e-03, grad_scale: 32.0
2024-09-23 11:17:23,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=246390.66666666666, ans=0.0
2024-09-23 11:17:45,474 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.78 vs. limit=22.5
2024-09-23 11:17:47,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=246437.33333333334, ans=0.0
2024-09-23 11:17:55,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=246484.0, ans=0.125
2024-09-23 11:18:06,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=246484.0, ans=0.0
2024-09-23 11:18:06,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=246484.0, ans=0.1
2024-09-23 11:18:07,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=246530.66666666666, ans=0.125
2024-09-23 11:18:31,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=246577.33333333334, ans=0.125
2024-09-23 11:18:43,399 INFO [train.py:1198] (1/4) Epoch 14, batch 2200, loss[loss=0.2129, ctc_loss=0.1455, cr_loss=0.3368, over 17174.00 frames. ], tot_loss[loss=0.2322, ctc_loss=0.1582, cr_loss=0.3701, over 3358198.54 frames. ], batch size: 41, lr: 8.60e-03, grad_scale: 32.0
2024-09-23 11:18:46,517 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.334e+02 1.425e+02 1.540e+02 2.133e+02, threshold=2.850e+02, percent-clipped=0.0
2024-09-23 11:18:54,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=246624.0, ans=0.125
2024-09-23 11:18:56,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=246624.0, ans=0.125
2024-09-23 11:19:09,680 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.66 vs. limit=15.0
2024-09-23 11:19:18,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=246717.33333333334, ans=0.04949747468305833
2024-09-23 11:19:24,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2.whitening_limit, batch_count=246717.33333333334, ans=15.0
2024-09-23 11:19:42,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=246764.0, ans=0.125
2024-09-23 11:19:44,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=246764.0, ans=0.1
2024-09-23 11:20:02,897 INFO [train.py:1198] (1/4) Epoch 14, batch 2250, loss[loss=0.2518, ctc_loss=0.1729, cr_loss=0.3946, over 17041.00 frames. ], tot_loss[loss=0.2335, ctc_loss=0.1592, cr_loss=0.3715, over 3343711.66 frames. ], batch size: 56, lr: 8.60e-03, grad_scale: 16.0
2024-09-23 11:20:08,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=246857.33333333334, ans=0.0
2024-09-23 11:20:28,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=246904.0, ans=0.125
2024-09-23 11:20:30,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=246904.0, ans=0.125
2024-09-23 11:20:39,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=246950.66666666666, ans=0.125
2024-09-23 11:21:00,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=246997.33333333334, ans=0.035
2024-09-23 11:21:01,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=246997.33333333334, ans=0.07
2024-09-23 11:21:22,504 INFO [train.py:1198] (1/4) Epoch 14, batch 2300, loss[loss=0.2288, ctc_loss=0.1566, cr_loss=0.3609, over 17112.00 frames. ], tot_loss[loss=0.2318, ctc_loss=0.1579, cr_loss=0.3699, over 3352856.84 frames. ], batch size: 49, lr: 8.59e-03, grad_scale: 16.0
2024-09-23 11:21:27,261 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.285e+02 1.398e+02 1.588e+02 2.479e+02, threshold=2.795e+02, percent-clipped=0.0
2024-09-23 11:21:48,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=247137.33333333334, ans=0.2
2024-09-23 11:22:07,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=247184.0, ans=0.0
2024-09-23 11:22:50,511 INFO [train.py:1198] (1/4) Epoch 14, batch 2350, loss[loss=0.1957, ctc_loss=0.1281, cr_loss=0.3382, over 17117.00 frames. ], tot_loss[loss=0.2312, ctc_loss=0.1574, cr_loss=0.3691, over 3356934.91 frames. ], batch size: 40, lr: 8.59e-03, grad_scale: 16.0
2024-09-23 11:22:54,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=247324.0, ans=0.125
2024-09-23 11:23:03,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=247324.0, ans=0.05
2024-09-23 11:23:37,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=247417.33333333334, ans=0.125
2024-09-23 11:23:41,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=247464.0, ans=0.0
2024-09-23 11:24:12,731 INFO [train.py:1198] (1/4) Epoch 14, batch 2400, loss[loss=0.21, ctc_loss=0.1386, cr_loss=0.3568, over 17024.00 frames. ], tot_loss[loss=0.2293, ctc_loss=0.1558, cr_loss=0.3676, over 3367769.03 frames. ], batch size: 44, lr: 8.58e-03, grad_scale: 32.0
2024-09-23 11:24:16,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=247557.33333333334, ans=0.05
2024-09-23 11:24:17,509 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.242e+02 1.312e+02 1.459e+02 2.054e+02, threshold=2.624e+02, percent-clipped=0.0
2024-09-23 11:24:17,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=247557.33333333334, ans=0.0
2024-09-23 11:24:19,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=247557.33333333334, ans=0.125
2024-09-23 11:24:52,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=247650.66666666666, ans=0.0
2024-09-23 11:24:59,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=247697.33333333334, ans=0.05
2024-09-23 11:25:05,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=247697.33333333334, ans=0.95
2024-09-23 11:25:32,183 INFO [train.py:1198] (1/4) Epoch 14, batch 2450, loss[loss=0.2252, ctc_loss=0.1531, cr_loss=0.3603, over 17174.00 frames. ], tot_loss[loss=0.23, ctc_loss=0.1563, cr_loss=0.3686, over 3372759.06 frames. ], batch size: 45, lr: 8.58e-03, grad_scale: 32.0
2024-09-23 11:25:35,004 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.80 vs. limit=6.0
2024-09-23 11:25:43,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=247790.66666666666, ans=0.0
2024-09-23 11:25:59,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=247837.33333333334, ans=0.0
2024-09-23 11:26:06,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=247884.0, ans=0.125
2024-09-23 11:26:54,748 INFO [train.py:1198] (1/4) Epoch 14, batch 2500, loss[loss=0.2285, ctc_loss=0.1564, cr_loss=0.3604, over 15987.00 frames. ], tot_loss[loss=0.2309, ctc_loss=0.157, cr_loss=0.3697, over 3368707.74 frames. ], batch size: 74, lr: 8.58e-03, grad_scale: 32.0
2024-09-23 11:26:59,531 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.311e+02 1.464e+02 1.674e+02 2.701e+02, threshold=2.928e+02, percent-clipped=1.0
2024-09-23 11:27:06,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=248024.0, ans=0.125
2024-09-23 11:27:16,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.min_positive, batch_count=248070.66666666666, ans=0.025
2024-09-23 11:28:21,882 INFO [train.py:1198] (1/4) Epoch 14, batch 2550, loss[loss=0.2472, ctc_loss=0.1675, cr_loss=0.3982, over 17010.00 frames. ], tot_loss[loss=0.2317, ctc_loss=0.1576, cr_loss=0.3704, over 3362938.88 frames. ], batch size: 53, lr: 8.57e-03, grad_scale: 32.0
2024-09-23 11:28:27,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=248257.33333333334, ans=0.0
2024-09-23 11:28:59,487 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.26 vs. limit=22.5
2024-09-23 11:29:10,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=248397.33333333334, ans=0.09899494936611666
2024-09-23 11:29:34,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=248444.0, ans=0.0
2024-09-23 11:29:42,049 INFO [train.py:1198] (1/4) Epoch 14, batch 2600, loss[loss=0.2404, ctc_loss=0.1646, cr_loss=0.3792, over 17220.00 frames. ], tot_loss[loss=0.2333, ctc_loss=0.1588, cr_loss=0.3726, over 3359719.52 frames. ], batch size: 50, lr: 8.57e-03, grad_scale: 32.0
2024-09-23 11:29:46,741 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.321e+02 1.442e+02 1.644e+02 2.368e+02, threshold=2.883e+02, percent-clipped=0.0
2024-09-23 11:29:48,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=248490.66666666666, ans=0.125
2024-09-23 11:30:23,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=248584.0, ans=0.2
2024-09-23 11:30:49,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=248677.33333333334, ans=0.125
2024-09-23 11:30:56,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=248677.33333333334, ans=0.125
2024-09-23 11:31:02,245 INFO [train.py:1198] (1/4) Epoch 14, batch 2650, loss[loss=0.2329, ctc_loss=0.1591, cr_loss=0.3693, over 17368.00 frames. ], tot_loss[loss=0.2329, ctc_loss=0.1587, cr_loss=0.3713, over 3351419.03 frames. ], batch size: 48, lr: 8.56e-03, grad_scale: 32.0
2024-09-23 11:31:04,329 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.19 vs. limit=15.0
2024-09-23 11:32:22,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=248910.66666666666, ans=0.1
2024-09-23 11:32:26,707 INFO [train.py:1198] (1/4) Epoch 14, batch 2700, loss[loss=0.2606, ctc_loss=0.1848, cr_loss=0.3788, over 16719.00 frames. ], tot_loss[loss=0.2333, ctc_loss=0.1589, cr_loss=0.3717, over 3341901.94 frames. ], batch size: 61, lr: 8.56e-03, grad_scale: 32.0
2024-09-23 11:32:31,436 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.327e+02 1.447e+02 1.619e+02 2.182e+02, threshold=2.895e+02, percent-clipped=0.0
2024-09-23 11:32:31,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=248957.33333333334, ans=0.1
2024-09-23 11:32:40,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=248957.33333333334, ans=0.125
2024-09-23 11:32:42,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=248957.33333333334, ans=0.025
2024-09-23 11:33:00,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=249004.0, ans=0.125
2024-09-23 11:33:03,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer_ff3.min_abs, batch_count=249050.66666666666, ans=0.2
2024-09-23 11:33:21,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=249097.33333333334, ans=0.125
2024-09-23 11:33:32,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=249097.33333333334, ans=0.0
2024-09-23 11:33:50,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=249190.66666666666, ans=0.125
2024-09-23 11:33:51,509 INFO [train.py:1198] (1/4) Epoch 14, batch 2750, loss[loss=0.2142, ctc_loss=0.1416, cr_loss=0.3631, over 17158.00 frames. ], tot_loss[loss=0.2338, ctc_loss=0.1592, cr_loss=0.373, over 3348381.74 frames. ], batch size: 45, lr: 8.56e-03, grad_scale: 32.0
2024-09-23 11:33:51,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=249190.66666666666, ans=0.125
2024-09-23 11:34:30,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=249284.0, ans=0.2
2024-09-23 11:34:33,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=249284.0, ans=0.2
2024-09-23 11:35:10,882 INFO [train.py:1198] (1/4) Epoch 14, batch 2800, loss[loss=0.2013, ctc_loss=0.1354, cr_loss=0.3294, over 17096.00 frames. ], tot_loss[loss=0.2324, ctc_loss=0.1582, cr_loss=0.371, over 3354225.74 frames. ], batch size: 43, lr: 8.55e-03, grad_scale: 32.0
2024-09-23 11:35:15,651 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.308e+02 1.382e+02 1.526e+02 2.267e+02, threshold=2.765e+02, percent-clipped=0.0
2024-09-23 11:35:20,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=249424.0, ans=0.0
2024-09-23 11:35:22,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=249424.0, ans=0.0
2024-09-23 11:35:41,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=249517.33333333334, ans=0.1
2024-09-23 11:35:41,840 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.25 vs. limit=15.0
2024-09-23 11:35:42,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=249517.33333333334, ans=0.2
2024-09-23 11:35:49,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=249517.33333333334, ans=0.125
2024-09-23 11:35:49,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=249517.33333333334, ans=0.125
2024-09-23 11:36:06,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=249564.0, ans=0.0
2024-09-23 11:36:26,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=249610.66666666666, ans=0.125
2024-09-23 11:36:31,110 INFO [train.py:1198] (1/4) Epoch 14, batch 2850, loss[loss=0.2029, ctc_loss=0.1337, cr_loss=0.346, over 16354.00 frames. ], tot_loss[loss=0.2319, ctc_loss=0.1577, cr_loss=0.3708, over 3360089.76 frames. ], batch size: 36, lr: 8.55e-03, grad_scale: 16.0
2024-09-23 11:36:38,340 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.78 vs. limit=12.0
2024-09-23 11:36:44,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=249657.33333333334, ans=0.0
2024-09-23 11:36:49,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_positive, batch_count=249704.0, ans=0.05
2024-09-23 11:37:42,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=249844.0, ans=0.035
2024-09-23 11:37:55,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=249844.0, ans=0.0
2024-09-23 11:37:56,796 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.37 vs. limit=5.0
2024-09-23 11:38:01,839 INFO [train.py:1198] (1/4) Epoch 14, batch 2900, loss[loss=0.2169, ctc_loss=0.1425, cr_loss=0.3722, over 17024.00 frames. ], tot_loss[loss=0.2313, ctc_loss=0.1572, cr_loss=0.3703, over 3359191.82 frames. ], batch size: 44, lr: 8.55e-03, grad_scale: 16.0
2024-09-23 11:38:08,318 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.270e+02 1.418e+02 1.645e+02 2.792e+02, threshold=2.835e+02, percent-clipped=1.0
2024-09-23 11:38:45,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=249984.0, ans=0.125
2024-09-23 11:38:52,172 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.84 vs. limit=22.5
2024-09-23 11:39:04,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=250077.33333333334, ans=0.125
2024-09-23 11:39:20,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=250124.0, ans=0.125
2024-09-23 11:39:21,419 INFO [train.py:1198] (1/4) Epoch 14, batch 2950, loss[loss=0.21, ctc_loss=0.1398, cr_loss=0.3511, over 17222.00 frames. ], tot_loss[loss=0.2325, ctc_loss=0.1582, cr_loss=0.3715, over 3352681.39 frames. ], batch size: 50, lr: 8.54e-03, grad_scale: 16.0
2024-09-23 11:39:21,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=250124.0, ans=0.1
2024-09-23 11:39:56,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=250217.33333333334, ans=0.0
2024-09-23 11:40:01,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=250217.33333333334, ans=0.1
2024-09-23 11:40:26,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=250310.66666666666, ans=0.1
2024-09-23 11:40:26,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=250310.66666666666, ans=0.09899494936611666
2024-09-23 11:40:32,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=250310.66666666666, ans=0.2
2024-09-23 11:40:35,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=250310.66666666666, ans=0.125
2024-09-23 11:40:40,235 INFO [train.py:1198] (1/4) Epoch 14, batch 3000, loss[loss=0.2264, ctc_loss=0.1538, cr_loss=0.3628, over 16896.00 frames. ], tot_loss[loss=0.2319, ctc_loss=0.1577, cr_loss=0.371, over 3354848.74 frames. ], batch size: 58, lr: 8.54e-03, grad_scale: 16.0
2024-09-23 11:40:40,236 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 11:40:55,478 INFO [train.py:1230] (1/4) Epoch 14, validation: loss=0.04331, ctc_loss=0.04331, cr_loss=7.532e-15, over 944034.00 frames. 
2024-09-23 11:40:55,478 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 11:40:55,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=250357.33333333334, ans=0.125
2024-09-23 11:40:58,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=250357.33333333334, ans=0.2
2024-09-23 11:41:01,568 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.302e+02 1.389e+02 1.457e+02 1.974e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-23 11:41:08,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=250357.33333333334, ans=0.125
2024-09-23 11:41:58,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=250544.0, ans=0.05
2024-09-23 11:42:12,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=250590.66666666666, ans=0.0
2024-09-23 11:42:13,881 INFO [train.py:1198] (1/4) Epoch 14, batch 3050, loss[loss=0.2241, ctc_loss=0.1479, cr_loss=0.3808, over 17021.00 frames. ], tot_loss[loss=0.2312, ctc_loss=0.1571, cr_loss=0.3705, over 3363954.95 frames. ], batch size: 44, lr: 8.53e-03, grad_scale: 16.0
2024-09-23 11:42:53,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=250684.0, ans=0.0
2024-09-23 11:42:58,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=250684.0, ans=0.2
2024-09-23 11:43:02,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=250730.66666666666, ans=0.125
2024-09-23 11:43:05,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=250730.66666666666, ans=0.125
2024-09-23 11:43:34,614 INFO [train.py:1198] (1/4) Epoch 14, batch 3100, loss[loss=0.2454, ctc_loss=0.1692, cr_loss=0.3809, over 15930.00 frames. ], tot_loss[loss=0.2298, ctc_loss=0.156, cr_loss=0.3689, over 3367457.24 frames. ], batch size: 74, lr: 8.53e-03, grad_scale: 16.0
2024-09-23 11:43:36,585 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 11:43:40,946 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.263e+02 1.328e+02 1.443e+02 2.080e+02, threshold=2.656e+02, percent-clipped=0.0
2024-09-23 11:43:42,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=250824.0, ans=0.125
2024-09-23 11:44:55,828 INFO [train.py:1198] (1/4) Epoch 14, batch 3150, loss[loss=0.2511, ctc_loss=0.1742, cr_loss=0.3844, over 17231.00 frames. ], tot_loss[loss=0.2294, ctc_loss=0.1557, cr_loss=0.3685, over 3369431.99 frames. ], batch size: 55, lr: 8.53e-03, grad_scale: 16.0
2024-09-23 11:45:44,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=251150.66666666666, ans=0.1
2024-09-23 11:46:03,861 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.71 vs. limit=15.0
2024-09-23 11:46:11,449 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.82 vs. limit=10.0
2024-09-23 11:46:18,649 INFO [train.py:1198] (1/4) Epoch 14, batch 3200, loss[loss=0.2704, ctc_loss=0.1939, cr_loss=0.3824, over 15098.00 frames. ], tot_loss[loss=0.2312, ctc_loss=0.157, cr_loss=0.3708, over 3362773.35 frames. ], batch size: 89, lr: 8.52e-03, grad_scale: 32.0
2024-09-23 11:46:24,742 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.267e+02 1.361e+02 1.514e+02 1.918e+02, threshold=2.723e+02, percent-clipped=0.0
2024-09-23 11:47:07,226 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.92 vs. limit=12.0
2024-09-23 11:47:19,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=251477.33333333334, ans=0.2
2024-09-23 11:47:36,178 INFO [train.py:1198] (1/4) Epoch 14, batch 3250, loss[loss=0.215, ctc_loss=0.1466, cr_loss=0.342, over 17027.00 frames. ], tot_loss[loss=0.2304, ctc_loss=0.1564, cr_loss=0.3698, over 3369220.84 frames. ], batch size: 44, lr: 8.52e-03, grad_scale: 32.0
2024-09-23 11:47:46,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=251524.0, ans=0.5
2024-09-23 11:47:46,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=8.28 vs. limit=12.0
2024-09-23 11:47:49,209 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.52 vs. limit=22.5
2024-09-23 11:47:52,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=251570.66666666666, ans=0.0
2024-09-23 11:47:54,039 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.31 vs. limit=15.0
2024-09-23 11:47:54,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.72 vs. limit=15.0
2024-09-23 11:47:59,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=251570.66666666666, ans=0.125
2024-09-23 11:48:09,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=251617.33333333334, ans=0.125
2024-09-23 11:48:24,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=251664.0, ans=0.09899494936611666
2024-09-23 11:48:26,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=251664.0, ans=0.2
2024-09-23 11:48:35,123 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.37 vs. limit=22.5
2024-09-23 11:48:45,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=251710.66666666666, ans=0.025
2024-09-23 11:48:51,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=251710.66666666666, ans=0.0
2024-09-23 11:48:54,096 INFO [train.py:1198] (1/4) Epoch 14, batch 3300, loss[loss=0.2165, ctc_loss=0.1466, cr_loss=0.3497, over 17025.00 frames. ], tot_loss[loss=0.2309, ctc_loss=0.1568, cr_loss=0.3702, over 3370146.23 frames. ], batch size: 44, lr: 8.51e-03, grad_scale: 32.0
2024-09-23 11:49:00,432 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.303e+02 1.410e+02 1.606e+02 3.318e+02, threshold=2.819e+02, percent-clipped=1.0
2024-09-23 11:49:05,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=251757.33333333334, ans=0.0
2024-09-23 11:49:18,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=251804.0, ans=0.1
2024-09-23 11:49:43,529 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.02 vs. limit=10.0
2024-09-23 11:50:09,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=251944.0, ans=0.125
2024-09-23 11:50:12,288 INFO [train.py:1198] (1/4) Epoch 14, batch 3350, loss[loss=0.2342, ctc_loss=0.1573, cr_loss=0.3843, over 16999.00 frames. ], tot_loss[loss=0.231, ctc_loss=0.157, cr_loss=0.3698, over 3362634.39 frames. ], batch size: 44, lr: 8.51e-03, grad_scale: 16.0
2024-09-23 11:50:51,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=252084.0, ans=0.125
2024-09-23 11:50:52,584 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.60 vs. limit=15.0
2024-09-23 11:50:58,242 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 11:51:16,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=252177.33333333334, ans=0.1
2024-09-23 11:51:30,492 INFO [train.py:1198] (1/4) Epoch 14, batch 3400, loss[loss=0.2596, ctc_loss=0.1743, cr_loss=0.4269, over 16903.00 frames. ], tot_loss[loss=0.2303, ctc_loss=0.1565, cr_loss=0.369, over 3371346.85 frames. ], batch size: 58, lr: 8.51e-03, grad_scale: 16.0
2024-09-23 11:51:35,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=252224.0, ans=0.0
2024-09-23 11:51:38,162 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.275e+02 1.402e+02 1.543e+02 4.509e+02, threshold=2.804e+02, percent-clipped=1.0
2024-09-23 11:51:43,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=252224.0, ans=0.2
2024-09-23 11:51:55,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=252270.66666666666, ans=0.0
2024-09-23 11:51:57,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=252270.66666666666, ans=0.1
2024-09-23 11:52:01,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.min_positive, batch_count=252317.33333333334, ans=0.05
2024-09-23 11:52:08,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=252317.33333333334, ans=0.2
2024-09-23 11:52:17,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.57 vs. limit=22.5
2024-09-23 11:52:38,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=252410.66666666666, ans=0.1
2024-09-23 11:52:48,079 INFO [train.py:1198] (1/4) Epoch 14, batch 3450, loss[loss=0.2387, ctc_loss=0.1614, cr_loss=0.3862, over 17291.00 frames. ], tot_loss[loss=0.2292, ctc_loss=0.1557, cr_loss=0.3674, over 3373439.38 frames. ], batch size: 49, lr: 8.50e-03, grad_scale: 16.0
2024-09-23 11:52:48,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=252457.33333333334, ans=0.125
2024-09-23 11:53:10,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=252504.0, ans=0.0
2024-09-23 11:53:13,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=252504.0, ans=0.05
2024-09-23 11:53:19,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=252550.66666666666, ans=0.2
2024-09-23 11:53:32,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=252550.66666666666, ans=0.0
2024-09-23 11:54:08,250 INFO [train.py:1198] (1/4) Epoch 14, batch 3500, loss[loss=0.2458, ctc_loss=0.1697, cr_loss=0.3807, over 16997.00 frames. ], tot_loss[loss=0.229, ctc_loss=0.1555, cr_loss=0.3673, over 3369095.79 frames. ], batch size: 53, lr: 8.50e-03, grad_scale: 16.0
2024-09-23 11:54:13,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=252690.66666666666, ans=0.125
2024-09-23 11:54:16,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=252690.66666666666, ans=0.0
2024-09-23 11:54:18,110 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.252e+02 1.352e+02 1.458e+02 2.935e+02, threshold=2.705e+02, percent-clipped=1.0
2024-09-23 11:54:33,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=252737.33333333334, ans=0.125
2024-09-23 11:54:36,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=252737.33333333334, ans=0.125
2024-09-23 11:54:54,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=252784.0, ans=0.125
2024-09-23 11:54:55,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=252830.66666666666, ans=0.0
2024-09-23 11:55:07,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=252830.66666666666, ans=15.0
2024-09-23 11:55:32,279 INFO [train.py:1198] (1/4) Epoch 14, batch 3550, loss[loss=0.2093, ctc_loss=0.1431, cr_loss=0.331, over 17081.00 frames. ], tot_loss[loss=0.2292, ctc_loss=0.1556, cr_loss=0.3678, over 3368161.38 frames. ], batch size: 43, lr: 8.49e-03, grad_scale: 16.0
2024-09-23 11:55:46,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=252970.66666666666, ans=0.2
2024-09-23 11:56:09,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=253017.33333333334, ans=0.125
2024-09-23 11:56:33,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=253110.66666666666, ans=0.0
2024-09-23 11:56:51,144 INFO [train.py:1198] (1/4) Epoch 14, batch 3600, loss[loss=0.2267, ctc_loss=0.1501, cr_loss=0.3829, over 17246.00 frames. ], tot_loss[loss=0.2291, ctc_loss=0.1556, cr_loss=0.3675, over 3364905.11 frames. ], batch size: 50, lr: 8.49e-03, grad_scale: 32.0
2024-09-23 11:56:58,882 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.239e+02 1.349e+02 1.491e+02 2.999e+02, threshold=2.699e+02, percent-clipped=1.0
2024-09-23 11:57:10,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=253204.0, ans=0.2
2024-09-23 11:57:13,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=253204.0, ans=0.04949747468305833
2024-09-23 11:57:22,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=253250.66666666666, ans=0.125
2024-09-23 11:57:23,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=253250.66666666666, ans=0.0
2024-09-23 11:57:31,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=253250.66666666666, ans=0.1
2024-09-23 11:57:53,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=253344.0, ans=0.0
2024-09-23 11:57:59,088 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=7.76 vs. limit=15.0
2024-09-23 11:58:08,967 INFO [train.py:1198] (1/4) Epoch 14, batch 3650, loss[loss=0.2019, ctc_loss=0.1378, cr_loss=0.3208, over 17289.00 frames. ], tot_loss[loss=0.2291, ctc_loss=0.1556, cr_loss=0.3675, over 3369676.69 frames. ], batch size: 42, lr: 8.49e-03, grad_scale: 32.0
2024-09-23 11:58:13,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=253390.66666666666, ans=0.125
2024-09-23 11:58:36,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=253437.33333333334, ans=0.125
2024-09-23 11:59:28,444 INFO [train.py:1198] (1/4) Epoch 14, batch 3700, loss[loss=0.2177, ctc_loss=0.1441, cr_loss=0.3681, over 17307.00 frames. ], tot_loss[loss=0.2294, ctc_loss=0.1557, cr_loss=0.3683, over 3370047.22 frames. ], batch size: 46, lr: 8.48e-03, grad_scale: 32.0
2024-09-23 11:59:31,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=253624.0, ans=0.2
2024-09-23 11:59:36,288 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.294e+02 1.387e+02 1.607e+02 1.987e+02, threshold=2.774e+02, percent-clipped=0.0
2024-09-23 11:59:50,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=253670.66666666666, ans=0.1
2024-09-23 12:00:04,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=253717.33333333334, ans=0.125
2024-09-23 12:00:06,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.57 vs. limit=22.5
2024-09-23 12:00:11,509 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.43 vs. limit=22.5
2024-09-23 12:00:46,835 INFO [train.py:1198] (1/4) Epoch 14, batch 3750, loss[loss=0.2538, ctc_loss=0.171, cr_loss=0.4138, over 17015.00 frames. ], tot_loss[loss=0.2286, ctc_loss=0.1551, cr_loss=0.3677, over 3375618.31 frames. ], batch size: 53, lr: 8.48e-03, grad_scale: 32.0
2024-09-23 12:01:09,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=253904.0, ans=0.125
2024-09-23 12:01:10,586 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 12:01:13,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=253904.0, ans=0.125
2024-09-23 12:01:21,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=253950.66666666666, ans=0.0
2024-09-23 12:01:21,720 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.22 vs. limit=6.0
2024-09-23 12:01:23,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=20.32 vs. limit=22.5
2024-09-23 12:01:55,185 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=9.87 vs. limit=15.0
2024-09-23 12:02:05,049 INFO [train.py:1198] (1/4) Epoch 14, batch 3800, loss[loss=0.2058, ctc_loss=0.1395, cr_loss=0.3316, over 16960.00 frames. ], tot_loss[loss=0.2299, ctc_loss=0.156, cr_loss=0.3694, over 3368660.41 frames. ], batch size: 42, lr: 8.48e-03, grad_scale: 32.0
2024-09-23 12:02:13,078 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.380e+02 1.551e+02 1.777e+02 3.575e+02, threshold=3.102e+02, percent-clipped=2.0
2024-09-23 12:02:22,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=254137.33333333334, ans=0.125
2024-09-23 12:02:25,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=254137.33333333334, ans=0.025
2024-09-23 12:02:49,529 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.42 vs. limit=10.0
2024-09-23 12:03:24,331 INFO [train.py:1198] (1/4) Epoch 14, batch 3850, loss[loss=0.2314, ctc_loss=0.1574, cr_loss=0.3703, over 16921.00 frames. ], tot_loss[loss=0.2307, ctc_loss=0.1569, cr_loss=0.3688, over 3329393.68 frames. ], batch size: 58, lr: 8.47e-03, grad_scale: 32.0
2024-09-23 12:03:55,088 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.79 vs. limit=10.0
2024-09-23 12:03:59,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=254417.33333333334, ans=0.125
2024-09-23 12:05:28,602 INFO [train.py:1198] (1/4) Epoch 15, batch 0, loss[loss=0.2321, ctc_loss=0.1563, cr_loss=0.3791, over 17283.00 frames. ], tot_loss[loss=0.2321, ctc_loss=0.1563, cr_loss=0.3791, over 17283.00 frames. ], batch size: 46, lr: 8.18e-03, grad_scale: 32.0
2024-09-23 12:05:28,603 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 12:05:41,534 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.2.self_attn_weights, attn_weights_entropy = tensor([2.0280, 2.2730, 2.3389, 2.4828, 2.1981, 2.1706, 2.4134, 2.3842],
       device='cuda:1')
2024-09-23 12:05:45,280 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([1.9440, 3.4075, 3.2428, 3.3934, 3.4160, 2.8249, 3.2035, 1.9975],
       device='cuda:1')
2024-09-23 12:05:46,330 INFO [train.py:1230] (1/4) Epoch 15, validation: loss=0.0431, ctc_loss=0.0431, cr_loss=7.486e-15, over 944034.00 frames. 
2024-09-23 12:05:46,330 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 12:05:48,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=254538.66666666666, ans=0.0
2024-09-23 12:05:57,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=254538.66666666666, ans=10.0
2024-09-23 12:06:00,796 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.387e+02 1.561e+02 1.706e+02 2.670e+02, threshold=3.121e+02, percent-clipped=0.0
2024-09-23 12:06:20,277 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.02 vs. limit=15.0
2024-09-23 12:06:21,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=254632.0, ans=0.125
2024-09-23 12:06:33,257 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.54 vs. limit=12.0
2024-09-23 12:06:35,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=254678.66666666666, ans=0.1
2024-09-23 12:06:42,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=254678.66666666666, ans=0.0
2024-09-23 12:06:48,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=254725.33333333334, ans=0.1
2024-09-23 12:07:00,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=254725.33333333334, ans=0.125
2024-09-23 12:07:05,543 INFO [train.py:1198] (1/4) Epoch 15, batch 50, loss[loss=0.211, ctc_loss=0.1419, cr_loss=0.3458, over 16742.00 frames. ], tot_loss[loss=0.2339, ctc_loss=0.1593, cr_loss=0.3731, over 763627.45 frames. ], batch size: 37, lr: 8.18e-03, grad_scale: 32.0
2024-09-23 12:07:05,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=254772.0, ans=0.125
2024-09-23 12:08:12,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=254958.66666666666, ans=0.0
2024-09-23 12:08:28,617 INFO [train.py:1198] (1/4) Epoch 15, batch 100, loss[loss=0.2182, ctc_loss=0.143, cr_loss=0.3758, over 17207.00 frames. ], tot_loss[loss=0.2324, ctc_loss=0.1582, cr_loss=0.3714, over 1339353.14 frames. ], batch size: 55, lr: 8.17e-03, grad_scale: 32.0
2024-09-23 12:08:36,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=7.17 vs. limit=15.0
2024-09-23 12:08:42,830 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.954e+01 1.228e+02 1.306e+02 1.476e+02 1.867e+02, threshold=2.613e+02, percent-clipped=0.0
2024-09-23 12:08:43,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=255052.0, ans=0.0
2024-09-23 12:09:11,951 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 12:09:26,442 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.35 vs. limit=6.0
2024-09-23 12:09:33,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=255192.0, ans=0.125
2024-09-23 12:09:47,784 INFO [train.py:1198] (1/4) Epoch 15, batch 150, loss[loss=0.241, ctc_loss=0.1643, cr_loss=0.3837, over 16039.00 frames. ], tot_loss[loss=0.2339, ctc_loss=0.1592, cr_loss=0.3733, over 1789326.34 frames. ], batch size: 74, lr: 8.17e-03, grad_scale: 32.0
2024-09-23 12:09:54,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=255238.66666666666, ans=0.125
2024-09-23 12:10:01,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=255238.66666666666, ans=0.125
2024-09-23 12:10:26,389 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.99 vs. limit=22.5
2024-09-23 12:10:54,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=255378.66666666666, ans=0.125
2024-09-23 12:11:05,795 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.73 vs. limit=15.0
2024-09-23 12:11:14,511 INFO [train.py:1198] (1/4) Epoch 15, batch 200, loss[loss=0.2475, ctc_loss=0.1689, cr_loss=0.3929, over 17310.00 frames. ], tot_loss[loss=0.2333, ctc_loss=0.1587, cr_loss=0.3729, over 2136853.29 frames. ], batch size: 51, lr: 8.16e-03, grad_scale: 32.0
2024-09-23 12:11:28,889 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.243e+02 1.308e+02 1.422e+02 1.839e+02, threshold=2.616e+02, percent-clipped=0.0
2024-09-23 12:12:05,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=255612.0, ans=0.125
2024-09-23 12:12:10,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=255612.0, ans=0.1
2024-09-23 12:12:12,631 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=18.75 vs. limit=22.5
2024-09-23 12:12:33,973 INFO [train.py:1198] (1/4) Epoch 15, batch 250, loss[loss=0.2002, ctc_loss=0.1333, cr_loss=0.3346, over 17092.00 frames. ], tot_loss[loss=0.2319, ctc_loss=0.1579, cr_loss=0.37, over 2398034.08 frames. ], batch size: 40, lr: 8.16e-03, grad_scale: 32.0
2024-09-23 12:12:39,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=255705.33333333334, ans=0.025
2024-09-23 12:12:51,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=255752.0, ans=0.1
2024-09-23 12:12:58,316 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.02 vs. limit=22.5
2024-09-23 12:13:02,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=255752.0, ans=0.125
2024-09-23 12:13:44,802 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.80 vs. limit=15.0
2024-09-23 12:13:50,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=255892.0, ans=0.125
2024-09-23 12:13:56,717 INFO [train.py:1198] (1/4) Epoch 15, batch 300, loss[loss=0.2597, ctc_loss=0.1799, cr_loss=0.3988, over 16918.00 frames. ], tot_loss[loss=0.2311, ctc_loss=0.1572, cr_loss=0.3695, over 2615274.63 frames. ], batch size: 58, lr: 8.16e-03, grad_scale: 32.0
2024-09-23 12:14:01,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=255938.66666666666, ans=0.125
2024-09-23 12:14:10,831 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.341e+02 1.470e+02 1.683e+02 2.993e+02, threshold=2.941e+02, percent-clipped=1.0
2024-09-23 12:14:15,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=255985.33333333334, ans=0.125
2024-09-23 12:14:19,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=9.89 vs. limit=15.0
2024-09-23 12:14:28,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=256032.0, ans=0.0
2024-09-23 12:14:33,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=256032.0, ans=0.125
2024-09-23 12:14:36,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=256032.0, ans=0.1
2024-09-23 12:14:58,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=256078.66666666666, ans=0.2
2024-09-23 12:15:13,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=256125.33333333334, ans=0.0
2024-09-23 12:15:21,580 INFO [train.py:1198] (1/4) Epoch 15, batch 350, loss[loss=0.227, ctc_loss=0.153, cr_loss=0.3699, over 17075.00 frames. ], tot_loss[loss=0.2317, ctc_loss=0.1576, cr_loss=0.3705, over 2781120.47 frames. ], batch size: 46, lr: 8.15e-03, grad_scale: 32.0
2024-09-23 12:15:21,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=256172.0, ans=0.2
2024-09-23 12:15:46,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=256218.66666666666, ans=0.125
2024-09-23 12:15:57,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=256265.33333333334, ans=0.125
2024-09-23 12:16:14,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=256312.0, ans=0.2
2024-09-23 12:16:27,448 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.71 vs. limit=15.0
2024-09-23 12:16:44,049 INFO [train.py:1198] (1/4) Epoch 15, batch 400, loss[loss=0.2364, ctc_loss=0.163, cr_loss=0.3667, over 17319.00 frames. ], tot_loss[loss=0.2311, ctc_loss=0.1571, cr_loss=0.37, over 2908315.60 frames. ], batch size: 51, lr: 8.15e-03, grad_scale: 32.0
2024-09-23 12:16:55,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=256405.33333333334, ans=0.125
2024-09-23 12:16:58,142 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.238e+02 1.377e+02 1.544e+02 2.269e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-23 12:17:13,154 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.67 vs. limit=15.0
2024-09-23 12:17:17,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=256498.66666666666, ans=0.04949747468305833
2024-09-23 12:17:43,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=256545.33333333334, ans=0.1
2024-09-23 12:17:48,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=256545.33333333334, ans=0.125
2024-09-23 12:17:58,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=256592.0, ans=0.125
2024-09-23 12:18:00,859 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.25 vs. limit=15.0
2024-09-23 12:18:06,461 INFO [train.py:1198] (1/4) Epoch 15, batch 450, loss[loss=0.2432, ctc_loss=0.1635, cr_loss=0.3983, over 17357.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.156, cr_loss=0.3681, over 3019775.66 frames. ], batch size: 48, lr: 8.15e-03, grad_scale: 32.0
2024-09-23 12:18:24,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=256685.33333333334, ans=0.1
2024-09-23 12:18:27,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=256685.33333333334, ans=0.07
2024-09-23 12:18:41,105 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.88 vs. limit=15.0
2024-09-23 12:19:20,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=256825.33333333334, ans=0.125
2024-09-23 12:19:25,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=256872.0, ans=0.125
2024-09-23 12:19:27,189 INFO [train.py:1198] (1/4) Epoch 15, batch 500, loss[loss=0.2267, ctc_loss=0.1512, cr_loss=0.3775, over 17137.00 frames. ], tot_loss[loss=0.2286, ctc_loss=0.1551, cr_loss=0.3675, over 3099214.53 frames. ], batch size: 48, lr: 8.14e-03, grad_scale: 32.0
2024-09-23 12:19:41,901 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.305e+02 1.439e+02 1.681e+02 2.242e+02, threshold=2.879e+02, percent-clipped=0.0
2024-09-23 12:19:57,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.max_abs, batch_count=256918.66666666666, ans=10.0
2024-09-23 12:20:13,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=256965.33333333334, ans=0.1
2024-09-23 12:20:24,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=257012.0, ans=0.125
2024-09-23 12:20:55,118 INFO [train.py:1198] (1/4) Epoch 15, batch 550, loss[loss=0.2311, ctc_loss=0.1574, cr_loss=0.3684, over 16920.00 frames. ], tot_loss[loss=0.2292, ctc_loss=0.1555, cr_loss=0.3682, over 3154018.71 frames. ], batch size: 58, lr: 8.14e-03, grad_scale: 32.0
2024-09-23 12:21:05,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=257105.33333333334, ans=0.0
2024-09-23 12:21:12,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=257152.0, ans=0.2
2024-09-23 12:21:12,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=257152.0, ans=0.2
2024-09-23 12:21:17,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=257152.0, ans=0.2
2024-09-23 12:21:22,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=257152.0, ans=0.125
2024-09-23 12:21:27,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=257198.66666666666, ans=0.125
2024-09-23 12:21:33,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=257198.66666666666, ans=0.1
2024-09-23 12:21:41,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=257245.33333333334, ans=0.05
2024-09-23 12:21:43,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=257245.33333333334, ans=0.125
2024-09-23 12:21:57,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=257292.0, ans=0.0
2024-09-23 12:22:12,890 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.65 vs. limit=15.0
2024-09-23 12:22:15,289 INFO [train.py:1198] (1/4) Epoch 15, batch 600, loss[loss=0.2559, ctc_loss=0.1785, cr_loss=0.3871, over 16227.00 frames. ], tot_loss[loss=0.2292, ctc_loss=0.1555, cr_loss=0.3686, over 3197001.38 frames. ], batch size: 74, lr: 8.14e-03, grad_scale: 32.0
2024-09-23 12:22:28,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=257338.66666666666, ans=0.025
2024-09-23 12:22:29,755 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.274e+02 1.386e+02 1.572e+02 2.356e+02, threshold=2.773e+02, percent-clipped=0.0
2024-09-23 12:22:39,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=257385.33333333334, ans=0.0
2024-09-23 12:22:47,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=257385.33333333334, ans=0.0
2024-09-23 12:22:52,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=257432.0, ans=0.125
2024-09-23 12:22:52,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=257432.0, ans=0.0
2024-09-23 12:22:53,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=257432.0, ans=0.2
2024-09-23 12:22:57,589 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.80 vs. limit=15.0
2024-09-23 12:23:06,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=257478.66666666666, ans=0.125
2024-09-23 12:23:24,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=257525.33333333334, ans=0.125
2024-09-23 12:23:38,342 INFO [train.py:1198] (1/4) Epoch 15, batch 650, loss[loss=0.2795, ctc_loss=0.1995, cr_loss=0.4, over 12099.00 frames. ], tot_loss[loss=0.2297, ctc_loss=0.1558, cr_loss=0.3691, over 3230860.30 frames. ], batch size: 123, lr: 8.13e-03, grad_scale: 32.0
2024-09-23 12:23:48,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=257572.0, ans=0.125
2024-09-23 12:24:01,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=257618.66666666666, ans=0.07
2024-09-23 12:24:02,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=257618.66666666666, ans=0.0
2024-09-23 12:24:34,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=257712.0, ans=0.04949747468305833
2024-09-23 12:25:01,760 INFO [train.py:1198] (1/4) Epoch 15, batch 700, loss[loss=0.2582, ctc_loss=0.175, cr_loss=0.4161, over 17039.00 frames. ], tot_loss[loss=0.2299, ctc_loss=0.1561, cr_loss=0.3691, over 3264965.83 frames. ], batch size: 56, lr: 8.13e-03, grad_scale: 32.0
2024-09-23 12:25:18,901 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.271e+02 1.381e+02 1.537e+02 2.206e+02, threshold=2.762e+02, percent-clipped=0.0
2024-09-23 12:25:39,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=257898.66666666666, ans=0.125
2024-09-23 12:25:59,978 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=257945.33333333334, ans=0.0
2024-09-23 12:26:11,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=257992.0, ans=0.0
2024-09-23 12:26:25,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=258038.66666666666, ans=0.125
2024-09-23 12:26:26,614 INFO [train.py:1198] (1/4) Epoch 15, batch 750, loss[loss=0.2412, ctc_loss=0.1636, cr_loss=0.3879, over 17018.00 frames. ], tot_loss[loss=0.2298, ctc_loss=0.156, cr_loss=0.3688, over 3279165.42 frames. ], batch size: 51, lr: 8.12e-03, grad_scale: 32.0
2024-09-23 12:26:27,747 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.95 vs. limit=15.0
2024-09-23 12:26:30,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=258038.66666666666, ans=0.125
2024-09-23 12:26:41,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=258085.33333333334, ans=0.2
2024-09-23 12:27:01,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=258132.0, ans=0.0
2024-09-23 12:27:03,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=258132.0, ans=0.125
2024-09-23 12:27:19,722 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.80 vs. limit=15.0
2024-09-23 12:27:38,205 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 12:27:38,546 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.86 vs. limit=15.0
2024-09-23 12:27:48,955 INFO [train.py:1198] (1/4) Epoch 15, batch 800, loss[loss=0.2012, ctc_loss=0.1327, cr_loss=0.3424, over 17094.00 frames. ], tot_loss[loss=0.2292, ctc_loss=0.1556, cr_loss=0.3684, over 3297909.88 frames. ], batch size: 43, lr: 8.12e-03, grad_scale: 32.0
2024-09-23 12:27:51,430 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.33 vs. limit=22.5
2024-09-23 12:28:03,181 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.255e+02 1.392e+02 1.544e+02 3.619e+02, threshold=2.784e+02, percent-clipped=1.0
2024-09-23 12:28:16,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=258318.66666666666, ans=0.1
2024-09-23 12:28:26,054 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=7.78 vs. limit=12.0
2024-09-23 12:28:44,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=258412.0, ans=0.125
2024-09-23 12:28:49,969 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.89 vs. limit=15.0
2024-09-23 12:29:05,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=258458.66666666666, ans=0.0
2024-09-23 12:29:08,256 INFO [train.py:1198] (1/4) Epoch 15, batch 850, loss[loss=0.2628, ctc_loss=0.1847, cr_loss=0.3909, over 17237.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1558, cr_loss=0.369, over 3315494.61 frames. ], batch size: 55, lr: 8.12e-03, grad_scale: 32.0
2024-09-23 12:29:14,211 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.45 vs. limit=15.0
2024-09-23 12:29:25,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=258552.0, ans=0.05
2024-09-23 12:29:59,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=258645.33333333334, ans=0.125
2024-09-23 12:30:36,041 INFO [train.py:1198] (1/4) Epoch 15, batch 900, loss[loss=0.2077, ctc_loss=0.1393, cr_loss=0.3419, over 17144.00 frames. ], tot_loss[loss=0.229, ctc_loss=0.1552, cr_loss=0.3687, over 3334434.99 frames. ], batch size: 48, lr: 8.11e-03, grad_scale: 32.0
2024-09-23 12:30:50,308 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.250e+02 1.335e+02 1.491e+02 2.252e+02, threshold=2.670e+02, percent-clipped=0.0
2024-09-23 12:31:03,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=258785.33333333334, ans=0.025
2024-09-23 12:31:06,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=258832.0, ans=0.2
2024-09-23 12:31:56,134 INFO [train.py:1198] (1/4) Epoch 15, batch 950, loss[loss=0.2489, ctc_loss=0.1665, cr_loss=0.4122, over 16991.00 frames. ], tot_loss[loss=0.2295, ctc_loss=0.1556, cr_loss=0.3693, over 3340180.38 frames. ], batch size: 53, lr: 8.11e-03, grad_scale: 16.0
2024-09-23 12:32:09,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=258972.0, ans=0.1
2024-09-23 12:32:43,956 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=17.24 vs. limit=22.5
2024-09-23 12:32:49,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=259112.0, ans=0.125
2024-09-23 12:32:51,927 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=9.17 vs. limit=15.0
2024-09-23 12:32:57,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=259112.0, ans=0.2
2024-09-23 12:33:05,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=259158.66666666666, ans=0.025
2024-09-23 12:33:17,945 INFO [train.py:1198] (1/4) Epoch 15, batch 1000, loss[loss=0.224, ctc_loss=0.1522, cr_loss=0.3591, over 17361.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1558, cr_loss=0.3691, over 3347805.14 frames. ], batch size: 48, lr: 8.11e-03, grad_scale: 16.0
2024-09-23 12:33:33,769 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.224e+02 1.330e+02 1.426e+02 2.141e+02, threshold=2.660e+02, percent-clipped=0.0
2024-09-23 12:34:11,728 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=6.36 vs. limit=15.0
2024-09-23 12:34:22,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=259392.0, ans=0.125
2024-09-23 12:34:30,768 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.31 vs. limit=15.0
2024-09-23 12:34:40,676 INFO [train.py:1198] (1/4) Epoch 15, batch 1050, loss[loss=0.3095, ctc_loss=0.2212, cr_loss=0.4415, over 15002.00 frames. ], tot_loss[loss=0.2308, ctc_loss=0.1567, cr_loss=0.3703, over 3342100.46 frames. ], batch size: 89, lr: 8.10e-03, grad_scale: 16.0
2024-09-23 12:34:48,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=259438.66666666666, ans=0.125
2024-09-23 12:34:57,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=259485.33333333334, ans=0.025
2024-09-23 12:35:06,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=259485.33333333334, ans=0.0
2024-09-23 12:35:18,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=259532.0, ans=0.1
2024-09-23 12:35:24,996 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.89 vs. limit=12.0
2024-09-23 12:35:47,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=259578.66666666666, ans=0.0
2024-09-23 12:36:05,844 INFO [train.py:1198] (1/4) Epoch 15, batch 1100, loss[loss=0.2149, ctc_loss=0.1415, cr_loss=0.3666, over 17081.00 frames. ], tot_loss[loss=0.2301, ctc_loss=0.1563, cr_loss=0.3695, over 3343287.08 frames. ], batch size: 49, lr: 8.10e-03, grad_scale: 16.0
2024-09-23 12:36:21,570 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.305e+02 1.420e+02 1.545e+02 2.157e+02, threshold=2.840e+02, percent-clipped=0.0
2024-09-23 12:36:23,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=259718.66666666666, ans=0.0
2024-09-23 12:36:25,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=259718.66666666666, ans=0.025
2024-09-23 12:36:29,020 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.86 vs. limit=22.5
2024-09-23 12:36:48,544 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.57 vs. limit=15.0
2024-09-23 12:37:04,412 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.73 vs. limit=15.0
2024-09-23 12:37:19,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=259858.66666666666, ans=0.125
2024-09-23 12:37:25,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=259858.66666666666, ans=0.2
2024-09-23 12:37:28,267 INFO [train.py:1198] (1/4) Epoch 15, batch 1150, loss[loss=0.2228, ctc_loss=0.1502, cr_loss=0.3627, over 15908.00 frames. ], tot_loss[loss=0.2301, ctc_loss=0.1562, cr_loss=0.3694, over 3343813.44 frames. ], batch size: 74, lr: 8.10e-03, grad_scale: 16.0
2024-09-23 12:37:38,642 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.27 vs. limit=22.5
2024-09-23 12:37:42,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=259952.0, ans=0.0
2024-09-23 12:37:51,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=259952.0, ans=0.125
2024-09-23 12:37:51,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=259952.0, ans=0.025
2024-09-23 12:38:19,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=260045.33333333334, ans=0.1
2024-09-23 12:38:21,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=260045.33333333334, ans=0.5
2024-09-23 12:38:45,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=260092.0, ans=0.1
2024-09-23 12:38:48,574 INFO [train.py:1198] (1/4) Epoch 15, batch 1200, loss[loss=0.2944, ctc_loss=0.2068, cr_loss=0.4382, over 15117.00 frames. ], tot_loss[loss=0.2299, ctc_loss=0.156, cr_loss=0.3694, over 3352651.83 frames. ], batch size: 89, lr: 8.09e-03, grad_scale: 32.0
2024-09-23 12:38:48,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=260138.66666666666, ans=0.125
2024-09-23 12:38:58,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=260138.66666666666, ans=0.125
2024-09-23 12:39:04,648 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.045e+02 1.247e+02 1.362e+02 1.504e+02 2.311e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-23 12:39:08,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.88 vs. limit=15.0
2024-09-23 12:39:22,758 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 12:39:33,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=260232.0, ans=0.0
2024-09-23 12:39:36,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=260278.66666666666, ans=0.125
2024-09-23 12:39:49,425 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.14 vs. limit=10.0
2024-09-23 12:40:13,698 INFO [train.py:1198] (1/4) Epoch 15, batch 1250, loss[loss=0.2064, ctc_loss=0.1369, cr_loss=0.3477, over 17113.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1557, cr_loss=0.3696, over 3361217.22 frames. ], batch size: 40, lr: 8.09e-03, grad_scale: 32.0
2024-09-23 12:40:18,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=260372.0, ans=0.0
2024-09-23 12:40:29,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.45 vs. limit=12.0
2024-09-23 12:40:31,280 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.18 vs. limit=22.5
2024-09-23 12:41:04,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=260512.0, ans=0.125
2024-09-23 12:41:15,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=260512.0, ans=0.0
2024-09-23 12:41:20,599 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 12:41:35,958 INFO [train.py:1198] (1/4) Epoch 15, batch 1300, loss[loss=0.2575, ctc_loss=0.1787, cr_loss=0.394, over 16002.00 frames. ], tot_loss[loss=0.2284, ctc_loss=0.1549, cr_loss=0.3677, over 3355402.26 frames. ], batch size: 74, lr: 8.09e-03, grad_scale: 16.0
2024-09-23 12:41:36,520 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=21.66 vs. limit=22.5
2024-09-23 12:41:39,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=260605.33333333334, ans=0.125
2024-09-23 12:41:49,357 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.56 vs. limit=6.0
2024-09-23 12:41:52,162 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 12:41:53,387 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.270e+02 1.373e+02 1.516e+02 2.157e+02, threshold=2.746e+02, percent-clipped=0.0
2024-09-23 12:42:28,806 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.74 vs. limit=15.0
2024-09-23 12:42:58,448 INFO [train.py:1198] (1/4) Epoch 15, batch 1350, loss[loss=0.2455, ctc_loss=0.1712, cr_loss=0.3715, over 15933.00 frames. ], tot_loss[loss=0.2299, ctc_loss=0.156, cr_loss=0.369, over 3349014.21 frames. ], batch size: 74, lr: 8.08e-03, grad_scale: 16.0
2024-09-23 12:43:06,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=260838.66666666666, ans=10.0
2024-09-23 12:43:08,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=260838.66666666666, ans=0.0
2024-09-23 12:43:08,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=260838.66666666666, ans=0.0
2024-09-23 12:43:09,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=260838.66666666666, ans=0.025
2024-09-23 12:44:11,438 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.93 vs. limit=10.0
2024-09-23 12:44:18,620 INFO [train.py:1198] (1/4) Epoch 15, batch 1400, loss[loss=0.2273, ctc_loss=0.1515, cr_loss=0.3794, over 17145.00 frames. ], tot_loss[loss=0.2304, ctc_loss=0.1564, cr_loss=0.3703, over 3353931.15 frames. ], batch size: 48, lr: 8.08e-03, grad_scale: 16.0
2024-09-23 12:44:22,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.32 vs. limit=10.0
2024-09-23 12:44:35,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=261118.66666666666, ans=0.0
2024-09-23 12:44:36,421 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.225e+02 1.306e+02 1.392e+02 2.119e+02, threshold=2.612e+02, percent-clipped=0.0
2024-09-23 12:45:16,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=261212.0, ans=0.0
2024-09-23 12:45:46,010 INFO [train.py:1198] (1/4) Epoch 15, batch 1450, loss[loss=0.1829, ctc_loss=0.1227, cr_loss=0.3013, over 17009.00 frames. ], tot_loss[loss=0.2303, ctc_loss=0.1563, cr_loss=0.3698, over 3346698.73 frames. ], batch size: 44, lr: 8.07e-03, grad_scale: 16.0
2024-09-23 12:46:00,125 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.51 vs. limit=22.5
2024-09-23 12:46:11,333 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.35 vs. limit=6.0
2024-09-23 12:46:16,276 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=15.23 vs. limit=15.0
2024-09-23 12:46:31,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=261398.66666666666, ans=0.125
2024-09-23 12:46:42,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=261445.33333333334, ans=0.0
2024-09-23 12:46:46,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=261445.33333333334, ans=0.1
2024-09-23 12:46:49,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=261445.33333333334, ans=0.015
2024-09-23 12:47:00,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=261492.0, ans=0.09899494936611666
2024-09-23 12:47:08,081 INFO [train.py:1198] (1/4) Epoch 15, batch 1500, loss[loss=0.2256, ctc_loss=0.1531, cr_loss=0.363, over 17024.00 frames. ], tot_loss[loss=0.2295, ctc_loss=0.1557, cr_loss=0.3688, over 3342108.87 frames. ], batch size: 44, lr: 8.07e-03, grad_scale: 16.0
2024-09-23 12:47:19,844 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.84 vs. limit=15.0
2024-09-23 12:47:25,760 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.263e+02 1.374e+02 1.561e+02 5.695e+02, threshold=2.748e+02, percent-clipped=2.0
2024-09-23 12:47:30,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=261585.33333333334, ans=0.0
2024-09-23 12:47:46,895 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.20 vs. limit=22.5
2024-09-23 12:47:53,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=261632.0, ans=0.2
2024-09-23 12:48:00,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=261678.66666666666, ans=0.0
2024-09-23 12:48:11,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=261678.66666666666, ans=0.125
2024-09-23 12:48:13,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=261725.33333333334, ans=0.125
2024-09-23 12:48:24,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=261725.33333333334, ans=0.0
2024-09-23 12:48:30,873 INFO [train.py:1198] (1/4) Epoch 15, batch 1550, loss[loss=0.2174, ctc_loss=0.144, cr_loss=0.3668, over 17008.00 frames. ], tot_loss[loss=0.2295, ctc_loss=0.1557, cr_loss=0.369, over 3343359.06 frames. ], batch size: 44, lr: 8.07e-03, grad_scale: 16.0
2024-09-23 12:49:19,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.56 vs. limit=22.5
2024-09-23 12:49:53,461 INFO [train.py:1198] (1/4) Epoch 15, batch 1600, loss[loss=0.2082, ctc_loss=0.1374, cr_loss=0.3539, over 17053.00 frames. ], tot_loss[loss=0.2297, ctc_loss=0.1559, cr_loss=0.3688, over 3339827.44 frames. ], batch size: 46, lr: 8.06e-03, grad_scale: 32.0
2024-09-23 12:49:58,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=262005.33333333334, ans=0.0
2024-09-23 12:50:00,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=262005.33333333334, ans=0.0
2024-09-23 12:50:13,468 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.270e+02 1.410e+02 1.627e+02 2.274e+02, threshold=2.820e+02, percent-clipped=0.0
2024-09-23 12:50:28,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=262098.66666666666, ans=0.125
2024-09-23 12:50:31,283 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.68 vs. limit=15.0
2024-09-23 12:51:17,820 INFO [train.py:1198] (1/4) Epoch 15, batch 1650, loss[loss=0.2283, ctc_loss=0.1501, cr_loss=0.391, over 17359.00 frames. ], tot_loss[loss=0.2309, ctc_loss=0.1568, cr_loss=0.3708, over 3340804.71 frames. ], batch size: 48, lr: 8.06e-03, grad_scale: 32.0
2024-09-23 12:51:34,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.28 vs. limit=15.0
2024-09-23 12:51:53,451 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.63 vs. limit=22.5
2024-09-23 12:52:04,838 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.90 vs. limit=15.0
2024-09-23 12:52:18,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=262378.6666666667, ans=0.125
2024-09-23 12:52:21,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=262425.3333333333, ans=0.0
2024-09-23 12:52:27,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=262425.3333333333, ans=0.035
2024-09-23 12:52:39,797 INFO [train.py:1198] (1/4) Epoch 15, batch 1700, loss[loss=0.2316, ctc_loss=0.1562, cr_loss=0.3774, over 17268.00 frames. ], tot_loss[loss=0.231, ctc_loss=0.1568, cr_loss=0.3711, over 3347988.26 frames. ], batch size: 44, lr: 8.06e-03, grad_scale: 32.0
2024-09-23 12:52:46,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=262472.0, ans=0.2
2024-09-23 12:52:57,186 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.272e+02 1.393e+02 1.539e+02 2.504e+02, threshold=2.785e+02, percent-clipped=0.0
2024-09-23 12:53:32,493 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.53 vs. limit=6.0
2024-09-23 12:53:43,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=262658.6666666667, ans=0.0
2024-09-23 12:53:58,954 INFO [train.py:1198] (1/4) Epoch 15, batch 1750, loss[loss=0.2152, ctc_loss=0.1437, cr_loss=0.358, over 17195.00 frames. ], tot_loss[loss=0.2313, ctc_loss=0.157, cr_loss=0.3718, over 3353734.94 frames. ], batch size: 41, lr: 8.05e-03, grad_scale: 32.0
2024-09-23 12:54:02,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=262705.3333333333, ans=0.125
2024-09-23 12:54:21,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=262752.0, ans=0.2
2024-09-23 12:54:49,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=262845.3333333333, ans=0.0
2024-09-23 12:55:00,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=262845.3333333333, ans=0.09899494936611666
2024-09-23 12:55:26,237 INFO [train.py:1198] (1/4) Epoch 15, batch 1800, loss[loss=0.2316, ctc_loss=0.1549, cr_loss=0.3832, over 17218.00 frames. ], tot_loss[loss=0.2308, ctc_loss=0.1566, cr_loss=0.3713, over 3357916.13 frames. ], batch size: 50, lr: 8.05e-03, grad_scale: 32.0
2024-09-23 12:55:43,903 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.283e+02 1.353e+02 1.483e+02 2.243e+02, threshold=2.705e+02, percent-clipped=0.0
2024-09-23 12:55:45,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=262985.3333333333, ans=0.0
2024-09-23 12:55:47,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=262985.3333333333, ans=0.1
2024-09-23 12:55:48,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=262985.3333333333, ans=0.035
2024-09-23 12:55:52,100 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=262985.3333333333, ans=0.125
2024-09-23 12:55:55,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=262985.3333333333, ans=0.125
2024-09-23 12:56:14,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=263078.6666666667, ans=0.2
2024-09-23 12:56:16,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=263078.6666666667, ans=0.0
2024-09-23 12:56:31,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=263125.3333333333, ans=0.1
2024-09-23 12:56:39,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=263125.3333333333, ans=0.125
2024-09-23 12:56:46,104 INFO [train.py:1198] (1/4) Epoch 15, batch 1850, loss[loss=0.2793, ctc_loss=0.1966, cr_loss=0.4137, over 14932.00 frames. ], tot_loss[loss=0.2302, ctc_loss=0.1562, cr_loss=0.3699, over 3353342.26 frames. ], batch size: 89, lr: 8.05e-03, grad_scale: 32.0
2024-09-23 12:56:54,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=263172.0, ans=0.125
2024-09-23 12:56:57,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=263172.0, ans=0.2
2024-09-23 12:57:17,005 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 12:57:18,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=263265.3333333333, ans=0.125
2024-09-23 12:57:40,672 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.06 vs. limit=15.0
2024-09-23 12:57:54,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=263358.6666666667, ans=0.125
2024-09-23 12:57:59,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=263358.6666666667, ans=0.0
2024-09-23 12:58:08,513 INFO [train.py:1198] (1/4) Epoch 15, batch 1900, loss[loss=0.25, ctc_loss=0.1755, cr_loss=0.3721, over 15402.00 frames. ], tot_loss[loss=0.2302, ctc_loss=0.1562, cr_loss=0.37, over 3360563.61 frames. ], batch size: 89, lr: 8.04e-03, grad_scale: 32.0
2024-09-23 12:58:23,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=263452.0, ans=0.0
2024-09-23 12:58:24,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=263452.0, ans=0.0
2024-09-23 12:58:26,130 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.271e+02 1.385e+02 1.549e+02 2.232e+02, threshold=2.769e+02, percent-clipped=0.0
2024-09-23 12:59:27,838 INFO [train.py:1198] (1/4) Epoch 15, batch 1950, loss[loss=0.2465, ctc_loss=0.1659, cr_loss=0.4031, over 17104.00 frames. ], tot_loss[loss=0.231, ctc_loss=0.1568, cr_loss=0.3711, over 3346938.48 frames. ], batch size: 49, lr: 8.04e-03, grad_scale: 16.0
2024-09-23 12:59:31,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=263638.6666666667, ans=0.025
2024-09-23 12:59:35,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=263638.6666666667, ans=0.5
2024-09-23 12:59:43,823 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.24 vs. limit=15.0
2024-09-23 13:00:05,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=263732.0, ans=0.0
2024-09-23 13:00:07,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=263732.0, ans=0.125
2024-09-23 13:00:08,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=263732.0, ans=0.07
2024-09-23 13:00:21,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=263778.6666666667, ans=0.5
2024-09-23 13:00:35,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=263778.6666666667, ans=0.035
2024-09-23 13:00:54,761 INFO [train.py:1198] (1/4) Epoch 15, batch 2000, loss[loss=0.2173, ctc_loss=0.1452, cr_loss=0.3605, over 16732.00 frames. ], tot_loss[loss=0.2304, ctc_loss=0.1563, cr_loss=0.3705, over 3342955.75 frames. ], batch size: 61, lr: 8.04e-03, grad_scale: 32.0
2024-09-23 13:01:12,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=263918.6666666667, ans=0.0
2024-09-23 13:01:14,001 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.286e+02 1.410e+02 1.590e+02 2.174e+02, threshold=2.819e+02, percent-clipped=0.0
2024-09-23 13:01:16,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.73 vs. limit=15.0
2024-09-23 13:01:30,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=263965.3333333333, ans=0.0
2024-09-23 13:01:34,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=263965.3333333333, ans=0.125
2024-09-23 13:01:58,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=264058.6666666667, ans=0.125
2024-09-23 13:02:00,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=264058.6666666667, ans=0.0
2024-09-23 13:02:16,688 INFO [train.py:1198] (1/4) Epoch 15, batch 2050, loss[loss=0.2616, ctc_loss=0.183, cr_loss=0.3932, over 15947.00 frames. ], tot_loss[loss=0.2307, ctc_loss=0.1566, cr_loss=0.3705, over 3329757.24 frames. ], batch size: 74, lr: 8.03e-03, grad_scale: 32.0
2024-09-23 13:02:24,109 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.09 vs. limit=12.0
2024-09-23 13:02:28,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=264105.3333333333, ans=0.125
2024-09-23 13:02:41,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=264152.0, ans=0.125
2024-09-23 13:02:53,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=13.09 vs. limit=22.5
2024-09-23 13:03:04,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=264245.3333333333, ans=0.2
2024-09-23 13:03:25,760 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.55 vs. limit=15.0
2024-09-23 13:03:36,265 INFO [train.py:1198] (1/4) Epoch 15, batch 2100, loss[loss=0.2387, ctc_loss=0.1633, cr_loss=0.3767, over 17308.00 frames. ], tot_loss[loss=0.2297, ctc_loss=0.1559, cr_loss=0.3694, over 3341577.41 frames. ], batch size: 49, lr: 8.03e-03, grad_scale: 32.0
2024-09-23 13:03:55,211 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.282e+02 1.387e+02 1.573e+02 2.128e+02, threshold=2.775e+02, percent-clipped=0.0
2024-09-23 13:04:04,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=264385.3333333333, ans=0.5
2024-09-23 13:04:11,770 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.08 vs. limit=15.0
2024-09-23 13:04:14,849 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.46 vs. limit=15.0
2024-09-23 13:04:20,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=264432.0, ans=0.2
2024-09-23 13:04:24,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=264478.6666666667, ans=0.0
2024-09-23 13:04:34,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=264478.6666666667, ans=0.0
2024-09-23 13:04:38,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=264478.6666666667, ans=0.0
2024-09-23 13:04:49,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=264525.3333333333, ans=0.1
2024-09-23 13:04:53,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=264525.3333333333, ans=0.125
2024-09-23 13:05:03,753 INFO [train.py:1198] (1/4) Epoch 15, batch 2150, loss[loss=0.206, ctc_loss=0.1389, cr_loss=0.3356, over 17105.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1558, cr_loss=0.369, over 3344833.95 frames. ], batch size: 40, lr: 8.03e-03, grad_scale: 32.0
2024-09-23 13:05:19,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=264618.6666666667, ans=0.025
2024-09-23 13:05:29,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=264618.6666666667, ans=0.2
2024-09-23 13:05:31,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=264618.6666666667, ans=0.125
2024-09-23 13:05:34,912 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.61 vs. limit=12.0
2024-09-23 13:05:40,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=264665.3333333333, ans=0.125
2024-09-23 13:05:58,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=264712.0, ans=0.025
2024-09-23 13:05:59,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=264712.0, ans=0.95
2024-09-23 13:06:09,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=264758.6666666667, ans=0.1
2024-09-23 13:06:17,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=264758.6666666667, ans=0.1
2024-09-23 13:06:19,801 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.24 vs. limit=22.5
2024-09-23 13:06:23,715 INFO [train.py:1198] (1/4) Epoch 15, batch 2200, loss[loss=0.2634, ctc_loss=0.1862, cr_loss=0.3859, over 15207.00 frames. ], tot_loss[loss=0.2298, ctc_loss=0.1558, cr_loss=0.3699, over 3351443.30 frames. ], batch size: 89, lr: 8.02e-03, grad_scale: 32.0
2024-09-23 13:06:42,813 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.291e+02 1.443e+02 1.638e+02 2.419e+02, threshold=2.885e+02, percent-clipped=0.0
2024-09-23 13:06:52,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=264852.0, ans=0.125
2024-09-23 13:07:29,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=264992.0, ans=0.125
2024-09-23 13:07:29,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=264992.0, ans=0.1
2024-09-23 13:07:32,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=264992.0, ans=0.05
2024-09-23 13:07:46,502 INFO [train.py:1198] (1/4) Epoch 15, batch 2250, loss[loss=0.2648, ctc_loss=0.1864, cr_loss=0.392, over 15215.00 frames. ], tot_loss[loss=0.2303, ctc_loss=0.1561, cr_loss=0.3709, over 3355206.24 frames. ], batch size: 89, lr: 8.02e-03, grad_scale: 32.0
2024-09-23 13:08:09,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=265085.3333333333, ans=0.125
2024-09-23 13:08:13,314 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.02 vs. limit=15.0
2024-09-23 13:08:17,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=265132.0, ans=0.1
2024-09-23 13:08:20,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=265132.0, ans=0.5
2024-09-23 13:08:25,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=265132.0, ans=0.125
2024-09-23 13:08:30,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=265132.0, ans=0.125
2024-09-23 13:08:33,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=265178.6666666667, ans=0.125
2024-09-23 13:09:06,702 INFO [train.py:1198] (1/4) Epoch 15, batch 2300, loss[loss=0.2609, ctc_loss=0.1777, cr_loss=0.416, over 16546.00 frames. ], tot_loss[loss=0.2302, ctc_loss=0.1562, cr_loss=0.3704, over 3351140.51 frames. ], batch size: 66, lr: 8.02e-03, grad_scale: 16.0
2024-09-23 13:09:08,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=265272.0, ans=0.2
2024-09-23 13:09:23,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=265318.6666666667, ans=0.2
2024-09-23 13:09:27,201 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.234e+02 1.309e+02 1.490e+02 3.155e+02, threshold=2.619e+02, percent-clipped=1.0
2024-09-23 13:09:43,505 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.85 vs. limit=22.5
2024-09-23 13:09:49,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=265365.3333333333, ans=0.1
2024-09-23 13:10:06,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=265412.0, ans=0.0
2024-09-23 13:10:21,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=265458.6666666667, ans=0.2
2024-09-23 13:10:24,021 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=16.92 vs. limit=22.5
2024-09-23 13:10:31,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=265458.6666666667, ans=0.2
2024-09-23 13:10:33,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=265505.3333333333, ans=0.125
2024-09-23 13:10:34,409 INFO [train.py:1198] (1/4) Epoch 15, batch 2350, loss[loss=0.2637, ctc_loss=0.1892, cr_loss=0.3728, over 11535.00 frames. ], tot_loss[loss=0.23, ctc_loss=0.1561, cr_loss=0.3695, over 3338160.98 frames. ], batch size: 123, lr: 8.01e-03, grad_scale: 16.0
2024-09-23 13:11:50,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=265692.0, ans=0.125
2024-09-23 13:11:53,674 INFO [train.py:1198] (1/4) Epoch 15, batch 2400, loss[loss=0.2667, ctc_loss=0.183, cr_loss=0.4189, over 17042.00 frames. ], tot_loss[loss=0.2292, ctc_loss=0.1554, cr_loss=0.3692, over 3355404.66 frames. ], batch size: 52, lr: 8.01e-03, grad_scale: 16.0
2024-09-23 13:12:18,951 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.338e+02 1.450e+02 1.574e+02 3.453e+02, threshold=2.900e+02, percent-clipped=1.0
2024-09-23 13:12:29,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=265832.0, ans=0.05
2024-09-23 13:12:35,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=265832.0, ans=0.1
2024-09-23 13:12:42,120 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.20 vs. limit=10.0
2024-09-23 13:12:57,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=265878.6666666667, ans=0.125
2024-09-23 13:13:07,479 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 13:13:16,576 INFO [train.py:1198] (1/4) Epoch 15, batch 2450, loss[loss=0.2193, ctc_loss=0.1447, cr_loss=0.373, over 17257.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1556, cr_loss=0.37, over 3353089.07 frames. ], batch size: 44, lr: 8.00e-03, grad_scale: 16.0
2024-09-23 13:13:49,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=266065.3333333333, ans=0.1
2024-09-23 13:14:14,428 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=266112.0, ans=0.1
2024-09-23 13:14:22,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=266158.6666666667, ans=0.125
2024-09-23 13:14:29,136 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.27 vs. limit=6.0
2024-09-23 13:14:37,961 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.43 vs. limit=22.5
2024-09-23 13:14:38,832 INFO [train.py:1198] (1/4) Epoch 15, batch 2500, loss[loss=0.1862, ctc_loss=0.1218, cr_loss=0.3217, over 16999.00 frames. ], tot_loss[loss=0.2289, ctc_loss=0.155, cr_loss=0.3694, over 3367171.88 frames. ], batch size: 39, lr: 8.00e-03, grad_scale: 16.0
2024-09-23 13:15:01,485 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 13:15:05,874 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.240e+02 1.331e+02 1.479e+02 2.623e+02, threshold=2.663e+02, percent-clipped=0.0
2024-09-23 13:15:54,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=266392.0, ans=0.2
2024-09-23 13:16:01,182 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.85 vs. limit=15.0
2024-09-23 13:16:03,517 INFO [train.py:1198] (1/4) Epoch 15, batch 2550, loss[loss=0.1993, ctc_loss=0.1321, cr_loss=0.336, over 17100.00 frames. ], tot_loss[loss=0.2283, ctc_loss=0.1546, cr_loss=0.3684, over 3367740.84 frames. ], batch size: 43, lr: 8.00e-03, grad_scale: 16.0
2024-09-23 13:16:12,121 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.29 vs. limit=22.5
2024-09-23 13:16:20,499 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.21 vs. limit=10.0
2024-09-23 13:17:25,063 INFO [train.py:1198] (1/4) Epoch 15, batch 2600, loss[loss=0.2456, ctc_loss=0.1674, cr_loss=0.3911, over 16730.00 frames. ], tot_loss[loss=0.2295, ctc_loss=0.1555, cr_loss=0.3699, over 3357291.31 frames. ], batch size: 61, lr: 7.99e-03, grad_scale: 16.0
2024-09-23 13:17:47,325 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.291e+02 1.396e+02 1.519e+02 2.239e+02, threshold=2.791e+02, percent-clipped=0.0
2024-09-23 13:17:52,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=266718.6666666667, ans=0.0
2024-09-23 13:18:14,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=266812.0, ans=0.035
2024-09-23 13:18:16,458 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.86 vs. limit=15.0
2024-09-23 13:18:40,195 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 13:18:44,566 INFO [train.py:1198] (1/4) Epoch 15, batch 2650, loss[loss=0.241, ctc_loss=0.1655, cr_loss=0.3774, over 17047.00 frames. ], tot_loss[loss=0.2297, ctc_loss=0.1557, cr_loss=0.3703, over 3349722.48 frames. ], batch size: 53, lr: 7.99e-03, grad_scale: 16.0
2024-09-23 13:18:57,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=266905.3333333333, ans=0.125
2024-09-23 13:19:12,764 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.47 vs. limit=15.0
2024-09-23 13:19:42,008 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.72 vs. limit=6.0
2024-09-23 13:19:42,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=267045.3333333333, ans=0.0
2024-09-23 13:20:03,283 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.67 vs. limit=15.0
2024-09-23 13:20:04,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=267092.0, ans=0.07
2024-09-23 13:20:12,065 INFO [train.py:1198] (1/4) Epoch 15, batch 2700, loss[loss=0.2906, ctc_loss=0.2116, cr_loss=0.3954, over 11743.00 frames. ], tot_loss[loss=0.2289, ctc_loss=0.155, cr_loss=0.3693, over 3351776.06 frames. ], batch size: 123, lr: 7.99e-03, grad_scale: 16.0
2024-09-23 13:20:34,472 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.269e+02 1.346e+02 1.479e+02 2.065e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-23 13:20:42,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=267232.0, ans=0.0
2024-09-23 13:21:00,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=267278.6666666667, ans=0.125
2024-09-23 13:21:01,910 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 13:21:06,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=267278.6666666667, ans=0.125
2024-09-23 13:21:16,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=267325.3333333333, ans=0.0
2024-09-23 13:21:16,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=267325.3333333333, ans=0.125
2024-09-23 13:21:31,860 INFO [train.py:1198] (1/4) Epoch 15, batch 2750, loss[loss=0.255, ctc_loss=0.1766, cr_loss=0.392, over 17034.00 frames. ], tot_loss[loss=0.2294, ctc_loss=0.1555, cr_loss=0.3696, over 3352119.43 frames. ], batch size: 56, lr: 7.98e-03, grad_scale: 16.0
2024-09-23 13:21:43,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=267372.0, ans=0.0
2024-09-23 13:22:26,475 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=11.84 vs. limit=22.5
2024-09-23 13:22:46,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=267558.6666666667, ans=0.025
2024-09-23 13:22:53,941 INFO [train.py:1198] (1/4) Epoch 15, batch 2800, loss[loss=0.2154, ctc_loss=0.1413, cr_loss=0.3706, over 16693.00 frames. ], tot_loss[loss=0.2301, ctc_loss=0.1559, cr_loss=0.3707, over 3345513.50 frames. ], batch size: 37, lr: 7.98e-03, grad_scale: 32.0
2024-09-23 13:22:54,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=267605.3333333333, ans=0.1
2024-09-23 13:23:17,755 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.278e+02 1.476e+02 1.758e+02 2.429e+02, threshold=2.952e+02, percent-clipped=0.0
2024-09-23 13:23:48,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=267745.3333333333, ans=0.95
2024-09-23 13:23:51,915 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.81 vs. limit=15.0
2024-09-23 13:24:10,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=267792.0, ans=0.125
2024-09-23 13:24:13,372 INFO [train.py:1198] (1/4) Epoch 15, batch 2850, loss[loss=0.2362, ctc_loss=0.159, cr_loss=0.3862, over 17280.00 frames. ], tot_loss[loss=0.229, ctc_loss=0.155, cr_loss=0.3702, over 3359805.64 frames. ], batch size: 46, lr: 7.98e-03, grad_scale: 16.0
2024-09-23 13:24:25,890 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.43 vs. limit=22.5
2024-09-23 13:24:34,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=267885.3333333333, ans=0.1
2024-09-23 13:24:47,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=267885.3333333333, ans=0.015
2024-09-23 13:25:11,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=267978.6666666667, ans=0.0
2024-09-23 13:25:19,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=267978.6666666667, ans=0.125
2024-09-23 13:25:32,862 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.92 vs. limit=6.0
2024-09-23 13:25:41,408 INFO [train.py:1198] (1/4) Epoch 15, batch 2900, loss[loss=0.2266, ctc_loss=0.1516, cr_loss=0.3747, over 17250.00 frames. ], tot_loss[loss=0.2293, ctc_loss=0.1552, cr_loss=0.3701, over 3354183.18 frames. ], batch size: 44, lr: 7.97e-03, grad_scale: 16.0
2024-09-23 13:25:49,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=268072.0, ans=0.0
2024-09-23 13:25:51,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=268072.0, ans=0.1
2024-09-23 13:26:03,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=268118.6666666667, ans=0.2
2024-09-23 13:26:04,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=268118.6666666667, ans=0.125
2024-09-23 13:26:05,969 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.264e+02 1.372e+02 1.552e+02 2.806e+02, threshold=2.744e+02, percent-clipped=0.0
2024-09-23 13:26:29,112 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.81 vs. limit=15.0
2024-09-23 13:26:29,506 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.02 vs. limit=15.0
2024-09-23 13:26:31,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=268212.0, ans=0.2
2024-09-23 13:26:33,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=268212.0, ans=0.125
2024-09-23 13:26:36,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=268212.0, ans=0.125
2024-09-23 13:27:03,973 INFO [train.py:1198] (1/4) Epoch 15, batch 2950, loss[loss=0.1976, ctc_loss=0.1313, cr_loss=0.3314, over 17090.00 frames. ], tot_loss[loss=0.2302, ctc_loss=0.156, cr_loss=0.371, over 3349177.73 frames. ], batch size: 43, lr: 7.97e-03, grad_scale: 16.0
2024-09-23 13:27:05,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=268305.3333333333, ans=0.125
2024-09-23 13:27:13,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=268305.3333333333, ans=0.1
2024-09-23 13:27:32,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=268352.0, ans=0.0
2024-09-23 13:27:49,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=268398.6666666667, ans=0.125
2024-09-23 13:27:53,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=268445.3333333333, ans=0.0
2024-09-23 13:28:23,347 INFO [train.py:1198] (1/4) Epoch 15, batch 3000, loss[loss=0.2498, ctc_loss=0.1684, cr_loss=0.4067, over 16967.00 frames. ], tot_loss[loss=0.2289, ctc_loss=0.155, cr_loss=0.3697, over 3350912.63 frames. ], batch size: 42, lr: 7.97e-03, grad_scale: 16.0
2024-09-23 13:28:23,347 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 13:28:33,130 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.2.self_attn_weights, attn_weights_entropy = tensor([2.3260, 2.7953, 2.7660, 3.0753, 2.5845, 2.6737, 3.0491, 3.1376],
       device='cuda:1')
2024-09-23 13:28:38,973 INFO [train.py:1230] (1/4) Epoch 15, validation: loss=0.04166, ctc_loss=0.04166, cr_loss=7.464e-15, over 944034.00 frames. 
2024-09-23 13:28:38,973 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 13:28:41,678 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=9.43 vs. limit=15.0
2024-09-23 13:29:02,495 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.290e+02 1.376e+02 1.476e+02 2.234e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-23 13:29:20,481 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.17 vs. limit=10.0
2024-09-23 13:29:35,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=268678.6666666667, ans=0.125
2024-09-23 13:29:35,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=268678.6666666667, ans=0.125
2024-09-23 13:29:35,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=268678.6666666667, ans=0.125
2024-09-23 13:29:56,909 INFO [train.py:1198] (1/4) Epoch 15, batch 3050, loss[loss=0.1977, ctc_loss=0.1309, cr_loss=0.334, over 17099.00 frames. ], tot_loss[loss=0.2295, ctc_loss=0.1554, cr_loss=0.3702, over 3339162.54 frames. ], batch size: 43, lr: 7.96e-03, grad_scale: 16.0
2024-09-23 13:30:04,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=268772.0, ans=0.125
2024-09-23 13:30:17,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=268818.6666666667, ans=0.125
2024-09-23 13:30:36,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=268865.3333333333, ans=0.125
2024-09-23 13:30:57,544 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.90 vs. limit=15.0
2024-09-23 13:31:01,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=268912.0, ans=0.2
2024-09-23 13:31:02,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=268912.0, ans=0.125
2024-09-23 13:31:12,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=268958.6666666667, ans=0.125
2024-09-23 13:31:22,605 INFO [train.py:1198] (1/4) Epoch 15, batch 3100, loss[loss=0.2571, ctc_loss=0.1713, cr_loss=0.4289, over 17007.00 frames. ], tot_loss[loss=0.2302, ctc_loss=0.1559, cr_loss=0.3714, over 3326052.67 frames. ], batch size: 51, lr: 7.96e-03, grad_scale: 16.0
2024-09-23 13:31:32,386 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 13:31:41,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=269052.0, ans=0.125
2024-09-23 13:31:46,078 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.323e+02 1.397e+02 1.534e+02 2.167e+02, threshold=2.795e+02, percent-clipped=0.0
2024-09-23 13:31:46,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=269052.0, ans=0.125
2024-09-23 13:31:49,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=269052.0, ans=15.0
2024-09-23 13:32:03,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=269098.6666666667, ans=0.1
2024-09-23 13:32:17,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=269145.3333333333, ans=0.125
2024-09-23 13:32:41,149 INFO [train.py:1198] (1/4) Epoch 15, batch 3150, loss[loss=0.248, ctc_loss=0.168, cr_loss=0.4003, over 17205.00 frames. ], tot_loss[loss=0.23, ctc_loss=0.1557, cr_loss=0.3713, over 3336889.87 frames. ], batch size: 50, lr: 7.96e-03, grad_scale: 16.0
2024-09-23 13:32:58,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=269285.3333333333, ans=0.1
2024-09-23 13:33:25,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=269332.0, ans=0.09899494936611666
2024-09-23 13:33:37,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=269378.6666666667, ans=0.125
2024-09-23 13:33:44,762 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=11.97 vs. limit=22.5
2024-09-23 13:33:46,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=269425.3333333333, ans=0.0
2024-09-23 13:33:59,102 INFO [train.py:1198] (1/4) Epoch 15, batch 3200, loss[loss=0.2193, ctc_loss=0.1475, cr_loss=0.3594, over 17072.00 frames. ], tot_loss[loss=0.2311, ctc_loss=0.1566, cr_loss=0.3726, over 3345136.45 frames. ], batch size: 46, lr: 7.95e-03, grad_scale: 32.0
2024-09-23 13:34:18,902 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.95 vs. limit=22.5
2024-09-23 13:34:22,376 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.268e+02 1.359e+02 1.495e+02 2.696e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-23 13:34:28,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=269565.3333333333, ans=0.1
2024-09-23 13:34:44,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=269612.0, ans=0.125
2024-09-23 13:35:17,051 INFO [train.py:1198] (1/4) Epoch 15, batch 3250, loss[loss=0.2022, ctc_loss=0.1359, cr_loss=0.3316, over 17180.00 frames. ], tot_loss[loss=0.2306, ctc_loss=0.1563, cr_loss=0.3717, over 3339379.37 frames. ], batch size: 41, lr: 7.95e-03, grad_scale: 32.0
2024-09-23 13:35:20,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=269705.3333333333, ans=0.04949747468305833
2024-09-23 13:35:25,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=269705.3333333333, ans=0.07
2024-09-23 13:35:36,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=269752.0, ans=0.0
2024-09-23 13:35:49,099 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=269798.6666666667, ans=0.2
2024-09-23 13:35:52,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=269798.6666666667, ans=0.2
2024-09-23 13:36:03,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=269798.6666666667, ans=0.125
2024-09-23 13:36:14,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=269845.3333333333, ans=0.0
2024-09-23 13:36:28,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=269892.0, ans=0.0
2024-09-23 13:36:33,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=269892.0, ans=0.1
2024-09-23 13:36:37,977 INFO [train.py:1198] (1/4) Epoch 15, batch 3300, loss[loss=0.2145, ctc_loss=0.1417, cr_loss=0.3643, over 17280.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1556, cr_loss=0.3701, over 3341190.99 frames. ], batch size: 46, lr: 7.95e-03, grad_scale: 32.0
2024-09-23 13:36:50,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=269938.6666666667, ans=0.07
2024-09-23 13:36:52,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=269985.3333333333, ans=0.0
2024-09-23 13:37:01,532 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.277e+02 1.362e+02 1.536e+02 4.994e+02, threshold=2.723e+02, percent-clipped=1.0
2024-09-23 13:37:34,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=270078.6666666667, ans=0.025
2024-09-23 13:37:55,866 INFO [train.py:1198] (1/4) Epoch 15, batch 3350, loss[loss=0.2083, ctc_loss=0.1413, cr_loss=0.3351, over 17234.00 frames. ], tot_loss[loss=0.2291, ctc_loss=0.1553, cr_loss=0.369, over 3347790.82 frames. ], batch size: 47, lr: 7.94e-03, grad_scale: 32.0
2024-09-23 13:38:05,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=270172.0, ans=0.125
2024-09-23 13:38:14,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=270218.6666666667, ans=0.1
2024-09-23 13:38:16,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=270218.6666666667, ans=0.1
2024-09-23 13:38:22,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=270218.6666666667, ans=0.0
2024-09-23 13:38:25,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=270265.3333333333, ans=0.0
2024-09-23 13:38:35,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=270265.3333333333, ans=0.0
2024-09-23 13:38:44,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=270312.0, ans=0.0
2024-09-23 13:39:01,715 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 13:39:14,019 INFO [train.py:1198] (1/4) Epoch 15, batch 3400, loss[loss=0.255, ctc_loss=0.1727, cr_loss=0.411, over 17035.00 frames. ], tot_loss[loss=0.2293, ctc_loss=0.1554, cr_loss=0.3697, over 3346429.48 frames. ], batch size: 52, lr: 7.94e-03, grad_scale: 32.0
2024-09-23 13:39:14,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=270405.3333333333, ans=0.125
2024-09-23 13:39:23,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=270405.3333333333, ans=0.09899494936611666
2024-09-23 13:39:37,359 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.273e+02 1.371e+02 1.521e+02 2.083e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-23 13:39:43,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=270498.6666666667, ans=0.125
2024-09-23 13:39:47,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=270498.6666666667, ans=0.1
2024-09-23 13:39:59,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=270545.3333333333, ans=0.2
2024-09-23 13:40:01,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=270545.3333333333, ans=0.125
2024-09-23 13:40:02,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=270545.3333333333, ans=0.1
2024-09-23 13:40:20,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=270592.0, ans=0.125
2024-09-23 13:40:25,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.65 vs. limit=22.5
2024-09-23 13:40:32,268 INFO [train.py:1198] (1/4) Epoch 15, batch 3450, loss[loss=0.2185, ctc_loss=0.144, cr_loss=0.3726, over 17005.00 frames. ], tot_loss[loss=0.2291, ctc_loss=0.1552, cr_loss=0.3695, over 3352017.62 frames. ], batch size: 56, lr: 7.94e-03, grad_scale: 32.0
2024-09-23 13:40:32,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=270638.6666666667, ans=0.125
2024-09-23 13:40:42,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=270638.6666666667, ans=0.125
2024-09-23 13:40:53,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=270685.3333333333, ans=0.025
2024-09-23 13:41:44,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=270825.3333333333, ans=0.125
2024-09-23 13:41:56,575 INFO [train.py:1198] (1/4) Epoch 15, batch 3500, loss[loss=0.2184, ctc_loss=0.1469, cr_loss=0.3577, over 17337.00 frames. ], tot_loss[loss=0.229, ctc_loss=0.1552, cr_loss=0.3693, over 3352060.42 frames. ], batch size: 48, lr: 7.93e-03, grad_scale: 32.0
2024-09-23 13:42:18,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=270918.6666666667, ans=0.0
2024-09-23 13:42:19,966 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.027e+02 1.268e+02 1.406e+02 1.575e+02 2.426e+02, threshold=2.811e+02, percent-clipped=0.0
2024-09-23 13:42:33,737 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.01 vs. limit=6.0
2024-09-23 13:43:15,425 INFO [train.py:1198] (1/4) Epoch 15, batch 3550, loss[loss=0.2177, ctc_loss=0.1463, cr_loss=0.3568, over 17219.00 frames. ], tot_loss[loss=0.2292, ctc_loss=0.1554, cr_loss=0.3692, over 3350275.88 frames. ], batch size: 47, lr: 7.93e-03, grad_scale: 32.0
2024-09-23 13:43:59,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=271198.6666666667, ans=0.125
2024-09-23 13:44:07,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.20 vs. limit=15.0
2024-09-23 13:44:18,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=271292.0, ans=0.125
2024-09-23 13:44:33,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.93 vs. limit=12.0
2024-09-23 13:44:33,800 INFO [train.py:1198] (1/4) Epoch 15, batch 3600, loss[loss=0.2738, ctc_loss=0.189, cr_loss=0.4239, over 15362.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1558, cr_loss=0.369, over 3339227.85 frames. ], batch size: 89, lr: 7.93e-03, grad_scale: 32.0
2024-09-23 13:44:36,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=271338.6666666667, ans=0.125
2024-09-23 13:44:55,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=271385.3333333333, ans=0.125
2024-09-23 13:44:57,246 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.288e+02 1.401e+02 1.561e+02 2.194e+02, threshold=2.802e+02, percent-clipped=0.0
2024-09-23 13:45:53,864 INFO [train.py:1198] (1/4) Epoch 15, batch 3650, loss[loss=0.226, ctc_loss=0.1535, cr_loss=0.3626, over 16861.00 frames. ], tot_loss[loss=0.2291, ctc_loss=0.1555, cr_loss=0.3682, over 3344107.52 frames. ], batch size: 58, lr: 7.92e-03, grad_scale: 16.0
2024-09-23 13:45:58,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=271572.0, ans=0.125
2024-09-23 13:46:57,649 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.32 vs. limit=15.0
2024-09-23 13:47:12,692 INFO [train.py:1198] (1/4) Epoch 15, batch 3700, loss[loss=0.3126, ctc_loss=0.2191, cr_loss=0.4676, over 14986.00 frames. ], tot_loss[loss=0.2296, ctc_loss=0.1558, cr_loss=0.3689, over 3338082.28 frames. ], batch size: 90, lr: 7.92e-03, grad_scale: 16.0
2024-09-23 13:47:12,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=271805.3333333333, ans=0.125
2024-09-23 13:47:19,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=271805.3333333333, ans=0.0
2024-09-23 13:47:37,539 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.268e+02 1.357e+02 1.522e+02 2.745e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-23 13:47:53,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=271898.6666666667, ans=0.025
2024-09-23 13:48:01,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=271945.3333333333, ans=0.125
2024-09-23 13:48:31,495 INFO [train.py:1198] (1/4) Epoch 15, batch 3750, loss[loss=0.2398, ctc_loss=0.1677, cr_loss=0.3602, over 15117.00 frames. ], tot_loss[loss=0.2289, ctc_loss=0.1553, cr_loss=0.3679, over 3341842.27 frames. ], batch size: 89, lr: 7.92e-03, grad_scale: 16.0
2024-09-23 13:48:31,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=272038.6666666667, ans=0.025
2024-09-23 13:49:03,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=272132.0, ans=0.125
2024-09-23 13:49:11,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=272132.0, ans=0.1
2024-09-23 13:49:11,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=272132.0, ans=0.125
2024-09-23 13:49:50,617 INFO [train.py:1198] (1/4) Epoch 15, batch 3800, loss[loss=0.2092, ctc_loss=0.1387, cr_loss=0.3526, over 17167.00 frames. ], tot_loss[loss=0.2287, ctc_loss=0.1551, cr_loss=0.368, over 3321828.56 frames. ], batch size: 41, lr: 7.91e-03, grad_scale: 16.0
2024-09-23 13:49:57,856 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=5.72 vs. limit=15.0
2024-09-23 13:50:16,159 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.263e+02 1.390e+02 1.534e+02 1.887e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-23 13:50:25,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=272365.3333333333, ans=0.125
2024-09-23 13:50:38,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.min_positive, batch_count=272412.0, ans=0.05
2024-09-23 13:50:41,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=272412.0, ans=0.1
2024-09-23 13:51:08,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=272458.6666666667, ans=0.0
2024-09-23 13:51:11,675 INFO [train.py:1198] (1/4) Epoch 15, batch 3850, loss[loss=0.2282, ctc_loss=0.1498, cr_loss=0.3921, over 17306.00 frames. ], tot_loss[loss=0.23, ctc_loss=0.1561, cr_loss=0.3694, over 3296673.19 frames. ], batch size: 46, lr: 7.91e-03, grad_scale: 16.0
2024-09-23 13:51:33,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=272552.0, ans=0.035
2024-09-23 13:51:36,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=272552.0, ans=0.035
2024-09-23 13:51:40,571 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.70 vs. limit=15.0
2024-09-23 13:51:47,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=272598.6666666667, ans=0.125
2024-09-23 13:51:56,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=272645.3333333333, ans=0.125
2024-09-23 13:53:14,135 INFO [train.py:1198] (1/4) Epoch 16, batch 0, loss[loss=0.2638, ctc_loss=0.1904, cr_loss=0.3671, over 12089.00 frames. ], tot_loss[loss=0.2638, ctc_loss=0.1904, cr_loss=0.3671, over 12089.00 frames. ], batch size: 124, lr: 7.65e-03, grad_scale: 32.0
2024-09-23 13:53:14,135 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 13:53:26,826 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.1.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([5.9982, 5.9285, 5.5864, 5.7284], device='cuda:1')
2024-09-23 13:53:30,201 INFO [train.py:1230] (1/4) Epoch 16, validation: loss=0.04222, ctc_loss=0.04222, cr_loss=7.738e-15, over 944034.00 frames. 
2024-09-23 13:53:30,202 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 13:53:35,571 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.84 vs. limit=15.0
2024-09-23 13:54:02,002 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.456e+02 1.611e+02 1.770e+02 2.340e+02, threshold=3.223e+02, percent-clipped=0.0
2024-09-23 13:54:32,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=272906.6666666667, ans=0.125
2024-09-23 13:54:32,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=272906.6666666667, ans=0.1
2024-09-23 13:54:39,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=272906.6666666667, ans=0.025
2024-09-23 13:54:50,285 INFO [train.py:1198] (1/4) Epoch 16, batch 50, loss[loss=0.2214, ctc_loss=0.1473, cr_loss=0.3703, over 17299.00 frames. ], tot_loss[loss=0.2237, ctc_loss=0.1509, cr_loss=0.3639, over 757998.30 frames. ], batch size: 46, lr: 7.65e-03, grad_scale: 32.0
2024-09-23 13:55:31,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=273046.6666666667, ans=0.0
2024-09-23 13:56:09,766 INFO [train.py:1198] (1/4) Epoch 16, batch 100, loss[loss=0.2378, ctc_loss=0.1647, cr_loss=0.3655, over 17348.00 frames. ], tot_loss[loss=0.2265, ctc_loss=0.153, cr_loss=0.3677, over 1331753.35 frames. ], batch size: 48, lr: 7.65e-03, grad_scale: 32.0
2024-09-23 13:56:10,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=273186.6666666667, ans=0.125
2024-09-23 13:56:21,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=273186.6666666667, ans=0.1
2024-09-23 13:56:29,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=273186.6666666667, ans=0.125
2024-09-23 13:56:51,567 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.239e+02 1.320e+02 1.510e+02 1.777e+02, threshold=2.639e+02, percent-clipped=0.0
2024-09-23 13:57:17,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=273326.6666666667, ans=0.07
2024-09-23 13:57:21,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=273373.3333333333, ans=0.2
2024-09-23 13:57:38,812 INFO [train.py:1198] (1/4) Epoch 16, batch 150, loss[loss=0.2133, ctc_loss=0.1435, cr_loss=0.3489, over 17215.00 frames. ], tot_loss[loss=0.2276, ctc_loss=0.1536, cr_loss=0.3702, over 1790027.78 frames. ], batch size: 47, lr: 7.64e-03, grad_scale: 32.0
2024-09-23 13:58:03,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=273466.6666666667, ans=0.1
2024-09-23 13:58:22,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=273513.3333333333, ans=0.0
2024-09-23 13:58:52,076 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.85 vs. limit=15.0
2024-09-23 13:58:59,125 INFO [train.py:1198] (1/4) Epoch 16, batch 200, loss[loss=0.2901, ctc_loss=0.2105, cr_loss=0.3979, over 11860.00 frames. ], tot_loss[loss=0.2283, ctc_loss=0.1544, cr_loss=0.3697, over 2127243.38 frames. ], batch size: 123, lr: 7.64e-03, grad_scale: 32.0
2024-09-23 13:59:10,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=273653.3333333333, ans=0.035
2024-09-23 13:59:26,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=273700.0, ans=0.125
2024-09-23 13:59:31,015 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.259e+02 1.374e+02 1.473e+02 2.348e+02, threshold=2.748e+02, percent-clipped=0.0
2024-09-23 13:59:31,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=273746.6666666667, ans=0.125
2024-09-23 13:59:36,187 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 13:59:56,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=273793.3333333333, ans=0.125
2024-09-23 13:59:58,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=273793.3333333333, ans=0.125
2024-09-23 14:00:04,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=273840.0, ans=0.125
2024-09-23 14:00:07,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=273840.0, ans=0.125
2024-09-23 14:00:14,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=273840.0, ans=0.0
2024-09-23 14:00:18,769 INFO [train.py:1198] (1/4) Epoch 16, batch 250, loss[loss=0.2214, ctc_loss=0.1487, cr_loss=0.3637, over 17352.00 frames. ], tot_loss[loss=0.229, ctc_loss=0.1549, cr_loss=0.3705, over 2399446.82 frames. ], batch size: 48, lr: 7.64e-03, grad_scale: 32.0
2024-09-23 14:00:19,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=273886.6666666667, ans=0.2
2024-09-23 14:00:23,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=273886.6666666667, ans=0.2
2024-09-23 14:00:37,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=273933.3333333333, ans=0.0
2024-09-23 14:01:14,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=274026.6666666667, ans=0.125
2024-09-23 14:01:46,158 INFO [train.py:1198] (1/4) Epoch 16, batch 300, loss[loss=0.2324, ctc_loss=0.1581, cr_loss=0.3715, over 16993.00 frames. ], tot_loss[loss=0.2274, ctc_loss=0.1537, cr_loss=0.3682, over 2622626.96 frames. ], batch size: 44, lr: 7.63e-03, grad_scale: 32.0
2024-09-23 14:02:22,709 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.271e+02 1.353e+02 1.525e+02 2.781e+02, threshold=2.705e+02, percent-clipped=1.0
2024-09-23 14:02:28,143 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.28 vs. limit=22.5
2024-09-23 14:02:50,252 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 14:03:08,654 INFO [train.py:1198] (1/4) Epoch 16, batch 350, loss[loss=0.2355, ctc_loss=0.1582, cr_loss=0.3866, over 16762.00 frames. ], tot_loss[loss=0.2278, ctc_loss=0.154, cr_loss=0.369, over 2790873.28 frames. ], batch size: 61, lr: 7.63e-03, grad_scale: 16.0
2024-09-23 14:03:23,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=274400.0, ans=0.5
2024-09-23 14:03:35,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=274400.0, ans=0.125
2024-09-23 14:03:46,114 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.12 vs. limit=15.0
2024-09-23 14:03:52,398 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.61 vs. limit=22.5
2024-09-23 14:03:53,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=274446.6666666667, ans=0.0
2024-09-23 14:03:59,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=274493.3333333333, ans=0.09899494936611666
2024-09-23 14:04:28,539 INFO [train.py:1198] (1/4) Epoch 16, batch 400, loss[loss=0.22, ctc_loss=0.1498, cr_loss=0.3509, over 17276.00 frames. ], tot_loss[loss=0.2261, ctc_loss=0.1528, cr_loss=0.3668, over 2922102.49 frames. ], batch size: 51, lr: 7.63e-03, grad_scale: 32.0
2024-09-23 14:04:36,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=274586.6666666667, ans=0.0
2024-09-23 14:04:51,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=274633.3333333333, ans=0.125
2024-09-23 14:04:55,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=274633.3333333333, ans=0.125
2024-09-23 14:04:59,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.53 vs. limit=15.0
2024-09-23 14:05:01,858 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.271e+02 1.344e+02 1.500e+02 2.841e+02, threshold=2.689e+02, percent-clipped=1.0
2024-09-23 14:05:02,230 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 14:05:32,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=274773.3333333333, ans=0.025
2024-09-23 14:05:43,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=274773.3333333333, ans=0.025
2024-09-23 14:05:43,965 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.42 vs. limit=15.0
2024-09-23 14:05:47,970 INFO [train.py:1198] (1/4) Epoch 16, batch 450, loss[loss=0.1892, ctc_loss=0.1279, cr_loss=0.3067, over 17137.00 frames. ], tot_loss[loss=0.2271, ctc_loss=0.1535, cr_loss=0.3677, over 3017649.62 frames. ], batch size: 45, lr: 7.62e-03, grad_scale: 32.0
2024-09-23 14:06:02,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=274866.6666666667, ans=0.1
2024-09-23 14:06:30,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=274913.3333333333, ans=0.125
2024-09-23 14:07:00,119 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 14:07:00,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=275006.6666666667, ans=0.0
2024-09-23 14:07:08,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=275006.6666666667, ans=0.025
2024-09-23 14:07:15,703 INFO [train.py:1198] (1/4) Epoch 16, batch 500, loss[loss=0.1916, ctc_loss=0.1272, cr_loss=0.3217, over 17129.00 frames. ], tot_loss[loss=0.2256, ctc_loss=0.1524, cr_loss=0.3662, over 3103267.10 frames. ], batch size: 40, lr: 7.62e-03, grad_scale: 16.0
2024-09-23 14:07:22,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=275053.3333333333, ans=0.125
2024-09-23 14:07:30,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=275100.0, ans=0.0
2024-09-23 14:07:33,970 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.01 vs. limit=15.0
2024-09-23 14:07:38,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=275100.0, ans=0.0
2024-09-23 14:07:52,534 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.254e+02 1.336e+02 1.469e+02 8.615e+02, threshold=2.672e+02, percent-clipped=1.0
2024-09-23 14:08:06,219 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.54 vs. limit=10.0
2024-09-23 14:08:18,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=275240.0, ans=0.0
2024-09-23 14:08:28,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=275240.0, ans=0.125
2024-09-23 14:08:31,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=275240.0, ans=0.125
2024-09-23 14:08:36,360 INFO [train.py:1198] (1/4) Epoch 16, batch 550, loss[loss=0.2355, ctc_loss=0.1583, cr_loss=0.3861, over 17068.00 frames. ], tot_loss[loss=0.225, ctc_loss=0.1518, cr_loss=0.3659, over 3169747.57 frames. ], batch size: 43, lr: 7.62e-03, grad_scale: 8.0
2024-09-23 14:08:38,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=275286.6666666667, ans=0.1
2024-09-23 14:08:39,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=275286.6666666667, ans=0.125
2024-09-23 14:08:59,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=275333.3333333333, ans=0.2
2024-09-23 14:09:05,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=275333.3333333333, ans=0.025
2024-09-23 14:09:56,355 INFO [train.py:1198] (1/4) Epoch 16, batch 600, loss[loss=0.2932, ctc_loss=0.2085, cr_loss=0.4234, over 11620.00 frames. ], tot_loss[loss=0.2251, ctc_loss=0.1519, cr_loss=0.3661, over 3205117.94 frames. ], batch size: 124, lr: 7.61e-03, grad_scale: 8.0
2024-09-23 14:10:10,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=275566.6666666667, ans=0.0
2024-09-23 14:10:32,682 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.280e+02 1.400e+02 1.549e+02 2.106e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-23 14:10:53,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=275660.0, ans=0.0
2024-09-23 14:11:03,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=275706.6666666667, ans=0.125
2024-09-23 14:11:20,853 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.93 vs. limit=22.5
2024-09-23 14:11:21,416 INFO [train.py:1198] (1/4) Epoch 16, batch 650, loss[loss=0.2005, ctc_loss=0.1341, cr_loss=0.3319, over 17117.00 frames. ], tot_loss[loss=0.2246, ctc_loss=0.1516, cr_loss=0.3648, over 3237093.83 frames. ], batch size: 40, lr: 7.61e-03, grad_scale: 8.0
2024-09-23 14:12:07,534 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.74 vs. limit=15.0
2024-09-23 14:12:15,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=275893.3333333333, ans=0.2
2024-09-23 14:12:43,638 INFO [train.py:1198] (1/4) Epoch 16, batch 700, loss[loss=0.2423, ctc_loss=0.1643, cr_loss=0.3902, over 16877.00 frames. ], tot_loss[loss=0.2259, ctc_loss=0.1526, cr_loss=0.3666, over 3265874.02 frames. ], batch size: 58, lr: 7.61e-03, grad_scale: 8.0
2024-09-23 14:12:45,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=275986.6666666667, ans=0.0
2024-09-23 14:13:20,906 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.263e+02 1.367e+02 1.500e+02 2.228e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-23 14:13:41,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=276126.6666666667, ans=6.0
2024-09-23 14:13:46,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=276173.3333333333, ans=0.125
2024-09-23 14:14:03,949 INFO [train.py:1198] (1/4) Epoch 16, batch 750, loss[loss=0.2487, ctc_loss=0.1697, cr_loss=0.395, over 17239.00 frames. ], tot_loss[loss=0.2264, ctc_loss=0.1529, cr_loss=0.3676, over 3290447.12 frames. ], batch size: 55, lr: 7.61e-03, grad_scale: 8.0
2024-09-23 14:14:05,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=276220.0, ans=0.0
2024-09-23 14:14:06,220 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.46 vs. limit=15.0
2024-09-23 14:14:49,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.37 vs. limit=15.0
2024-09-23 14:15:23,830 INFO [train.py:1198] (1/4) Epoch 16, batch 800, loss[loss=0.1891, ctc_loss=0.1263, cr_loss=0.314, over 17079.00 frames. ], tot_loss[loss=0.2275, ctc_loss=0.1537, cr_loss=0.3689, over 3305476.21 frames. ], batch size: 43, lr: 7.60e-03, grad_scale: 16.0
2024-09-23 14:15:30,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=276453.3333333333, ans=0.125
2024-09-23 14:15:38,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=276500.0, ans=0.125
2024-09-23 14:15:44,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=276500.0, ans=0.025
2024-09-23 14:15:58,434 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.42 vs. limit=22.5
2024-09-23 14:15:59,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=276546.6666666667, ans=0.0
2024-09-23 14:16:00,467 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.250e+02 1.343e+02 1.464e+02 3.040e+02, threshold=2.687e+02, percent-clipped=1.0
2024-09-23 14:16:25,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=276593.3333333333, ans=0.1
2024-09-23 14:16:37,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=276640.0, ans=0.1
2024-09-23 14:16:51,108 INFO [train.py:1198] (1/4) Epoch 16, batch 850, loss[loss=0.2107, ctc_loss=0.1407, cr_loss=0.35, over 17169.00 frames. ], tot_loss[loss=0.2278, ctc_loss=0.1539, cr_loss=0.3691, over 3321915.28 frames. ], batch size: 41, lr: 7.60e-03, grad_scale: 16.0
2024-09-23 14:16:53,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=276686.6666666667, ans=0.025
2024-09-23 14:17:12,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=276733.3333333333, ans=0.125
2024-09-23 14:17:20,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=276733.3333333333, ans=0.125
2024-09-23 14:17:31,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.max_abs, batch_count=276780.0, ans=10.0
2024-09-23 14:17:32,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=276780.0, ans=0.125
2024-09-23 14:17:41,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=276826.6666666667, ans=0.0
2024-09-23 14:18:04,495 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=13.50 vs. limit=22.5
2024-09-23 14:18:06,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=276873.3333333333, ans=0.1
2024-09-23 14:18:08,959 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.11 vs. limit=15.0
2024-09-23 14:18:11,440 INFO [train.py:1198] (1/4) Epoch 16, batch 900, loss[loss=0.219, ctc_loss=0.1462, cr_loss=0.364, over 17192.00 frames. ], tot_loss[loss=0.2273, ctc_loss=0.1537, cr_loss=0.3681, over 3333562.39 frames. ], batch size: 47, lr: 7.60e-03, grad_scale: 16.0
2024-09-23 14:18:19,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=276920.0, ans=0.0
2024-09-23 14:18:19,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=276920.0, ans=0.125
2024-09-23 14:18:23,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.22 vs. limit=15.0
2024-09-23 14:18:29,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=276966.6666666667, ans=0.0
2024-09-23 14:18:43,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=277013.3333333333, ans=0.125
2024-09-23 14:18:47,947 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.278e+02 1.412e+02 1.648e+02 4.971e+02, threshold=2.824e+02, percent-clipped=1.0
2024-09-23 14:18:48,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=277013.3333333333, ans=0.1
2024-09-23 14:19:07,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=277060.0, ans=0.125
2024-09-23 14:19:26,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=277106.6666666667, ans=0.125
2024-09-23 14:19:30,110 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.18 vs. limit=15.0
2024-09-23 14:19:30,870 INFO [train.py:1198] (1/4) Epoch 16, batch 950, loss[loss=0.253, ctc_loss=0.1704, cr_loss=0.413, over 17230.00 frames. ], tot_loss[loss=0.227, ctc_loss=0.1535, cr_loss=0.3678, over 3338364.42 frames. ], batch size: 55, lr: 7.59e-03, grad_scale: 16.0
2024-09-23 14:19:43,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=277153.3333333333, ans=0.125
2024-09-23 14:20:04,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=277246.6666666667, ans=0.2
2024-09-23 14:20:06,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=277246.6666666667, ans=0.0
2024-09-23 14:20:06,809 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.38 vs. limit=22.5
2024-09-23 14:20:14,131 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 14:20:29,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=277293.3333333333, ans=0.1
2024-09-23 14:20:50,274 INFO [train.py:1198] (1/4) Epoch 16, batch 1000, loss[loss=0.2331, ctc_loss=0.1611, cr_loss=0.3599, over 16842.00 frames. ], tot_loss[loss=0.2266, ctc_loss=0.1532, cr_loss=0.3675, over 3347039.17 frames. ], batch size: 61, lr: 7.59e-03, grad_scale: 16.0
2024-09-23 14:20:50,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=277386.6666666667, ans=0.05
2024-09-23 14:21:17,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=277433.3333333333, ans=0.125
2024-09-23 14:21:22,867 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.79 vs. limit=10.0
2024-09-23 14:21:25,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=277433.3333333333, ans=0.0
2024-09-23 14:21:37,360 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.289e+02 1.371e+02 1.510e+02 2.522e+02, threshold=2.742e+02, percent-clipped=0.0
2024-09-23 14:21:40,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=277480.0, ans=0.125
2024-09-23 14:21:59,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=277526.6666666667, ans=0.125
2024-09-23 14:22:01,968 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.87 vs. limit=15.0
2024-09-23 14:22:09,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=277573.3333333333, ans=0.125
2024-09-23 14:22:14,315 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 14:22:20,155 INFO [train.py:1198] (1/4) Epoch 16, batch 1050, loss[loss=0.2271, ctc_loss=0.1517, cr_loss=0.3771, over 17258.00 frames. ], tot_loss[loss=0.2259, ctc_loss=0.1527, cr_loss=0.366, over 3349395.30 frames. ], batch size: 44, lr: 7.59e-03, grad_scale: 16.0
2024-09-23 14:22:27,348 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.70 vs. limit=15.0
2024-09-23 14:22:33,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=277620.0, ans=0.0
2024-09-23 14:22:37,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=277666.6666666667, ans=0.1
2024-09-23 14:22:48,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=277666.6666666667, ans=0.125
2024-09-23 14:23:37,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=277806.6666666667, ans=0.0
2024-09-23 14:23:39,992 INFO [train.py:1198] (1/4) Epoch 16, batch 1100, loss[loss=0.2177, ctc_loss=0.148, cr_loss=0.3486, over 17105.00 frames. ], tot_loss[loss=0.2255, ctc_loss=0.1523, cr_loss=0.3658, over 3361847.06 frames. ], batch size: 49, lr: 7.58e-03, grad_scale: 16.0
2024-09-23 14:24:06,256 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=7.89 vs. limit=15.0
2024-09-23 14:24:12,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=277946.6666666667, ans=0.125
2024-09-23 14:24:16,322 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.265e+02 1.345e+02 1.521e+02 2.827e+02, threshold=2.691e+02, percent-clipped=1.0
2024-09-23 14:24:20,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.18 vs. limit=22.5
2024-09-23 14:24:37,638 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.83 vs. limit=22.5
2024-09-23 14:24:40,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=277993.3333333333, ans=0.1
2024-09-23 14:24:58,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=278086.6666666667, ans=0.125
2024-09-23 14:24:59,427 INFO [train.py:1198] (1/4) Epoch 16, batch 1150, loss[loss=0.2058, ctc_loss=0.1387, cr_loss=0.3353, over 16953.00 frames. ], tot_loss[loss=0.2245, ctc_loss=0.1516, cr_loss=0.3643, over 3365244.31 frames. ], batch size: 42, lr: 7.58e-03, grad_scale: 16.0
2024-09-23 14:25:09,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=278086.6666666667, ans=0.1
2024-09-23 14:25:32,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=278180.0, ans=0.125
2024-09-23 14:25:38,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=278180.0, ans=0.125
2024-09-23 14:25:56,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=278226.6666666667, ans=0.0
2024-09-23 14:26:10,882 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.52 vs. limit=15.0
2024-09-23 14:26:13,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=278273.3333333333, ans=0.125
2024-09-23 14:26:19,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=278273.3333333333, ans=0.1
2024-09-23 14:26:27,404 INFO [train.py:1198] (1/4) Epoch 16, batch 1200, loss[loss=0.2444, ctc_loss=0.1655, cr_loss=0.3945, over 17287.00 frames. ], tot_loss[loss=0.2247, ctc_loss=0.1518, cr_loss=0.3648, over 3364026.98 frames. ], batch size: 49, lr: 7.58e-03, grad_scale: 16.0
2024-09-23 14:26:29,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=278320.0, ans=0.125
2024-09-23 14:26:52,815 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.07 vs. limit=15.0
2024-09-23 14:27:08,387 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.263e+02 1.375e+02 1.495e+02 2.178e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-23 14:27:16,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=278460.0, ans=0.05
2024-09-23 14:27:24,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=278460.0, ans=0.0
2024-09-23 14:27:26,757 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.57 vs. limit=15.0
2024-09-23 14:27:29,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.25 vs. limit=22.5
2024-09-23 14:27:49,956 INFO [train.py:1198] (1/4) Epoch 16, batch 1250, loss[loss=0.2129, ctc_loss=0.1395, cr_loss=0.3668, over 17300.00 frames. ], tot_loss[loss=0.2252, ctc_loss=0.1521, cr_loss=0.3659, over 3367026.93 frames. ], batch size: 46, lr: 7.57e-03, grad_scale: 16.0
2024-09-23 14:27:53,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.80 vs. limit=15.0
2024-09-23 14:28:06,735 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.05 vs. limit=15.0
2024-09-23 14:28:18,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=278600.0, ans=0.1
2024-09-23 14:28:36,899 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=93.66 vs. limit=15.0
2024-09-23 14:28:39,925 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 14:29:10,195 INFO [train.py:1198] (1/4) Epoch 16, batch 1300, loss[loss=0.2542, ctc_loss=0.1722, cr_loss=0.4099, over 17001.00 frames. ], tot_loss[loss=0.2257, ctc_loss=0.1524, cr_loss=0.3667, over 3367168.54 frames. ], batch size: 56, lr: 7.57e-03, grad_scale: 16.0
2024-09-23 14:29:43,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=278880.0, ans=0.025
2024-09-23 14:29:48,132 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.264e+02 1.359e+02 1.551e+02 2.304e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-23 14:30:29,687 INFO [train.py:1198] (1/4) Epoch 16, batch 1350, loss[loss=0.208, ctc_loss=0.1368, cr_loss=0.3565, over 17075.00 frames. ], tot_loss[loss=0.2261, ctc_loss=0.1528, cr_loss=0.3665, over 3359382.16 frames. ], batch size: 39, lr: 7.57e-03, grad_scale: 16.0
2024-09-23 14:30:41,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=279020.0, ans=0.125
2024-09-23 14:31:14,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=279113.3333333333, ans=0.125
2024-09-23 14:31:56,197 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.80 vs. limit=10.0
2024-09-23 14:32:00,029 INFO [train.py:1198] (1/4) Epoch 16, batch 1400, loss[loss=0.2275, ctc_loss=0.1519, cr_loss=0.3779, over 17160.00 frames. ], tot_loss[loss=0.2256, ctc_loss=0.1524, cr_loss=0.366, over 3361581.04 frames. ], batch size: 45, lr: 7.56e-03, grad_scale: 16.0
2024-09-23 14:32:03,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=279253.3333333333, ans=0.125
2024-09-23 14:32:11,808 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=7.99 vs. limit=12.0
2024-09-23 14:32:27,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.31 vs. limit=22.5
2024-09-23 14:32:37,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=279346.6666666667, ans=0.125
2024-09-23 14:32:38,349 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.268e+02 1.363e+02 1.512e+02 2.184e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-23 14:32:51,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=279393.3333333333, ans=0.1
2024-09-23 14:32:56,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=279393.3333333333, ans=0.2
2024-09-23 14:32:59,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=279393.3333333333, ans=0.0
2024-09-23 14:33:17,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=279440.0, ans=0.125
2024-09-23 14:33:20,075 INFO [train.py:1198] (1/4) Epoch 16, batch 1450, loss[loss=0.2313, ctc_loss=0.1571, cr_loss=0.3706, over 16866.00 frames. ], tot_loss[loss=0.2244, ctc_loss=0.1515, cr_loss=0.3647, over 3367855.36 frames. ], batch size: 58, lr: 7.56e-03, grad_scale: 16.0
2024-09-23 14:33:26,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=279486.6666666667, ans=0.125
2024-09-23 14:33:55,901 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.32 vs. limit=15.0
2024-09-23 14:34:25,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=279673.3333333333, ans=0.0
2024-09-23 14:34:27,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=279673.3333333333, ans=0.0
2024-09-23 14:34:37,453 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.79 vs. limit=6.0
2024-09-23 14:34:39,814 INFO [train.py:1198] (1/4) Epoch 16, batch 1500, loss[loss=0.2303, ctc_loss=0.1539, cr_loss=0.3822, over 17254.00 frames. ], tot_loss[loss=0.2254, ctc_loss=0.1522, cr_loss=0.366, over 3367239.82 frames. ], batch size: 44, lr: 7.56e-03, grad_scale: 16.0
2024-09-23 14:34:44,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=279720.0, ans=0.1
2024-09-23 14:34:48,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=279720.0, ans=0.125
2024-09-23 14:35:15,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=279813.3333333333, ans=0.125
2024-09-23 14:35:16,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=279813.3333333333, ans=0.0
2024-09-23 14:35:18,097 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.287e+02 1.379e+02 1.521e+02 2.599e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-23 14:35:52,771 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.58 vs. limit=10.0
2024-09-23 14:36:06,618 INFO [train.py:1198] (1/4) Epoch 16, batch 1550, loss[loss=0.2156, ctc_loss=0.1442, cr_loss=0.3566, over 17306.00 frames. ], tot_loss[loss=0.2243, ctc_loss=0.1514, cr_loss=0.3645, over 3369086.32 frames. ], batch size: 46, lr: 7.56e-03, grad_scale: 16.0
2024-09-23 14:36:37,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=280000.0, ans=0.2
2024-09-23 14:36:37,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=280000.0, ans=0.0
2024-09-23 14:36:51,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=280046.6666666667, ans=0.05
2024-09-23 14:37:07,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=280093.3333333333, ans=0.1
2024-09-23 14:37:13,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=280140.0, ans=0.025
2024-09-23 14:37:23,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=280140.0, ans=0.0
2024-09-23 14:37:30,855 INFO [train.py:1198] (1/4) Epoch 16, batch 1600, loss[loss=0.2457, ctc_loss=0.1679, cr_loss=0.3894, over 17248.00 frames. ], tot_loss[loss=0.2245, ctc_loss=0.1514, cr_loss=0.3651, over 3369977.09 frames. ], batch size: 55, lr: 7.55e-03, grad_scale: 32.0
2024-09-23 14:37:46,033 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.32 vs. limit=15.0
2024-09-23 14:37:48,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=280233.3333333333, ans=0.0
2024-09-23 14:38:00,552 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.16 vs. limit=15.0
2024-09-23 14:38:08,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=280280.0, ans=0.2
2024-09-23 14:38:10,324 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.239e+02 1.329e+02 1.436e+02 2.606e+02, threshold=2.657e+02, percent-clipped=0.0
2024-09-23 14:38:15,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=280280.0, ans=0.0
2024-09-23 14:38:17,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=280326.6666666667, ans=0.1
2024-09-23 14:38:40,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=280373.3333333333, ans=0.025
2024-09-23 14:38:50,312 INFO [train.py:1198] (1/4) Epoch 16, batch 1650, loss[loss=0.2426, ctc_loss=0.1599, cr_loss=0.4135, over 17057.00 frames. ], tot_loss[loss=0.226, ctc_loss=0.1526, cr_loss=0.3667, over 3357157.19 frames. ], batch size: 46, lr: 7.55e-03, grad_scale: 16.0
2024-09-23 14:38:53,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=280420.0, ans=10.0
2024-09-23 14:39:01,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=280420.0, ans=0.2
2024-09-23 14:39:08,484 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.98 vs. limit=10.0
2024-09-23 14:39:31,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=280513.3333333333, ans=0.0
2024-09-23 14:39:46,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=280560.0, ans=0.125
2024-09-23 14:39:57,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=280606.6666666667, ans=0.1
2024-09-23 14:40:02,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=280606.6666666667, ans=0.1
2024-09-23 14:40:09,864 INFO [train.py:1198] (1/4) Epoch 16, batch 1700, loss[loss=0.235, ctc_loss=0.1618, cr_loss=0.3659, over 16380.00 frames. ], tot_loss[loss=0.2254, ctc_loss=0.1521, cr_loss=0.3666, over 3364357.41 frames. ], batch size: 66, lr: 7.55e-03, grad_scale: 16.0
2024-09-23 14:40:13,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=280653.3333333333, ans=0.125
2024-09-23 14:40:31,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=280700.0, ans=0.125
2024-09-23 14:40:38,019 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.97 vs. limit=22.5
2024-09-23 14:40:43,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=280746.6666666667, ans=0.125
2024-09-23 14:40:52,361 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.241e+02 1.314e+02 1.415e+02 2.347e+02, threshold=2.628e+02, percent-clipped=0.0
2024-09-23 14:41:09,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=280793.3333333333, ans=0.025
2024-09-23 14:41:09,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=280793.3333333333, ans=0.1
2024-09-23 14:41:15,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=280793.3333333333, ans=0.2
2024-09-23 14:41:17,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=280793.3333333333, ans=0.1
2024-09-23 14:41:40,568 INFO [train.py:1198] (1/4) Epoch 16, batch 1750, loss[loss=0.2474, ctc_loss=0.17, cr_loss=0.387, over 17213.00 frames. ], tot_loss[loss=0.2258, ctc_loss=0.1524, cr_loss=0.3671, over 3362917.19 frames. ], batch size: 55, lr: 7.54e-03, grad_scale: 16.0
2024-09-23 14:41:47,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=280886.6666666667, ans=0.0
2024-09-23 14:41:57,558 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.65 vs. limit=15.0
2024-09-23 14:42:05,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=280933.3333333333, ans=0.125
2024-09-23 14:42:08,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.max_positive, batch_count=280933.3333333333, ans=0.95
2024-09-23 14:42:17,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=280980.0, ans=0.125
2024-09-23 14:43:00,732 INFO [train.py:1198] (1/4) Epoch 16, batch 1800, loss[loss=0.2523, ctc_loss=0.1734, cr_loss=0.3943, over 16789.00 frames. ], tot_loss[loss=0.2252, ctc_loss=0.1519, cr_loss=0.3664, over 3363707.68 frames. ], batch size: 61, lr: 7.54e-03, grad_scale: 16.0
2024-09-23 14:43:18,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=281166.6666666667, ans=0.05
2024-09-23 14:43:27,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=281166.6666666667, ans=0.0
2024-09-23 14:43:36,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.75 vs. limit=12.0
2024-09-23 14:43:40,372 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.265e+02 1.365e+02 1.521e+02 2.085e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-23 14:44:09,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=281306.6666666667, ans=0.125
2024-09-23 14:44:20,023 INFO [train.py:1198] (1/4) Epoch 16, batch 1850, loss[loss=0.239, ctc_loss=0.1623, cr_loss=0.3837, over 16997.00 frames. ], tot_loss[loss=0.2262, ctc_loss=0.1528, cr_loss=0.3671, over 3358280.43 frames. ], batch size: 53, lr: 7.54e-03, grad_scale: 16.0
2024-09-23 14:44:32,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=281353.3333333333, ans=0.125
2024-09-23 14:44:34,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=281400.0, ans=0.125
2024-09-23 14:44:36,322 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.76 vs. limit=15.0
2024-09-23 14:44:56,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=281446.6666666667, ans=0.125
2024-09-23 14:45:23,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=281540.0, ans=0.07
2024-09-23 14:45:42,170 INFO [train.py:1198] (1/4) Epoch 16, batch 1900, loss[loss=0.2078, ctc_loss=0.1386, cr_loss=0.3461, over 17113.00 frames. ], tot_loss[loss=0.2259, ctc_loss=0.1525, cr_loss=0.3667, over 3359798.42 frames. ], batch size: 49, lr: 7.53e-03, grad_scale: 16.0
2024-09-23 14:45:43,237 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.03 vs. limit=12.0
2024-09-23 14:45:56,621 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=281586.6666666667, ans=0.04949747468305833
2024-09-23 14:45:58,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=281586.6666666667, ans=0.125
2024-09-23 14:45:58,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=281586.6666666667, ans=0.0
2024-09-23 14:46:27,054 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.270e+02 1.346e+02 1.420e+02 2.030e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-23 14:47:09,367 INFO [train.py:1198] (1/4) Epoch 16, batch 1950, loss[loss=0.2432, ctc_loss=0.1692, cr_loss=0.3699, over 17353.00 frames. ], tot_loss[loss=0.2263, ctc_loss=0.1528, cr_loss=0.3676, over 3366541.89 frames. ], batch size: 48, lr: 7.53e-03, grad_scale: 16.0
2024-09-23 14:47:09,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=281820.0, ans=0.04949747468305833
2024-09-23 14:47:16,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=281820.0, ans=0.0
2024-09-23 14:47:17,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.min_positive, batch_count=281820.0, ans=0.025
2024-09-23 14:47:20,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=281820.0, ans=0.0
2024-09-23 14:47:26,568 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.42 vs. limit=15.0
2024-09-23 14:48:04,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=281960.0, ans=0.2
2024-09-23 14:48:06,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=281960.0, ans=0.2
2024-09-23 14:48:29,653 INFO [train.py:1198] (1/4) Epoch 16, batch 2000, loss[loss=0.1909, ctc_loss=0.1287, cr_loss=0.3111, over 17168.00 frames. ], tot_loss[loss=0.2261, ctc_loss=0.1527, cr_loss=0.367, over 3370997.83 frames. ], batch size: 45, lr: 7.53e-03, grad_scale: 32.0
2024-09-23 14:48:42,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=282053.3333333333, ans=0.0
2024-09-23 14:49:09,154 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.302e+02 1.409e+02 1.610e+02 3.619e+02, threshold=2.818e+02, percent-clipped=1.0
2024-09-23 14:49:17,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=282193.3333333333, ans=0.0
2024-09-23 14:49:26,141 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.43 vs. limit=6.0
2024-09-23 14:49:49,170 INFO [train.py:1198] (1/4) Epoch 16, batch 2050, loss[loss=0.2336, ctc_loss=0.1546, cr_loss=0.395, over 17048.00 frames. ], tot_loss[loss=0.2252, ctc_loss=0.152, cr_loss=0.3659, over 3371704.93 frames. ], batch size: 52, lr: 7.52e-03, grad_scale: 32.0
2024-09-23 14:50:00,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=282286.6666666667, ans=0.0
2024-09-23 14:50:18,549 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.90 vs. limit=15.0
2024-09-23 14:50:19,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=282380.0, ans=0.0
2024-09-23 14:50:48,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=282426.6666666667, ans=0.1
2024-09-23 14:51:04,528 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.80 vs. limit=22.5
2024-09-23 14:51:13,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=282473.3333333333, ans=0.125
2024-09-23 14:51:16,373 INFO [train.py:1198] (1/4) Epoch 16, batch 2100, loss[loss=0.1933, ctc_loss=0.1304, cr_loss=0.3147, over 17280.00 frames. ], tot_loss[loss=0.2246, ctc_loss=0.1516, cr_loss=0.3652, over 3374088.82 frames. ], batch size: 42, lr: 7.52e-03, grad_scale: 32.0
2024-09-23 14:51:33,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=282566.6666666667, ans=0.125
2024-09-23 14:51:58,360 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.277e+02 1.347e+02 1.477e+02 3.259e+02, threshold=2.693e+02, percent-clipped=1.0
2024-09-23 14:52:38,057 INFO [train.py:1198] (1/4) Epoch 16, batch 2150, loss[loss=0.2107, ctc_loss=0.1386, cr_loss=0.3606, over 17302.00 frames. ], tot_loss[loss=0.2254, ctc_loss=0.1522, cr_loss=0.3662, over 3370386.78 frames. ], batch size: 46, lr: 7.52e-03, grad_scale: 32.0
2024-09-23 14:52:48,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=282753.3333333333, ans=0.2
2024-09-23 14:52:56,771 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.31 vs. limit=6.0
2024-09-23 14:52:57,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=282800.0, ans=0.0
2024-09-23 14:53:25,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=282893.3333333333, ans=0.125
2024-09-23 14:53:33,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=282893.3333333333, ans=0.2
2024-09-23 14:53:40,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=282940.0, ans=0.125
2024-09-23 14:53:58,127 INFO [train.py:1198] (1/4) Epoch 16, batch 2200, loss[loss=0.229, ctc_loss=0.1522, cr_loss=0.3838, over 17196.00 frames. ], tot_loss[loss=0.2242, ctc_loss=0.1512, cr_loss=0.3651, over 3375318.87 frames. ], batch size: 47, lr: 7.52e-03, grad_scale: 32.0
2024-09-23 14:54:19,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=283033.3333333333, ans=0.125
2024-09-23 14:54:21,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=283033.3333333333, ans=0.2
2024-09-23 14:54:38,169 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.245e+02 1.360e+02 1.489e+02 2.276e+02, threshold=2.720e+02, percent-clipped=0.0
2024-09-23 14:55:14,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=283173.3333333333, ans=0.125
2024-09-23 14:55:18,498 INFO [train.py:1198] (1/4) Epoch 16, batch 2250, loss[loss=0.2315, ctc_loss=0.1557, cr_loss=0.3787, over 17161.00 frames. ], tot_loss[loss=0.223, ctc_loss=0.1502, cr_loss=0.3643, over 3383534.64 frames. ], batch size: 48, lr: 7.51e-03, grad_scale: 32.0
2024-09-23 14:55:52,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=283266.6666666667, ans=0.025
2024-09-23 14:56:16,008 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.31 vs. limit=10.0
2024-09-23 14:56:18,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=283360.0, ans=0.125
2024-09-23 14:56:36,153 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=6.85 vs. limit=15.0
2024-09-23 14:56:43,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=283406.6666666667, ans=0.2
2024-09-23 14:56:47,891 INFO [train.py:1198] (1/4) Epoch 16, batch 2300, loss[loss=0.2401, ctc_loss=0.1609, cr_loss=0.3958, over 17337.00 frames. ], tot_loss[loss=0.2231, ctc_loss=0.1502, cr_loss=0.3643, over 3378534.80 frames. ], batch size: 48, lr: 7.51e-03, grad_scale: 32.0
2024-09-23 14:57:20,428 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.66 vs. limit=15.0
2024-09-23 14:57:27,769 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.282e+02 1.381e+02 1.559e+02 2.607e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-23 14:57:28,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.89 vs. limit=6.0
2024-09-23 14:57:40,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=283593.3333333333, ans=0.125
2024-09-23 14:57:56,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=283640.0, ans=0.0
2024-09-23 14:58:07,689 INFO [train.py:1198] (1/4) Epoch 16, batch 2350, loss[loss=0.2275, ctc_loss=0.155, cr_loss=0.3623, over 17203.00 frames. ], tot_loss[loss=0.224, ctc_loss=0.1509, cr_loss=0.3655, over 3370690.17 frames. ], batch size: 50, lr: 7.51e-03, grad_scale: 32.0
2024-09-23 14:58:51,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=283780.0, ans=0.125
2024-09-23 14:59:24,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=283873.3333333333, ans=0.2
2024-09-23 14:59:27,616 INFO [train.py:1198] (1/4) Epoch 16, batch 2400, loss[loss=0.2597, ctc_loss=0.178, cr_loss=0.4085, over 17162.00 frames. ], tot_loss[loss=0.2245, ctc_loss=0.1512, cr_loss=0.3663, over 3370962.09 frames. ], batch size: 48, lr: 7.50e-03, grad_scale: 32.0
2024-09-23 14:59:31,403 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.64 vs. limit=15.0
2024-09-23 14:59:32,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=283920.0, ans=0.125
2024-09-23 14:59:37,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=283920.0, ans=0.125
2024-09-23 14:59:46,048 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.57 vs. limit=6.0
2024-09-23 15:00:07,588 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.274e+02 1.411e+02 1.575e+02 2.045e+02, threshold=2.822e+02, percent-clipped=0.0
2024-09-23 15:00:09,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.81 vs. limit=15.0
2024-09-23 15:00:14,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=284060.0, ans=0.2
2024-09-23 15:00:20,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=284060.0, ans=0.125
2024-09-23 15:00:20,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=284060.0, ans=0.1
2024-09-23 15:00:22,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=284060.0, ans=15.0
2024-09-23 15:00:46,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=284106.6666666667, ans=0.125
2024-09-23 15:00:51,304 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.44 vs. limit=8.0
2024-09-23 15:00:54,749 INFO [train.py:1198] (1/4) Epoch 16, batch 2450, loss[loss=0.2011, ctc_loss=0.1323, cr_loss=0.3444, over 17192.00 frames. ], tot_loss[loss=0.2248, ctc_loss=0.1515, cr_loss=0.3666, over 3371340.26 frames. ], batch size: 41, lr: 7.50e-03, grad_scale: 32.0
2024-09-23 15:01:02,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=284153.3333333333, ans=0.1
2024-09-23 15:01:32,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=284246.6666666667, ans=0.125
2024-09-23 15:01:53,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=284293.3333333333, ans=0.125
2024-09-23 15:02:17,030 INFO [train.py:1198] (1/4) Epoch 16, batch 2500, loss[loss=0.2083, ctc_loss=0.1391, cr_loss=0.3461, over 17158.00 frames. ], tot_loss[loss=0.2246, ctc_loss=0.1513, cr_loss=0.3664, over 3360747.68 frames. ], batch size: 45, lr: 7.50e-03, grad_scale: 32.0
2024-09-23 15:02:20,812 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.90 vs. limit=10.0
2024-09-23 15:02:34,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=284433.3333333333, ans=0.125
2024-09-23 15:02:56,744 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.256e+02 1.352e+02 1.487e+02 2.472e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-23 15:03:06,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=284526.6666666667, ans=0.0
2024-09-23 15:03:13,312 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.49 vs. limit=15.0
2024-09-23 15:03:36,420 INFO [train.py:1198] (1/4) Epoch 16, batch 2550, loss[loss=0.2234, ctc_loss=0.1493, cr_loss=0.3702, over 17011.00 frames. ], tot_loss[loss=0.2247, ctc_loss=0.1514, cr_loss=0.3665, over 3357425.33 frames. ], batch size: 51, lr: 7.49e-03, grad_scale: 32.0
2024-09-23 15:03:41,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=284620.0, ans=0.125
2024-09-23 15:03:46,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=284620.0, ans=0.125
2024-09-23 15:04:01,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=284666.6666666667, ans=0.125
2024-09-23 15:04:05,489 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.25 vs. limit=8.0
2024-09-23 15:04:06,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=284666.6666666667, ans=0.07
2024-09-23 15:04:43,298 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.18 vs. limit=22.5
2024-09-23 15:04:55,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=284853.3333333333, ans=0.0
2024-09-23 15:04:56,919 INFO [train.py:1198] (1/4) Epoch 16, batch 2600, loss[loss=0.2023, ctc_loss=0.1358, cr_loss=0.3327, over 17207.00 frames. ], tot_loss[loss=0.2256, ctc_loss=0.152, cr_loss=0.3676, over 3358251.11 frames. ], batch size: 47, lr: 7.49e-03, grad_scale: 32.0
2024-09-23 15:04:58,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=284853.3333333333, ans=0.125
2024-09-23 15:05:02,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=284853.3333333333, ans=0.025
2024-09-23 15:05:06,679 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 15:05:41,680 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.252e+02 1.364e+02 1.589e+02 2.408e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-23 15:05:43,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=284946.6666666667, ans=0.1
2024-09-23 15:05:43,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=284946.6666666667, ans=0.2
2024-09-23 15:06:05,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=284993.3333333333, ans=0.0
2024-09-23 15:06:21,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=285040.0, ans=0.2
2024-09-23 15:06:24,112 INFO [train.py:1198] (1/4) Epoch 16, batch 2650, loss[loss=0.2275, ctc_loss=0.1529, cr_loss=0.3731, over 17303.00 frames. ], tot_loss[loss=0.2258, ctc_loss=0.1522, cr_loss=0.3678, over 3368433.43 frames. ], batch size: 49, lr: 7.49e-03, grad_scale: 32.0
2024-09-23 15:06:35,413 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.46 vs. limit=15.0
2024-09-23 15:06:52,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=285133.3333333333, ans=0.125
2024-09-23 15:07:07,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=285180.0, ans=0.5
2024-09-23 15:07:23,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=285226.6666666667, ans=0.0
2024-09-23 15:07:40,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=285273.3333333333, ans=0.125
2024-09-23 15:07:46,887 INFO [train.py:1198] (1/4) Epoch 16, batch 2700, loss[loss=0.1871, ctc_loss=0.121, cr_loss=0.3302, over 17280.00 frames. ], tot_loss[loss=0.2244, ctc_loss=0.1512, cr_loss=0.3659, over 3373861.98 frames. ], batch size: 42, lr: 7.48e-03, grad_scale: 32.0
2024-09-23 15:08:20,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=285413.3333333333, ans=0.1
2024-09-23 15:08:26,735 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.250e+02 1.328e+02 1.437e+02 2.275e+02, threshold=2.655e+02, percent-clipped=0.0
2024-09-23 15:08:28,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=285413.3333333333, ans=0.2
2024-09-23 15:08:50,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=285506.6666666667, ans=0.5
2024-09-23 15:08:52,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=285506.6666666667, ans=0.0
2024-09-23 15:09:06,362 INFO [train.py:1198] (1/4) Epoch 16, batch 2750, loss[loss=0.2045, ctc_loss=0.1382, cr_loss=0.3316, over 17014.00 frames. ], tot_loss[loss=0.2261, ctc_loss=0.1526, cr_loss=0.3677, over 3369629.55 frames. ], batch size: 44, lr: 7.48e-03, grad_scale: 32.0
2024-09-23 15:09:38,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=285646.6666666667, ans=0.2
2024-09-23 15:09:46,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=285646.6666666667, ans=0.09899494936611666
2024-09-23 15:10:23,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=285740.0, ans=0.125
2024-09-23 15:10:26,019 INFO [train.py:1198] (1/4) Epoch 16, batch 2800, loss[loss=0.2189, ctc_loss=0.1436, cr_loss=0.3766, over 17134.00 frames. ], tot_loss[loss=0.2258, ctc_loss=0.1524, cr_loss=0.3669, over 3366172.67 frames. ], batch size: 48, lr: 7.48e-03, grad_scale: 32.0
2024-09-23 15:10:32,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=285786.6666666667, ans=0.0
2024-09-23 15:11:11,237 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.268e+02 1.378e+02 1.547e+02 2.101e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-23 15:11:13,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=285880.0, ans=0.125
2024-09-23 15:11:53,455 INFO [train.py:1198] (1/4) Epoch 16, batch 2850, loss[loss=0.2652, ctc_loss=0.1786, cr_loss=0.4333, over 17017.00 frames. ], tot_loss[loss=0.226, ctc_loss=0.1525, cr_loss=0.3675, over 3367830.10 frames. ], batch size: 56, lr: 7.48e-03, grad_scale: 32.0
2024-09-23 15:11:56,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=286020.0, ans=0.1
2024-09-23 15:12:07,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=286066.6666666667, ans=0.0
2024-09-23 15:12:12,023 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.01 vs. limit=6.0
2024-09-23 15:12:14,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=286066.6666666667, ans=0.0
2024-09-23 15:12:34,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=286113.3333333333, ans=0.95
2024-09-23 15:12:38,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=286113.3333333333, ans=0.125
2024-09-23 15:13:13,161 INFO [train.py:1198] (1/4) Epoch 16, batch 2900, loss[loss=0.2551, ctc_loss=0.1749, cr_loss=0.4008, over 16531.00 frames. ], tot_loss[loss=0.2259, ctc_loss=0.1525, cr_loss=0.3673, over 3353911.55 frames. ], batch size: 66, lr: 7.47e-03, grad_scale: 32.0
2024-09-23 15:13:32,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=286300.0, ans=0.125
2024-09-23 15:13:37,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=286300.0, ans=0.025
2024-09-23 15:13:48,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=286346.6666666667, ans=0.2
2024-09-23 15:13:53,325 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.305e+02 1.434e+02 1.582e+02 2.466e+02, threshold=2.868e+02, percent-clipped=0.0
2024-09-23 15:14:10,189 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.58 vs. limit=22.5
2024-09-23 15:14:32,144 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 15:14:33,434 INFO [train.py:1198] (1/4) Epoch 16, batch 2950, loss[loss=0.2367, ctc_loss=0.1614, cr_loss=0.3763, over 16895.00 frames. ], tot_loss[loss=0.2265, ctc_loss=0.1529, cr_loss=0.3678, over 3360659.70 frames. ], batch size: 58, lr: 7.47e-03, grad_scale: 32.0
2024-09-23 15:14:54,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=286533.3333333333, ans=0.0
2024-09-23 15:15:00,129 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.14 vs. limit=15.0
2024-09-23 15:15:21,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=286580.0, ans=0.125
2024-09-23 15:15:36,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=286626.6666666667, ans=0.0
2024-09-23 15:15:38,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=286626.6666666667, ans=0.125
2024-09-23 15:15:58,477 INFO [train.py:1198] (1/4) Epoch 16, batch 3000, loss[loss=0.2128, ctc_loss=0.1394, cr_loss=0.3672, over 17096.00 frames. ], tot_loss[loss=0.227, ctc_loss=0.1534, cr_loss=0.3682, over 3357417.30 frames. ], batch size: 43, lr: 7.47e-03, grad_scale: 32.0
2024-09-23 15:15:58,478 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 15:16:14,048 INFO [train.py:1230] (1/4) Epoch 16, validation: loss=0.04215, ctc_loss=0.04215, cr_loss=7.551e-15, over 944034.00 frames. 
2024-09-23 15:16:14,048 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 15:16:39,428 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=286766.6666666667, ans=0.125
2024-09-23 15:16:39,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=286766.6666666667, ans=0.2
2024-09-23 15:16:53,330 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.253e+02 1.384e+02 1.509e+02 2.501e+02, threshold=2.769e+02, percent-clipped=0.0
2024-09-23 15:17:00,278 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.73 vs. limit=15.0
2024-09-23 15:17:31,988 INFO [train.py:1198] (1/4) Epoch 16, batch 3050, loss[loss=0.2131, ctc_loss=0.1427, cr_loss=0.352, over 16947.00 frames. ], tot_loss[loss=0.2255, ctc_loss=0.1522, cr_loss=0.3661, over 3366897.01 frames. ], batch size: 42, lr: 7.46e-03, grad_scale: 32.0
2024-09-23 15:18:05,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=287046.6666666667, ans=0.125
2024-09-23 15:18:16,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=287046.6666666667, ans=0.125
2024-09-23 15:18:38,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=287140.0, ans=0.0
2024-09-23 15:18:50,317 INFO [train.py:1198] (1/4) Epoch 16, batch 3100, loss[loss=0.2227, ctc_loss=0.1503, cr_loss=0.3619, over 16743.00 frames. ], tot_loss[loss=0.226, ctc_loss=0.1526, cr_loss=0.3671, over 3363240.23 frames. ], batch size: 61, lr: 7.46e-03, grad_scale: 32.0
2024-09-23 15:18:57,450 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.60 vs. limit=15.0
2024-09-23 15:18:58,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=287186.6666666667, ans=0.2
2024-09-23 15:19:06,924 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.64 vs. limit=15.0
2024-09-23 15:19:09,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=287233.3333333333, ans=0.125
2024-09-23 15:19:27,947 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 15:19:27,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=287280.0, ans=0.0
2024-09-23 15:19:29,118 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.248e+02 1.355e+02 1.474e+02 3.116e+02, threshold=2.710e+02, percent-clipped=1.0
2024-09-23 15:19:42,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=287326.6666666667, ans=0.125
2024-09-23 15:20:08,480 INFO [train.py:1198] (1/4) Epoch 16, batch 3150, loss[loss=0.2147, ctc_loss=0.1408, cr_loss=0.3697, over 17075.00 frames. ], tot_loss[loss=0.2257, ctc_loss=0.1522, cr_loss=0.3673, over 3351539.99 frames. ], batch size: 46, lr: 7.46e-03, grad_scale: 16.0
2024-09-23 15:20:23,419 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.86 vs. limit=22.5
2024-09-23 15:20:29,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=287466.6666666667, ans=0.125
2024-09-23 15:20:46,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=287513.3333333333, ans=0.2
2024-09-23 15:21:14,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=287606.6666666667, ans=0.0
2024-09-23 15:21:26,302 INFO [train.py:1198] (1/4) Epoch 16, batch 3200, loss[loss=0.2464, ctc_loss=0.1696, cr_loss=0.384, over 16975.00 frames. ], tot_loss[loss=0.2262, ctc_loss=0.1528, cr_loss=0.3674, over 3348435.11 frames. ], batch size: 53, lr: 7.45e-03, grad_scale: 32.0
2024-09-23 15:21:40,839 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.77 vs. limit=12.0
2024-09-23 15:21:59,778 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.42 vs. limit=15.0
2024-09-23 15:22:06,509 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.274e+02 1.402e+02 1.535e+02 1.896e+02, threshold=2.804e+02, percent-clipped=0.0
2024-09-23 15:22:32,232 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.99 vs. limit=12.0
2024-09-23 15:22:43,946 INFO [train.py:1198] (1/4) Epoch 16, batch 3250, loss[loss=0.2159, ctc_loss=0.1434, cr_loss=0.3624, over 17250.00 frames. ], tot_loss[loss=0.2263, ctc_loss=0.1529, cr_loss=0.3674, over 3349387.58 frames. ], batch size: 42, lr: 7.45e-03, grad_scale: 32.0
2024-09-23 15:22:57,242 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.45 vs. limit=22.5
2024-09-23 15:23:32,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=288026.6666666667, ans=0.1
2024-09-23 15:23:35,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=288026.6666666667, ans=0.1
2024-09-23 15:23:46,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=288073.3333333333, ans=0.0
2024-09-23 15:23:49,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=288073.3333333333, ans=0.125
2024-09-23 15:23:52,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=288073.3333333333, ans=0.125
2024-09-23 15:23:52,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=288073.3333333333, ans=0.125
2024-09-23 15:24:01,923 INFO [train.py:1198] (1/4) Epoch 16, batch 3300, loss[loss=0.2262, ctc_loss=0.1495, cr_loss=0.3837, over 17175.00 frames. ], tot_loss[loss=0.2246, ctc_loss=0.1514, cr_loss=0.3657, over 3361834.22 frames. ], batch size: 45, lr: 7.45e-03, grad_scale: 32.0
2024-09-23 15:24:22,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.59 vs. limit=15.0
2024-09-23 15:24:33,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=288213.3333333333, ans=0.1
2024-09-23 15:24:46,430 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.267e+02 1.362e+02 1.524e+02 2.882e+02, threshold=2.723e+02, percent-clipped=1.0
2024-09-23 15:24:56,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=288260.0, ans=0.0
2024-09-23 15:25:18,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=288306.6666666667, ans=0.0
2024-09-23 15:25:24,524 INFO [train.py:1198] (1/4) Epoch 16, batch 3350, loss[loss=0.2504, ctc_loss=0.168, cr_loss=0.4123, over 16986.00 frames. ], tot_loss[loss=0.2242, ctc_loss=0.1513, cr_loss=0.3649, over 3358716.14 frames. ], batch size: 53, lr: 7.45e-03, grad_scale: 32.0
2024-09-23 15:25:34,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=288353.3333333333, ans=0.2
2024-09-23 15:25:39,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=288353.3333333333, ans=0.0
2024-09-23 15:25:49,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=288400.0, ans=0.125
2024-09-23 15:26:17,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=288493.3333333333, ans=0.0
2024-09-23 15:26:18,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=288493.3333333333, ans=0.125
2024-09-23 15:26:18,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=288493.3333333333, ans=0.07
2024-09-23 15:26:29,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=288540.0, ans=0.1
2024-09-23 15:26:45,093 INFO [train.py:1198] (1/4) Epoch 16, batch 3400, loss[loss=0.2329, ctc_loss=0.1557, cr_loss=0.3857, over 17293.00 frames. ], tot_loss[loss=0.2247, ctc_loss=0.1517, cr_loss=0.3654, over 3360571.60 frames. ], batch size: 51, lr: 7.44e-03, grad_scale: 32.0
2024-09-23 15:26:55,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.22 vs. limit=15.0
2024-09-23 15:26:56,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=288586.6666666667, ans=0.0
2024-09-23 15:27:27,928 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.234e+02 1.362e+02 1.557e+02 3.833e+02, threshold=2.724e+02, percent-clipped=1.0
2024-09-23 15:27:31,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=288680.0, ans=0.1
2024-09-23 15:27:40,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=288726.6666666667, ans=0.035
2024-09-23 15:28:05,748 INFO [train.py:1198] (1/4) Epoch 16, batch 3450, loss[loss=0.2465, ctc_loss=0.1685, cr_loss=0.3902, over 16733.00 frames. ], tot_loss[loss=0.2251, ctc_loss=0.152, cr_loss=0.3655, over 3350702.93 frames. ], batch size: 61, lr: 7.44e-03, grad_scale: 32.0
2024-09-23 15:28:37,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=288913.3333333333, ans=0.125
2024-09-23 15:28:43,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=288913.3333333333, ans=0.1
2024-09-23 15:29:23,287 INFO [train.py:1198] (1/4) Epoch 16, batch 3500, loss[loss=0.1946, ctc_loss=0.1284, cr_loss=0.3308, over 17189.00 frames. ], tot_loss[loss=0.225, ctc_loss=0.1518, cr_loss=0.3659, over 3343779.68 frames. ], batch size: 41, lr: 7.44e-03, grad_scale: 32.0
2024-09-23 15:30:01,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=289146.6666666667, ans=0.0
2024-09-23 15:30:03,925 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.282e+02 1.368e+02 1.525e+02 3.473e+02, threshold=2.736e+02, percent-clipped=1.0
2024-09-23 15:30:05,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=289146.6666666667, ans=0.1
2024-09-23 15:30:07,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=289146.6666666667, ans=0.125
2024-09-23 15:30:18,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=289193.3333333333, ans=0.0
2024-09-23 15:30:29,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=289240.0, ans=0.0
2024-09-23 15:30:41,432 INFO [train.py:1198] (1/4) Epoch 16, batch 3550, loss[loss=0.2071, ctc_loss=0.138, cr_loss=0.3454, over 17302.00 frames. ], tot_loss[loss=0.2246, ctc_loss=0.1516, cr_loss=0.3652, over 3346116.74 frames. ], batch size: 49, lr: 7.43e-03, grad_scale: 32.0
2024-09-23 15:30:51,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=289286.6666666667, ans=0.125
2024-09-23 15:30:54,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=289286.6666666667, ans=0.2
2024-09-23 15:31:09,259 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=11.38 vs. limit=15.0
2024-09-23 15:31:43,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=289473.3333333333, ans=0.125
2024-09-23 15:31:47,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=289473.3333333333, ans=0.125
2024-09-23 15:32:00,001 INFO [train.py:1198] (1/4) Epoch 16, batch 3600, loss[loss=0.185, ctc_loss=0.1253, cr_loss=0.2985, over 17041.00 frames. ], tot_loss[loss=0.2238, ctc_loss=0.151, cr_loss=0.364, over 3349781.99 frames. ], batch size: 39, lr: 7.43e-03, grad_scale: 32.0
2024-09-23 15:32:00,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=289520.0, ans=0.05
2024-09-23 15:32:01,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=289520.0, ans=0.1
2024-09-23 15:32:06,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=289520.0, ans=0.1
2024-09-23 15:32:12,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=289520.0, ans=0.125
2024-09-23 15:32:23,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=289566.6666666667, ans=0.025
2024-09-23 15:32:40,499 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.232e+02 1.305e+02 1.381e+02 1.906e+02, threshold=2.610e+02, percent-clipped=0.0
2024-09-23 15:32:42,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=289613.3333333333, ans=0.2
2024-09-23 15:33:01,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=289706.6666666667, ans=0.125
2024-09-23 15:33:12,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.17 vs. limit=15.0
2024-09-23 15:33:18,231 INFO [train.py:1198] (1/4) Epoch 16, batch 3650, loss[loss=0.2275, ctc_loss=0.1533, cr_loss=0.3714, over 17075.00 frames. ], tot_loss[loss=0.2236, ctc_loss=0.1507, cr_loss=0.3644, over 3351529.70 frames. ], batch size: 46, lr: 7.43e-03, grad_scale: 16.0
2024-09-23 15:33:41,338 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.94 vs. limit=10.0
2024-09-23 15:33:45,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=289800.0, ans=0.125
2024-09-23 15:33:51,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=289846.6666666667, ans=0.0
2024-09-23 15:33:58,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=289846.6666666667, ans=0.025
2024-09-23 15:34:27,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=289940.0, ans=0.1
2024-09-23 15:34:31,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=289940.0, ans=0.125
2024-09-23 15:34:40,767 INFO [train.py:1198] (1/4) Epoch 16, batch 3700, loss[loss=0.1944, ctc_loss=0.1292, cr_loss=0.326, over 16940.00 frames. ], tot_loss[loss=0.2236, ctc_loss=0.1509, cr_loss=0.364, over 3353495.18 frames. ], batch size: 42, lr: 7.43e-03, grad_scale: 16.0
2024-09-23 15:34:58,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=290033.3333333333, ans=0.0
2024-09-23 15:35:23,814 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.296e+02 1.389e+02 1.546e+02 2.430e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-23 15:35:27,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=290126.6666666667, ans=0.125
2024-09-23 15:35:47,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=290173.3333333333, ans=0.025
2024-09-23 15:35:55,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=290173.3333333333, ans=0.04949747468305833
2024-09-23 15:35:59,677 INFO [train.py:1198] (1/4) Epoch 16, batch 3750, loss[loss=0.2137, ctc_loss=0.1473, cr_loss=0.3322, over 17147.00 frames. ], tot_loss[loss=0.2254, ctc_loss=0.1521, cr_loss=0.3665, over 3359447.91 frames. ], batch size: 45, lr: 7.42e-03, grad_scale: 16.0
2024-09-23 15:36:46,191 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.57 vs. limit=15.0
2024-09-23 15:36:47,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=290360.0, ans=0.125
2024-09-23 15:36:54,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=290360.0, ans=0.2
2024-09-23 15:37:09,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=290406.6666666667, ans=0.125
2024-09-23 15:37:15,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=290406.6666666667, ans=0.025
2024-09-23 15:37:18,711 INFO [train.py:1198] (1/4) Epoch 16, batch 3800, loss[loss=0.2694, ctc_loss=0.183, cr_loss=0.4322, over 15206.00 frames. ], tot_loss[loss=0.2255, ctc_loss=0.1523, cr_loss=0.3663, over 3356487.79 frames. ], batch size: 89, lr: 7.42e-03, grad_scale: 16.0
2024-09-23 15:37:24,237 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=11.84 vs. limit=15.0
2024-09-23 15:37:37,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=290500.0, ans=0.1
2024-09-23 15:37:49,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=290546.6666666667, ans=0.07
2024-09-23 15:37:56,343 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.69 vs. limit=15.0
2024-09-23 15:38:00,416 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.261e+02 1.370e+02 1.506e+02 3.479e+02, threshold=2.739e+02, percent-clipped=1.0
2024-09-23 15:38:17,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=290593.3333333333, ans=0.125
2024-09-23 15:38:36,619 INFO [train.py:1198] (1/4) Epoch 16, batch 3850, loss[loss=0.3079, ctc_loss=0.2256, cr_loss=0.4115, over 11786.00 frames. ], tot_loss[loss=0.2281, ctc_loss=0.1545, cr_loss=0.368, over 3298283.40 frames. ], batch size: 124, lr: 7.42e-03, grad_scale: 16.0
2024-09-23 15:38:46,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=290686.6666666667, ans=0.0
2024-09-23 15:40:38,386 INFO [train.py:1198] (1/4) Epoch 17, batch 0, loss[loss=0.2737, ctc_loss=0.1873, cr_loss=0.432, over 17137.00 frames. ], tot_loss[loss=0.2737, ctc_loss=0.1873, cr_loss=0.432, over 17137.00 frames. ], batch size: 48, lr: 7.19e-03, grad_scale: 32.0
2024-09-23 15:40:38,387 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 15:40:46,320 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([4.4996, 4.2327, 4.2788, 3.9933], device='cuda:1')
2024-09-23 15:40:53,766 INFO [train.py:1230] (1/4) Epoch 17, validation: loss=0.04104, ctc_loss=0.04104, cr_loss=7.589e-15, over 944034.00 frames. 
2024-09-23 15:40:53,767 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 15:41:24,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=290948.0, ans=0.05
2024-09-23 15:41:24,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=290948.0, ans=0.125
2024-09-23 15:41:24,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=290948.0, ans=0.125
2024-09-23 15:41:25,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=290948.0, ans=0.2
2024-09-23 15:41:43,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=291041.3333333333, ans=0.025
2024-09-23 15:41:46,231 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.415e+02 1.552e+02 1.651e+02 2.695e+02, threshold=3.103e+02, percent-clipped=0.0
2024-09-23 15:41:48,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=291041.3333333333, ans=0.125
2024-09-23 15:42:18,131 INFO [train.py:1198] (1/4) Epoch 17, batch 50, loss[loss=0.2238, ctc_loss=0.1511, cr_loss=0.3637, over 17215.00 frames. ], tot_loss[loss=0.2261, ctc_loss=0.1527, cr_loss=0.3673, over 757194.18 frames. ], batch size: 50, lr: 7.19e-03, grad_scale: 16.0
2024-09-23 15:42:18,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=291134.6666666667, ans=0.125
2024-09-23 15:42:30,944 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 15:42:34,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=291181.3333333333, ans=0.1
2024-09-23 15:42:40,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=291181.3333333333, ans=0.125
2024-09-23 15:42:55,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=291228.0, ans=0.125
2024-09-23 15:43:30,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=291321.3333333333, ans=0.125
2024-09-23 15:43:30,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=291321.3333333333, ans=0.0
2024-09-23 15:43:39,789 INFO [train.py:1198] (1/4) Epoch 17, batch 100, loss[loss=0.2273, ctc_loss=0.1526, cr_loss=0.3735, over 17303.00 frames. ], tot_loss[loss=0.2248, ctc_loss=0.1518, cr_loss=0.3651, over 1328320.71 frames. ], batch size: 51, lr: 7.18e-03, grad_scale: 16.0
2024-09-23 15:44:08,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=291414.6666666667, ans=0.125
2024-09-23 15:44:31,115 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.243e+02 1.327e+02 1.449e+02 2.389e+02, threshold=2.654e+02, percent-clipped=0.0
2024-09-23 15:44:59,988 INFO [train.py:1198] (1/4) Epoch 17, batch 150, loss[loss=0.2423, ctc_loss=0.1631, cr_loss=0.3963, over 17020.00 frames. ], tot_loss[loss=0.2244, ctc_loss=0.1514, cr_loss=0.3652, over 1778360.24 frames. ], batch size: 51, lr: 7.18e-03, grad_scale: 16.0
2024-09-23 15:45:10,520 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.43 vs. limit=22.5
2024-09-23 15:45:19,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=291648.0, ans=0.125
2024-09-23 15:45:22,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=291648.0, ans=0.0
2024-09-23 15:45:36,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=291694.6666666667, ans=0.1
2024-09-23 15:45:49,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=291741.3333333333, ans=0.0
2024-09-23 15:45:56,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=291741.3333333333, ans=0.95
2024-09-23 15:46:00,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=291741.3333333333, ans=0.0
2024-09-23 15:46:21,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=291788.0, ans=0.0
2024-09-23 15:46:25,712 INFO [train.py:1198] (1/4) Epoch 17, batch 200, loss[loss=0.2246, ctc_loss=0.1491, cr_loss=0.3775, over 17052.00 frames. ], tot_loss[loss=0.226, ctc_loss=0.1527, cr_loss=0.3665, over 2109756.00 frames. ], batch size: 39, lr: 7.18e-03, grad_scale: 16.0
2024-09-23 15:46:45,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=291881.3333333333, ans=0.125
2024-09-23 15:46:57,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.16 vs. limit=6.0
2024-09-23 15:47:18,310 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.306e+02 1.386e+02 1.614e+02 2.877e+02, threshold=2.773e+02, percent-clipped=1.0
2024-09-23 15:47:36,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=292021.3333333333, ans=0.125
2024-09-23 15:47:49,025 INFO [train.py:1198] (1/4) Epoch 17, batch 250, loss[loss=0.2179, ctc_loss=0.1465, cr_loss=0.3567, over 17032.00 frames. ], tot_loss[loss=0.2259, ctc_loss=0.1525, cr_loss=0.3669, over 2395925.79 frames. ], batch size: 44, lr: 7.18e-03, grad_scale: 16.0
2024-09-23 15:47:58,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=292068.0, ans=0.025
2024-09-23 15:47:58,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=292068.0, ans=0.125
2024-09-23 15:48:11,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=292114.6666666667, ans=0.125
2024-09-23 15:48:57,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=292254.6666666667, ans=0.125
2024-09-23 15:49:08,855 INFO [train.py:1198] (1/4) Epoch 17, batch 300, loss[loss=0.2211, ctc_loss=0.1488, cr_loss=0.3616, over 17204.00 frames. ], tot_loss[loss=0.2265, ctc_loss=0.153, cr_loss=0.3672, over 2595999.42 frames. ], batch size: 47, lr: 7.17e-03, grad_scale: 16.0
2024-09-23 15:49:45,240 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.54 vs. limit=15.0
2024-09-23 15:49:49,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=292394.6666666667, ans=0.025
2024-09-23 15:49:59,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=292441.3333333333, ans=0.125
2024-09-23 15:49:59,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=292441.3333333333, ans=0.0
2024-09-23 15:50:00,587 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.265e+02 1.346e+02 1.452e+02 2.269e+02, threshold=2.693e+02, percent-clipped=0.0
2024-09-23 15:50:07,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=292441.3333333333, ans=0.05
2024-09-23 15:50:07,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=292441.3333333333, ans=0.125
2024-09-23 15:50:18,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=292488.0, ans=0.125
2024-09-23 15:50:32,873 INFO [train.py:1198] (1/4) Epoch 17, batch 350, loss[loss=0.2272, ctc_loss=0.1503, cr_loss=0.3844, over 16751.00 frames. ], tot_loss[loss=0.2259, ctc_loss=0.1525, cr_loss=0.3671, over 2758095.08 frames. ], batch size: 61, lr: 7.17e-03, grad_scale: 16.0
2024-09-23 15:50:42,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=292534.6666666667, ans=0.125
2024-09-23 15:50:52,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=292581.3333333333, ans=0.95
2024-09-23 15:51:23,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=292674.6666666667, ans=0.125
2024-09-23 15:51:57,138 INFO [train.py:1198] (1/4) Epoch 17, batch 400, loss[loss=0.1935, ctc_loss=0.1254, cr_loss=0.3403, over 17273.00 frames. ], tot_loss[loss=0.2265, ctc_loss=0.1528, cr_loss=0.3685, over 2889891.26 frames. ], batch size: 44, lr: 7.17e-03, grad_scale: 32.0
2024-09-23 15:52:10,182 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 15:52:25,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=292814.6666666667, ans=10.0
2024-09-23 15:52:26,793 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=7.01 vs. limit=15.0
2024-09-23 15:52:35,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=292861.3333333333, ans=0.0
2024-09-23 15:52:41,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=292861.3333333333, ans=0.05
2024-09-23 15:52:50,106 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.274e+02 1.356e+02 1.494e+02 2.535e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-23 15:53:08,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=292954.6666666667, ans=0.125
2024-09-23 15:53:18,848 INFO [train.py:1198] (1/4) Epoch 17, batch 450, loss[loss=0.2374, ctc_loss=0.1647, cr_loss=0.3633, over 17145.00 frames. ], tot_loss[loss=0.2257, ctc_loss=0.1524, cr_loss=0.3667, over 2984771.36 frames. ], batch size: 48, lr: 7.16e-03, grad_scale: 32.0
2024-09-23 15:53:33,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=293048.0, ans=0.0
2024-09-23 15:53:37,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.74 vs. limit=6.0
2024-09-23 15:53:48,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=293048.0, ans=0.125
2024-09-23 15:53:49,978 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=293094.6666666667, ans=0.125
2024-09-23 15:54:21,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=293188.0, ans=0.125
2024-09-23 15:54:39,067 INFO [train.py:1198] (1/4) Epoch 17, batch 500, loss[loss=0.1961, ctc_loss=0.1287, cr_loss=0.337, over 16695.00 frames. ], tot_loss[loss=0.2252, ctc_loss=0.1518, cr_loss=0.3671, over 3070441.32 frames. ], batch size: 37, lr: 7.16e-03, grad_scale: 32.0
2024-09-23 15:54:53,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=293281.3333333333, ans=0.0
2024-09-23 15:55:13,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=293328.0, ans=0.125
2024-09-23 15:55:16,790 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.25 vs. limit=15.0
2024-09-23 15:55:32,746 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.054e+02 1.258e+02 1.318e+02 1.426e+02 2.177e+02, threshold=2.636e+02, percent-clipped=0.0
2024-09-23 15:55:44,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=293421.3333333333, ans=0.125
2024-09-23 15:56:03,960 INFO [train.py:1198] (1/4) Epoch 17, batch 550, loss[loss=0.2519, ctc_loss=0.1691, cr_loss=0.4144, over 17098.00 frames. ], tot_loss[loss=0.225, ctc_loss=0.1516, cr_loss=0.3674, over 3137165.24 frames. ], batch size: 49, lr: 7.16e-03, grad_scale: 32.0
2024-09-23 15:56:04,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=293468.0, ans=0.125
2024-09-23 15:56:07,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=293468.0, ans=0.125
2024-09-23 15:56:31,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=293514.6666666667, ans=0.125
2024-09-23 15:56:35,193 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.05 vs. limit=12.0
2024-09-23 15:56:53,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=293608.0, ans=0.0
2024-09-23 15:56:55,577 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.20 vs. limit=15.0
2024-09-23 15:57:01,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=293608.0, ans=0.1
2024-09-23 15:57:22,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=293654.6666666667, ans=0.125
2024-09-23 15:57:27,122 INFO [train.py:1198] (1/4) Epoch 17, batch 600, loss[loss=0.2666, ctc_loss=0.1851, cr_loss=0.4077, over 16805.00 frames. ], tot_loss[loss=0.224, ctc_loss=0.1508, cr_loss=0.366, over 3192927.67 frames. ], batch size: 61, lr: 7.16e-03, grad_scale: 32.0
2024-09-23 15:57:27,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=293701.3333333333, ans=0.1
2024-09-23 15:57:32,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=293701.3333333333, ans=0.125
2024-09-23 15:57:53,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=293748.0, ans=0.0
2024-09-23 15:58:16,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=293841.3333333333, ans=0.1
2024-09-23 15:58:20,482 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.285e+02 1.384e+02 1.494e+02 2.358e+02, threshold=2.768e+02, percent-clipped=0.0
2024-09-23 15:58:30,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=293841.3333333333, ans=0.125
2024-09-23 15:58:38,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=293888.0, ans=0.125
2024-09-23 15:58:40,998 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.48 vs. limit=15.0
2024-09-23 15:58:41,918 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 15:58:45,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=293888.0, ans=0.125
2024-09-23 15:58:49,413 INFO [train.py:1198] (1/4) Epoch 17, batch 650, loss[loss=0.2008, ctc_loss=0.1335, cr_loss=0.3368, over 17294.00 frames. ], tot_loss[loss=0.2237, ctc_loss=0.1506, cr_loss=0.3654, over 3237145.37 frames. ], batch size: 46, lr: 7.15e-03, grad_scale: 32.0
2024-09-23 15:59:31,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=294028.0, ans=0.125
2024-09-23 15:59:34,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=294028.0, ans=0.025
2024-09-23 16:00:02,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=294121.3333333333, ans=0.025
2024-09-23 16:00:07,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=294121.3333333333, ans=0.125
2024-09-23 16:00:09,840 INFO [train.py:1198] (1/4) Epoch 17, batch 700, loss[loss=0.2288, ctc_loss=0.1558, cr_loss=0.3652, over 17227.00 frames. ], tot_loss[loss=0.2232, ctc_loss=0.1502, cr_loss=0.3651, over 3268560.47 frames. ], batch size: 50, lr: 7.15e-03, grad_scale: 32.0
2024-09-23 16:00:13,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=294168.0, ans=0.125
2024-09-23 16:00:25,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=294168.0, ans=0.125
2024-09-23 16:00:28,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=294214.6666666667, ans=0.2
2024-09-23 16:00:41,553 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 16:01:06,301 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.243e+02 1.328e+02 1.463e+02 2.107e+02, threshold=2.656e+02, percent-clipped=0.0
2024-09-23 16:01:06,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=294308.0, ans=0.0
2024-09-23 16:01:09,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=294308.0, ans=0.125
2024-09-23 16:01:22,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=294354.6666666667, ans=0.125
2024-09-23 16:01:27,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=294354.6666666667, ans=0.0
2024-09-23 16:01:34,888 INFO [train.py:1198] (1/4) Epoch 17, batch 750, loss[loss=0.2343, ctc_loss=0.1597, cr_loss=0.3727, over 17031.00 frames. ], tot_loss[loss=0.2242, ctc_loss=0.1511, cr_loss=0.3657, over 3275790.99 frames. ], batch size: 52, lr: 7.15e-03, grad_scale: 32.0
2024-09-23 16:01:36,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=294401.3333333333, ans=0.05
2024-09-23 16:01:42,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=294401.3333333333, ans=0.0
2024-09-23 16:01:46,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=294401.3333333333, ans=0.125
2024-09-23 16:01:55,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=294448.0, ans=0.04949747468305833
2024-09-23 16:01:59,121 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.66 vs. limit=12.0
2024-09-23 16:02:02,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=294448.0, ans=0.125
2024-09-23 16:02:06,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=294448.0, ans=0.125
2024-09-23 16:02:25,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=294541.3333333333, ans=0.0
2024-09-23 16:02:29,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=294541.3333333333, ans=0.125
2024-09-23 16:02:49,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=294588.0, ans=0.1
2024-09-23 16:03:00,185 INFO [train.py:1198] (1/4) Epoch 17, batch 800, loss[loss=0.2238, ctc_loss=0.1505, cr_loss=0.3667, over 17339.00 frames. ], tot_loss[loss=0.2241, ctc_loss=0.1509, cr_loss=0.3657, over 3298857.87 frames. ], batch size: 48, lr: 7.14e-03, grad_scale: 32.0
2024-09-23 16:03:00,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=294634.6666666667, ans=0.125
2024-09-23 16:03:10,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=294634.6666666667, ans=0.1
2024-09-23 16:03:34,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=294728.0, ans=0.125
2024-09-23 16:03:45,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.51 vs. limit=15.0
2024-09-23 16:03:51,161 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.255e+02 1.363e+02 1.453e+02 2.011e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-23 16:04:00,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=294774.6666666667, ans=0.125
2024-09-23 16:04:00,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=294774.6666666667, ans=0.125
2024-09-23 16:04:19,634 INFO [train.py:1198] (1/4) Epoch 17, batch 850, loss[loss=0.2269, ctc_loss=0.152, cr_loss=0.3743, over 17025.00 frames. ], tot_loss[loss=0.2247, ctc_loss=0.1513, cr_loss=0.367, over 3315081.10 frames. ], batch size: 51, lr: 7.14e-03, grad_scale: 32.0
2024-09-23 16:04:26,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=294868.0, ans=0.125
2024-09-23 16:04:44,449 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.43 vs. limit=15.0
2024-09-23 16:04:55,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=294961.3333333333, ans=0.0
2024-09-23 16:05:25,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.18 vs. limit=6.0
2024-09-23 16:05:27,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=295054.6666666667, ans=0.015
2024-09-23 16:05:34,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=295054.6666666667, ans=0.04949747468305833
2024-09-23 16:05:37,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=295054.6666666667, ans=0.07
2024-09-23 16:05:41,832 INFO [train.py:1198] (1/4) Epoch 17, batch 900, loss[loss=0.2487, ctc_loss=0.1671, cr_loss=0.4078, over 16999.00 frames. ], tot_loss[loss=0.2248, ctc_loss=0.1514, cr_loss=0.367, over 3325012.25 frames. ], batch size: 53, lr: 7.14e-03, grad_scale: 32.0
2024-09-23 16:05:49,014 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.38 vs. limit=15.0
2024-09-23 16:06:07,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=295148.0, ans=0.125
2024-09-23 16:06:35,667 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.252e+02 1.345e+02 1.506e+02 2.061e+02, threshold=2.689e+02, percent-clipped=0.0
2024-09-23 16:06:46,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=295241.3333333333, ans=0.125
2024-09-23 16:06:48,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=295241.3333333333, ans=0.0
2024-09-23 16:07:07,053 INFO [train.py:1198] (1/4) Epoch 17, batch 950, loss[loss=0.2234, ctc_loss=0.1514, cr_loss=0.3597, over 16765.00 frames. ], tot_loss[loss=0.2251, ctc_loss=0.1516, cr_loss=0.3673, over 3329273.59 frames. ], batch size: 61, lr: 7.14e-03, grad_scale: 32.0
2024-09-23 16:07:12,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=295334.6666666667, ans=0.0
2024-09-23 16:07:21,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=295381.3333333333, ans=0.125
2024-09-23 16:07:32,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=295381.3333333333, ans=0.0
2024-09-23 16:07:46,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=295428.0, ans=0.125
2024-09-23 16:08:27,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=295568.0, ans=10.0
2024-09-23 16:08:29,164 INFO [train.py:1198] (1/4) Epoch 17, batch 1000, loss[loss=0.247, ctc_loss=0.1684, cr_loss=0.3928, over 17023.00 frames. ], tot_loss[loss=0.2247, ctc_loss=0.1513, cr_loss=0.3672, over 3333267.94 frames. ], batch size: 52, lr: 7.13e-03, grad_scale: 32.0
2024-09-23 16:08:40,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.47 vs. limit=15.0
2024-09-23 16:08:59,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=295661.3333333333, ans=0.1
2024-09-23 16:09:19,594 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.278e+02 1.364e+02 1.541e+02 2.196e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-23 16:09:37,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=295754.6666666667, ans=0.125
2024-09-23 16:09:47,780 INFO [train.py:1198] (1/4) Epoch 17, batch 1050, loss[loss=0.2499, ctc_loss=0.1649, cr_loss=0.4248, over 17299.00 frames. ], tot_loss[loss=0.2255, ctc_loss=0.1519, cr_loss=0.368, over 3337940.36 frames. ], batch size: 49, lr: 7.13e-03, grad_scale: 32.0
2024-09-23 16:10:00,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=295801.3333333333, ans=0.0
2024-09-23 16:10:40,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=295941.3333333333, ans=0.0
2024-09-23 16:10:40,726 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.04 vs. limit=22.5
2024-09-23 16:11:11,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.34 vs. limit=15.0
2024-09-23 16:11:12,826 INFO [train.py:1198] (1/4) Epoch 17, batch 1100, loss[loss=0.2002, ctc_loss=0.1343, cr_loss=0.3294, over 17296.00 frames. ], tot_loss[loss=0.2243, ctc_loss=0.151, cr_loss=0.3663, over 3339870.90 frames. ], batch size: 49, lr: 7.13e-03, grad_scale: 32.0
2024-09-23 16:11:54,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.32 vs. limit=12.0
2024-09-23 16:11:55,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=296128.0, ans=0.125
2024-09-23 16:12:06,482 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.273e+02 1.407e+02 1.625e+02 2.289e+02, threshold=2.813e+02, percent-clipped=0.0
2024-09-23 16:12:19,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=296221.3333333333, ans=0.125
2024-09-23 16:12:35,106 INFO [train.py:1198] (1/4) Epoch 17, batch 1150, loss[loss=0.1936, ctc_loss=0.127, cr_loss=0.333, over 17031.00 frames. ], tot_loss[loss=0.2243, ctc_loss=0.151, cr_loss=0.3666, over 3346776.68 frames. ], batch size: 39, lr: 7.13e-03, grad_scale: 32.0
2024-09-23 16:12:38,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=296268.0, ans=0.2
2024-09-23 16:12:44,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=296268.0, ans=0.125
2024-09-23 16:13:05,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=296314.6666666667, ans=0.125
2024-09-23 16:13:56,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=296501.3333333333, ans=0.125
2024-09-23 16:13:57,590 INFO [train.py:1198] (1/4) Epoch 17, batch 1200, loss[loss=0.2565, ctc_loss=0.1738, cr_loss=0.4134, over 16604.00 frames. ], tot_loss[loss=0.2236, ctc_loss=0.1504, cr_loss=0.3659, over 3350539.21 frames. ], batch size: 66, lr: 7.12e-03, grad_scale: 32.0
2024-09-23 16:14:05,855 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.47 vs. limit=15.0
2024-09-23 16:14:21,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=296548.0, ans=0.125
2024-09-23 16:14:34,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=296594.6666666667, ans=0.125
2024-09-23 16:14:38,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=296594.6666666667, ans=0.125
2024-09-23 16:14:43,848 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 16:14:49,695 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.250e+02 1.332e+02 1.452e+02 2.634e+02, threshold=2.664e+02, percent-clipped=0.0
2024-09-23 16:14:50,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=296641.3333333333, ans=0.2
2024-09-23 16:15:09,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=296688.0, ans=0.0
2024-09-23 16:15:19,279 INFO [train.py:1198] (1/4) Epoch 17, batch 1250, loss[loss=0.2047, ctc_loss=0.1357, cr_loss=0.345, over 17299.00 frames. ], tot_loss[loss=0.2225, ctc_loss=0.1495, cr_loss=0.365, over 3359158.98 frames. ], batch size: 51, lr: 7.12e-03, grad_scale: 32.0
2024-09-23 16:16:16,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=296874.6666666667, ans=0.1
2024-09-23 16:16:16,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=296874.6666666667, ans=0.0
2024-09-23 16:16:21,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=296874.6666666667, ans=0.125
2024-09-23 16:16:44,491 INFO [train.py:1198] (1/4) Epoch 17, batch 1300, loss[loss=0.1923, ctc_loss=0.1262, cr_loss=0.3308, over 17088.00 frames. ], tot_loss[loss=0.222, ctc_loss=0.1492, cr_loss=0.3639, over 3350108.35 frames. ], batch size: 43, lr: 7.12e-03, grad_scale: 32.0
2024-09-23 16:16:52,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=296968.0, ans=0.5
2024-09-23 16:17:03,025 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.46 vs. limit=15.0
2024-09-23 16:17:19,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=297061.3333333333, ans=0.125
2024-09-23 16:17:20,129 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.14 vs. limit=22.5
2024-09-23 16:17:36,805 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.242e+02 1.322e+02 1.445e+02 3.373e+02, threshold=2.644e+02, percent-clipped=1.0
2024-09-23 16:17:37,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=297108.0, ans=0.0
2024-09-23 16:17:38,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=297108.0, ans=0.125
2024-09-23 16:17:52,474 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=297154.6666666667, ans=0.0
2024-09-23 16:18:06,582 INFO [train.py:1198] (1/4) Epoch 17, batch 1350, loss[loss=0.2088, ctc_loss=0.1397, cr_loss=0.3458, over 17033.00 frames. ], tot_loss[loss=0.2223, ctc_loss=0.1495, cr_loss=0.3641, over 3347223.74 frames. ], batch size: 39, lr: 7.11e-03, grad_scale: 32.0
2024-09-23 16:18:13,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=297201.3333333333, ans=0.2
2024-09-23 16:18:18,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=297201.3333333333, ans=0.0
2024-09-23 16:18:48,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=297294.6666666667, ans=0.0
2024-09-23 16:19:05,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=297341.3333333333, ans=0.125
2024-09-23 16:19:08,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=297388.0, ans=0.025
2024-09-23 16:19:26,088 INFO [train.py:1198] (1/4) Epoch 17, batch 1400, loss[loss=0.2459, ctc_loss=0.1663, cr_loss=0.3977, over 17308.00 frames. ], tot_loss[loss=0.2222, ctc_loss=0.1493, cr_loss=0.3644, over 3357238.68 frames. ], batch size: 49, lr: 7.11e-03, grad_scale: 32.0
2024-09-23 16:19:32,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=297434.6666666667, ans=0.5
2024-09-23 16:19:40,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=297481.3333333333, ans=0.0
2024-09-23 16:19:48,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=297481.3333333333, ans=0.0
2024-09-23 16:19:50,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=297481.3333333333, ans=0.0
2024-09-23 16:19:59,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=297528.0, ans=0.025
2024-09-23 16:20:03,510 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.37 vs. limit=12.0
2024-09-23 16:20:13,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=297528.0, ans=0.0
2024-09-23 16:20:20,766 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.246e+02 1.345e+02 1.562e+02 2.130e+02, threshold=2.691e+02, percent-clipped=0.0
2024-09-23 16:20:29,757 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.90 vs. limit=15.0
2024-09-23 16:20:40,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=297621.3333333333, ans=0.0
2024-09-23 16:20:50,352 INFO [train.py:1198] (1/4) Epoch 17, batch 1450, loss[loss=0.198, ctc_loss=0.1276, cr_loss=0.352, over 16743.00 frames. ], tot_loss[loss=0.2228, ctc_loss=0.1496, cr_loss=0.3657, over 3361805.99 frames. ], batch size: 37, lr: 7.11e-03, grad_scale: 32.0
2024-09-23 16:21:08,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=297714.6666666667, ans=0.125
2024-09-23 16:21:38,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=297761.3333333333, ans=0.1
2024-09-23 16:22:04,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=297854.6666666667, ans=0.125
2024-09-23 16:22:12,530 INFO [train.py:1198] (1/4) Epoch 17, batch 1500, loss[loss=0.2298, ctc_loss=0.1518, cr_loss=0.3904, over 17090.00 frames. ], tot_loss[loss=0.2226, ctc_loss=0.1496, cr_loss=0.3655, over 3365038.33 frames. ], batch size: 49, lr: 7.11e-03, grad_scale: 32.0
2024-09-23 16:22:12,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=297901.3333333333, ans=0.125
2024-09-23 16:22:28,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=297948.0, ans=0.125
2024-09-23 16:22:43,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=297948.0, ans=0.1
2024-09-23 16:23:07,241 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.004e+02 1.245e+02 1.341e+02 1.437e+02 3.249e+02, threshold=2.682e+02, percent-clipped=1.0
2024-09-23 16:23:10,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=298041.3333333333, ans=0.0
2024-09-23 16:23:12,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=298041.3333333333, ans=0.125
2024-09-23 16:23:24,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=298088.0, ans=0.125
2024-09-23 16:23:34,452 INFO [train.py:1198] (1/4) Epoch 17, batch 1550, loss[loss=0.2046, ctc_loss=0.1357, cr_loss=0.3448, over 17062.00 frames. ], tot_loss[loss=0.2228, ctc_loss=0.1497, cr_loss=0.3655, over 3359674.01 frames. ], batch size: 39, lr: 7.10e-03, grad_scale: 32.0
2024-09-23 16:24:11,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=298228.0, ans=0.125
2024-09-23 16:24:49,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=298321.3333333333, ans=0.2
2024-09-23 16:24:51,759 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.96 vs. limit=10.0
2024-09-23 16:24:54,218 INFO [train.py:1198] (1/4) Epoch 17, batch 1600, loss[loss=0.1831, ctc_loss=0.1216, cr_loss=0.3077, over 17273.00 frames. ], tot_loss[loss=0.2222, ctc_loss=0.1492, cr_loss=0.3651, over 3364271.18 frames. ], batch size: 42, lr: 7.10e-03, grad_scale: 32.0
2024-09-23 16:25:28,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=298461.3333333333, ans=0.0
2024-09-23 16:25:43,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=298461.3333333333, ans=0.1
2024-09-23 16:25:51,724 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.268e+02 1.375e+02 1.538e+02 2.240e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-23 16:26:03,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.88 vs. limit=15.0
2024-09-23 16:26:09,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=298554.6666666667, ans=0.025
2024-09-23 16:26:18,870 INFO [train.py:1198] (1/4) Epoch 17, batch 1650, loss[loss=0.2185, ctc_loss=0.1422, cr_loss=0.3815, over 17251.00 frames. ], tot_loss[loss=0.2215, ctc_loss=0.1486, cr_loss=0.3642, over 3366027.63 frames. ], batch size: 42, lr: 7.10e-03, grad_scale: 32.0
2024-09-23 16:26:22,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_abs, batch_count=298601.3333333333, ans=0.5
2024-09-23 16:27:07,591 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.29 vs. limit=10.0
2024-09-23 16:27:15,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=298741.3333333333, ans=0.125
2024-09-23 16:27:27,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=298788.0, ans=0.125
2024-09-23 16:27:45,762 INFO [train.py:1198] (1/4) Epoch 17, batch 1700, loss[loss=0.275, ctc_loss=0.1966, cr_loss=0.392, over 11975.00 frames. ], tot_loss[loss=0.2216, ctc_loss=0.1488, cr_loss=0.3641, over 3357676.50 frames. ], batch size: 123, lr: 7.09e-03, grad_scale: 32.0
2024-09-23 16:28:19,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=298928.0, ans=0.125
2024-09-23 16:28:23,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer_ff2.min_abs, batch_count=298928.0, ans=0.1
2024-09-23 16:28:38,521 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.237e+02 1.323e+02 1.443e+02 1.876e+02, threshold=2.646e+02, percent-clipped=0.0
2024-09-23 16:28:41,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=298974.6666666667, ans=0.025
2024-09-23 16:28:51,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=299021.3333333333, ans=0.1
2024-09-23 16:28:54,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=299021.3333333333, ans=0.125
2024-09-23 16:29:05,468 INFO [train.py:1198] (1/4) Epoch 17, batch 1750, loss[loss=0.2406, ctc_loss=0.1648, cr_loss=0.3791, over 16996.00 frames. ], tot_loss[loss=0.2213, ctc_loss=0.1485, cr_loss=0.3641, over 3361025.89 frames. ], batch size: 53, lr: 7.09e-03, grad_scale: 32.0
2024-09-23 16:29:13,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=299068.0, ans=0.1
2024-09-23 16:29:13,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=299068.0, ans=0.125
2024-09-23 16:29:18,721 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.83 vs. limit=22.5
2024-09-23 16:29:43,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=299161.3333333333, ans=0.1
2024-09-23 16:29:47,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=299161.3333333333, ans=0.0
2024-09-23 16:29:50,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=299161.3333333333, ans=0.125
2024-09-23 16:30:08,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.57 vs. limit=10.0
2024-09-23 16:30:27,808 INFO [train.py:1198] (1/4) Epoch 17, batch 1800, loss[loss=0.2149, ctc_loss=0.1487, cr_loss=0.3312, over 17284.00 frames. ], tot_loss[loss=0.2213, ctc_loss=0.1485, cr_loss=0.3636, over 3364905.47 frames. ], batch size: 51, lr: 7.09e-03, grad_scale: 32.0
2024-09-23 16:30:38,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=299301.3333333333, ans=0.125
2024-09-23 16:30:56,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=299348.0, ans=0.1
2024-09-23 16:30:57,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=299348.0, ans=0.0
2024-09-23 16:31:23,056 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.259e+02 1.337e+02 1.488e+02 2.205e+02, threshold=2.673e+02, percent-clipped=0.0
2024-09-23 16:31:23,656 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.33 vs. limit=15.0
2024-09-23 16:31:40,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=299488.0, ans=0.1
2024-09-23 16:31:50,235 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.99 vs. limit=12.0
2024-09-23 16:31:52,512 INFO [train.py:1198] (1/4) Epoch 17, batch 1850, loss[loss=0.2363, ctc_loss=0.1602, cr_loss=0.3803, over 17355.00 frames. ], tot_loss[loss=0.2218, ctc_loss=0.1489, cr_loss=0.3645, over 3365762.35 frames. ], batch size: 52, lr: 7.09e-03, grad_scale: 32.0
2024-09-23 16:32:48,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=299674.6666666667, ans=0.0
2024-09-23 16:33:14,613 INFO [train.py:1198] (1/4) Epoch 17, batch 1900, loss[loss=0.1814, ctc_loss=0.1186, cr_loss=0.3143, over 17167.00 frames. ], tot_loss[loss=0.221, ctc_loss=0.1483, cr_loss=0.3634, over 3365870.53 frames. ], batch size: 41, lr: 7.08e-03, grad_scale: 32.0
2024-09-23 16:33:16,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=299768.0, ans=0.07
2024-09-23 16:33:35,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=299814.6666666667, ans=0.125
2024-09-23 16:33:37,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=299814.6666666667, ans=0.125
2024-09-23 16:34:02,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=299908.0, ans=0.125
2024-09-23 16:34:06,708 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.256e+02 1.309e+02 1.429e+02 1.873e+02, threshold=2.618e+02, percent-clipped=0.0
2024-09-23 16:34:15,344 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.48 vs. limit=15.0
2024-09-23 16:34:33,599 INFO [train.py:1198] (1/4) Epoch 17, batch 1950, loss[loss=0.2194, ctc_loss=0.1484, cr_loss=0.3549, over 17222.00 frames. ], tot_loss[loss=0.2231, ctc_loss=0.1499, cr_loss=0.3663, over 3361702.90 frames. ], batch size: 55, lr: 7.08e-03, grad_scale: 16.0
2024-09-23 16:34:33,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=300001.3333333333, ans=0.0
2024-09-23 16:35:02,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=300048.0, ans=0.0
2024-09-23 16:35:38,659 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.25 vs. limit=10.0
2024-09-23 16:35:55,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.min_abs, batch_count=300188.0, ans=0.5
2024-09-23 16:35:58,734 INFO [train.py:1198] (1/4) Epoch 17, batch 2000, loss[loss=0.1848, ctc_loss=0.1207, cr_loss=0.3202, over 16951.00 frames. ], tot_loss[loss=0.2228, ctc_loss=0.1498, cr_loss=0.3652, over 3350778.55 frames. ], batch size: 42, lr: 7.08e-03, grad_scale: 32.0
2024-09-23 16:36:27,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=300281.3333333333, ans=0.0
2024-09-23 16:36:39,298 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.98 vs. limit=22.5
2024-09-23 16:36:45,210 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=15.15 vs. limit=15.0
2024-09-23 16:36:45,297 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.28 vs. limit=6.0
2024-09-23 16:36:48,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=300374.6666666667, ans=0.0
2024-09-23 16:36:49,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer_na.min_abs, batch_count=300374.6666666667, ans=0.02
2024-09-23 16:36:52,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=300374.6666666667, ans=0.1
2024-09-23 16:36:55,626 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.280e+02 1.363e+02 1.514e+02 2.601e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-23 16:37:21,169 INFO [train.py:1198] (1/4) Epoch 17, batch 2050, loss[loss=0.237, ctc_loss=0.1587, cr_loss=0.3915, over 17198.00 frames. ], tot_loss[loss=0.2228, ctc_loss=0.1498, cr_loss=0.3651, over 3355139.57 frames. ], batch size: 55, lr: 7.08e-03, grad_scale: 32.0
2024-09-23 16:38:12,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=300608.0, ans=0.025
2024-09-23 16:38:17,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=300608.0, ans=0.0
2024-09-23 16:38:19,736 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.31 vs. limit=15.0
2024-09-23 16:38:43,058 INFO [train.py:1198] (1/4) Epoch 17, batch 2100, loss[loss=0.2059, ctc_loss=0.136, cr_loss=0.3492, over 17052.00 frames. ], tot_loss[loss=0.2232, ctc_loss=0.15, cr_loss=0.366, over 3361633.59 frames. ], batch size: 39, lr: 7.07e-03, grad_scale: 32.0
2024-09-23 16:39:15,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=300794.6666666667, ans=0.0
2024-09-23 16:39:37,336 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.012e+02 1.282e+02 1.378e+02 1.629e+02 2.500e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-23 16:39:53,686 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=300888.0, ans=0.05
2024-09-23 16:39:54,244 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.45 vs. limit=22.5
2024-09-23 16:40:05,318 INFO [train.py:1198] (1/4) Epoch 17, batch 2150, loss[loss=0.2184, ctc_loss=0.1437, cr_loss=0.3734, over 17305.00 frames. ], tot_loss[loss=0.2235, ctc_loss=0.1502, cr_loss=0.3668, over 3351695.04 frames. ], batch size: 51, lr: 7.07e-03, grad_scale: 32.0
2024-09-23 16:40:07,632 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.18 vs. limit=15.0
2024-09-23 16:40:33,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=300981.3333333333, ans=0.0
2024-09-23 16:41:03,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=301074.6666666667, ans=0.1
2024-09-23 16:41:13,497 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.50 vs. limit=15.0
2024-09-23 16:41:21,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=301121.3333333333, ans=0.2
2024-09-23 16:41:29,568 INFO [train.py:1198] (1/4) Epoch 17, batch 2200, loss[loss=0.2416, ctc_loss=0.1621, cr_loss=0.3975, over 17256.00 frames. ], tot_loss[loss=0.2243, ctc_loss=0.1508, cr_loss=0.3674, over 3362953.94 frames. ], batch size: 55, lr: 7.07e-03, grad_scale: 32.0
2024-09-23 16:41:59,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=301261.3333333333, ans=0.125
2024-09-23 16:42:12,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=301261.3333333333, ans=0.125
2024-09-23 16:42:23,223 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.227e+02 1.315e+02 1.424e+02 2.310e+02, threshold=2.629e+02, percent-clipped=0.0
2024-09-23 16:42:50,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=301401.3333333333, ans=0.125
2024-09-23 16:42:51,584 INFO [train.py:1198] (1/4) Epoch 17, batch 2250, loss[loss=0.1989, ctc_loss=0.1321, cr_loss=0.334, over 17173.00 frames. ], tot_loss[loss=0.2235, ctc_loss=0.1502, cr_loss=0.3662, over 3372026.15 frames. ], batch size: 41, lr: 7.07e-03, grad_scale: 32.0
2024-09-23 16:43:03,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=301401.3333333333, ans=0.0
2024-09-23 16:43:07,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=301448.0, ans=0.125
2024-09-23 16:43:15,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=301448.0, ans=0.0
2024-09-23 16:43:16,361 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.25 vs. limit=15.0
2024-09-23 16:43:22,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=301494.6666666667, ans=0.125
2024-09-23 16:43:28,657 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.35 vs. limit=15.0
2024-09-23 16:43:31,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=301494.6666666667, ans=0.025
2024-09-23 16:43:41,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=301541.3333333333, ans=0.125
2024-09-23 16:44:00,978 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.27 vs. limit=15.0
2024-09-23 16:44:11,220 INFO [train.py:1198] (1/4) Epoch 17, batch 2300, loss[loss=0.2434, ctc_loss=0.1663, cr_loss=0.3854, over 16500.00 frames. ], tot_loss[loss=0.2232, ctc_loss=0.15, cr_loss=0.3661, over 3369897.38 frames. ], batch size: 66, lr: 7.06e-03, grad_scale: 32.0
2024-09-23 16:44:25,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=301681.3333333333, ans=0.125
2024-09-23 16:44:25,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=301681.3333333333, ans=0.125
2024-09-23 16:44:41,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=301728.0, ans=0.125
2024-09-23 16:45:08,137 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.278e+02 1.376e+02 1.551e+02 2.468e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-23 16:45:23,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=301821.3333333333, ans=0.035
2024-09-23 16:45:35,977 INFO [train.py:1198] (1/4) Epoch 17, batch 2350, loss[loss=0.2019, ctc_loss=0.1305, cr_loss=0.3569, over 17269.00 frames. ], tot_loss[loss=0.2229, ctc_loss=0.1498, cr_loss=0.3656, over 3372148.83 frames. ], batch size: 42, lr: 7.06e-03, grad_scale: 32.0
2024-09-23 16:45:37,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=301868.0, ans=0.125
2024-09-23 16:46:04,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=301914.6666666667, ans=0.2
2024-09-23 16:46:06,747 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=6.16 vs. limit=15.0
2024-09-23 16:46:18,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=301961.3333333333, ans=0.125
2024-09-23 16:46:43,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=302054.6666666667, ans=0.125
2024-09-23 16:46:51,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=302054.6666666667, ans=0.125
2024-09-23 16:46:57,671 INFO [train.py:1198] (1/4) Epoch 17, batch 2400, loss[loss=0.2083, ctc_loss=0.1411, cr_loss=0.336, over 17093.00 frames. ], tot_loss[loss=0.223, ctc_loss=0.1498, cr_loss=0.3659, over 3370622.65 frames. ], batch size: 49, lr: 7.06e-03, grad_scale: 32.0
2024-09-23 16:47:09,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=302101.3333333333, ans=0.1
2024-09-23 16:47:54,481 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.300e+02 1.437e+02 1.590e+02 2.245e+02, threshold=2.874e+02, percent-clipped=0.0
2024-09-23 16:48:14,241 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.15 vs. limit=15.0
2024-09-23 16:48:19,880 INFO [train.py:1198] (1/4) Epoch 17, batch 2450, loss[loss=0.2347, ctc_loss=0.1579, cr_loss=0.384, over 17005.00 frames. ], tot_loss[loss=0.2231, ctc_loss=0.1499, cr_loss=0.3656, over 3370426.58 frames. ], batch size: 56, lr: 7.05e-03, grad_scale: 32.0
2024-09-23 16:48:20,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=302334.6666666667, ans=0.125
2024-09-23 16:48:44,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=302381.3333333333, ans=0.1
2024-09-23 16:49:20,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=302474.6666666667, ans=0.125
2024-09-23 16:49:27,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=302521.3333333333, ans=0.125
2024-09-23 16:49:39,967 INFO [train.py:1198] (1/4) Epoch 17, batch 2500, loss[loss=0.2524, ctc_loss=0.1717, cr_loss=0.4035, over 17220.00 frames. ], tot_loss[loss=0.2239, ctc_loss=0.1506, cr_loss=0.3665, over 3359423.62 frames. ], batch size: 50, lr: 7.05e-03, grad_scale: 32.0
2024-09-23 16:49:48,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=302568.0, ans=0.1
2024-09-23 16:50:11,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=302614.6666666667, ans=0.2
2024-09-23 16:50:30,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=302661.3333333333, ans=0.0
2024-09-23 16:50:30,885 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=7.76 vs. limit=15.0
2024-09-23 16:50:39,595 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.277e+02 1.416e+02 1.598e+02 3.065e+02, threshold=2.832e+02, percent-clipped=1.0
2024-09-23 16:51:07,612 INFO [train.py:1198] (1/4) Epoch 17, batch 2550, loss[loss=0.2322, ctc_loss=0.1567, cr_loss=0.3775, over 17216.00 frames. ], tot_loss[loss=0.2237, ctc_loss=0.1504, cr_loss=0.3665, over 3357312.29 frames. ], batch size: 50, lr: 7.05e-03, grad_scale: 32.0
2024-09-23 16:51:07,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=302801.3333333333, ans=0.125
2024-09-23 16:51:08,134 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.10 vs. limit=12.0
2024-09-23 16:51:14,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=302801.3333333333, ans=0.025
2024-09-23 16:51:27,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=302848.0, ans=0.0
2024-09-23 16:51:33,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=302848.0, ans=0.125
2024-09-23 16:51:44,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=302894.6666666667, ans=0.0
2024-09-23 16:51:47,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=302894.6666666667, ans=0.125
2024-09-23 16:51:53,257 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.16 vs. limit=10.0
2024-09-23 16:52:08,925 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.88 vs. limit=15.0
2024-09-23 16:52:11,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=302988.0, ans=0.0
2024-09-23 16:52:29,692 INFO [train.py:1198] (1/4) Epoch 17, batch 2600, loss[loss=0.2509, ctc_loss=0.1674, cr_loss=0.4173, over 17029.00 frames. ], tot_loss[loss=0.224, ctc_loss=0.1507, cr_loss=0.3665, over 3356276.17 frames. ], batch size: 56, lr: 7.05e-03, grad_scale: 32.0
2024-09-23 16:52:36,817 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.21 vs. limit=6.0
2024-09-23 16:52:57,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=303081.3333333333, ans=0.1
2024-09-23 16:53:20,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=303174.6666666667, ans=0.0
2024-09-23 16:53:22,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=303174.6666666667, ans=0.125
2024-09-23 16:53:23,708 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.309e+02 1.432e+02 1.509e+02 2.078e+02, threshold=2.863e+02, percent-clipped=0.0
2024-09-23 16:53:24,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=303174.6666666667, ans=0.1
2024-09-23 16:53:49,177 INFO [train.py:1198] (1/4) Epoch 17, batch 2650, loss[loss=0.264, ctc_loss=0.1822, cr_loss=0.4093, over 16521.00 frames. ], tot_loss[loss=0.2232, ctc_loss=0.1501, cr_loss=0.3659, over 3361272.30 frames. ], batch size: 66, lr: 7.04e-03, grad_scale: 32.0
2024-09-23 16:54:01,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=303268.0, ans=0.125
2024-09-23 16:54:14,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=303314.6666666667, ans=0.0
2024-09-23 16:54:19,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=303361.3333333333, ans=0.04949747468305833
2024-09-23 16:54:23,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=303361.3333333333, ans=0.07
2024-09-23 16:54:28,297 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.71 vs. limit=15.0
2024-09-23 16:54:34,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=303361.3333333333, ans=0.125
2024-09-23 16:54:46,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=303408.0, ans=0.1
2024-09-23 16:55:12,780 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.25 vs. limit=15.0
2024-09-23 16:55:13,734 INFO [train.py:1198] (1/4) Epoch 17, batch 2700, loss[loss=0.2293, ctc_loss=0.1532, cr_loss=0.3807, over 17334.00 frames. ], tot_loss[loss=0.2233, ctc_loss=0.1502, cr_loss=0.3653, over 3363362.40 frames. ], batch size: 52, lr: 7.04e-03, grad_scale: 32.0
2024-09-23 16:55:17,433 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.06 vs. limit=15.0
2024-09-23 16:55:25,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=303501.3333333333, ans=0.0
2024-09-23 16:55:50,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=303594.6666666667, ans=0.0
2024-09-23 16:55:53,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=303594.6666666667, ans=0.09899494936611666
2024-09-23 16:55:55,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=303594.6666666667, ans=0.07
2024-09-23 16:56:07,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=303641.3333333333, ans=15.0
2024-09-23 16:56:10,277 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.299e+02 1.395e+02 1.578e+02 3.213e+02, threshold=2.790e+02, percent-clipped=1.0
2024-09-23 16:56:13,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=303641.3333333333, ans=0.125
2024-09-23 16:56:35,876 INFO [train.py:1198] (1/4) Epoch 17, batch 2750, loss[loss=0.2449, ctc_loss=0.1626, cr_loss=0.4117, over 17317.00 frames. ], tot_loss[loss=0.224, ctc_loss=0.1507, cr_loss=0.3665, over 3345880.35 frames. ], batch size: 49, lr: 7.04e-03, grad_scale: 32.0
2024-09-23 16:56:48,198 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=14.14 vs. limit=15.0
2024-09-23 16:57:06,106 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.73 vs. limit=22.5
2024-09-23 16:57:08,334 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 16:57:30,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=303874.6666666667, ans=0.025
2024-09-23 16:57:31,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=303874.6666666667, ans=0.125
2024-09-23 16:57:49,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=303921.3333333333, ans=0.125
2024-09-23 16:57:51,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=303921.3333333333, ans=0.2
2024-09-23 16:57:58,357 INFO [train.py:1198] (1/4) Epoch 17, batch 2800, loss[loss=0.2305, ctc_loss=0.1542, cr_loss=0.3815, over 16737.00 frames. ], tot_loss[loss=0.224, ctc_loss=0.1507, cr_loss=0.3664, over 3351041.10 frames. ], batch size: 61, lr: 7.04e-03, grad_scale: 32.0
2024-09-23 16:58:16,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=304014.6666666667, ans=0.125
2024-09-23 16:58:22,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=304014.6666666667, ans=0.125
2024-09-23 16:58:31,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=304061.3333333333, ans=0.1
2024-09-23 16:58:53,780 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.311e+02 1.385e+02 1.466e+02 2.534e+02, threshold=2.770e+02, percent-clipped=0.0
2024-09-23 16:59:17,881 INFO [train.py:1198] (1/4) Epoch 17, batch 2850, loss[loss=0.1887, ctc_loss=0.1261, cr_loss=0.3131, over 17042.00 frames. ], tot_loss[loss=0.223, ctc_loss=0.15, cr_loss=0.3647, over 3351897.02 frames. ], batch size: 39, lr: 7.03e-03, grad_scale: 16.0
2024-09-23 16:59:32,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=304248.0, ans=0.05
2024-09-23 16:59:54,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=304294.6666666667, ans=0.025
2024-09-23 17:00:05,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=304294.6666666667, ans=0.125
2024-09-23 17:00:11,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=304341.3333333333, ans=0.125
2024-09-23 17:00:15,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=304341.3333333333, ans=0.025
2024-09-23 17:00:42,927 INFO [train.py:1198] (1/4) Epoch 17, batch 2900, loss[loss=0.2369, ctc_loss=0.1591, cr_loss=0.3891, over 17070.00 frames. ], tot_loss[loss=0.2228, ctc_loss=0.1499, cr_loss=0.3645, over 3349269.31 frames. ], batch size: 46, lr: 7.03e-03, grad_scale: 16.0
2024-09-23 17:00:49,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=304434.6666666667, ans=0.125
2024-09-23 17:01:30,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=304528.0, ans=0.125
2024-09-23 17:01:35,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=304574.6666666667, ans=0.125
2024-09-23 17:01:42,987 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.263e+02 1.350e+02 1.442e+02 2.620e+02, threshold=2.699e+02, percent-clipped=0.0
2024-09-23 17:01:46,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=304574.6666666667, ans=0.025
2024-09-23 17:01:48,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=304621.3333333333, ans=0.0
2024-09-23 17:01:49,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=304621.3333333333, ans=0.125
2024-09-23 17:02:04,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.22 vs. limit=6.0
2024-09-23 17:02:04,772 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.38 vs. limit=15.0
2024-09-23 17:02:05,521 INFO [train.py:1198] (1/4) Epoch 17, batch 2950, loss[loss=0.2219, ctc_loss=0.1443, cr_loss=0.388, over 17072.00 frames. ], tot_loss[loss=0.2234, ctc_loss=0.1503, cr_loss=0.3654, over 3348813.00 frames. ], batch size: 46, lr: 7.03e-03, grad_scale: 8.0
2024-09-23 17:02:15,750 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.25 vs. limit=15.0
2024-09-23 17:02:21,620 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:02:30,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=304714.6666666667, ans=0.1
2024-09-23 17:02:30,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=304714.6666666667, ans=0.125
2024-09-23 17:02:30,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=304714.6666666667, ans=0.0
2024-09-23 17:02:36,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=304714.6666666667, ans=0.125
2024-09-23 17:02:36,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=304714.6666666667, ans=0.1
2024-09-23 17:02:38,628 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.21 vs. limit=15.0
2024-09-23 17:02:52,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=304761.3333333333, ans=0.1
2024-09-23 17:03:26,627 INFO [train.py:1198] (1/4) Epoch 17, batch 3000, loss[loss=0.2384, ctc_loss=0.158, cr_loss=0.4016, over 17032.00 frames. ], tot_loss[loss=0.2236, ctc_loss=0.1505, cr_loss=0.3656, over 3348260.31 frames. ], batch size: 52, lr: 7.02e-03, grad_scale: 8.0
2024-09-23 17:03:26,628 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 17:03:42,428 INFO [train.py:1230] (1/4) Epoch 17, validation: loss=0.0409, ctc_loss=0.0409, cr_loss=7.678e-15, over 944034.00 frames. 
2024-09-23 17:03:42,429 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 17:03:55,488 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.63 vs. limit=10.0
2024-09-23 17:04:27,748 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:04:36,068 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=7.97 vs. limit=15.0
2024-09-23 17:04:38,119 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.276e+02 1.374e+02 1.513e+02 2.906e+02, threshold=2.749e+02, percent-clipped=1.0
2024-09-23 17:04:58,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=305134.6666666667, ans=0.125
2024-09-23 17:04:59,729 INFO [train.py:1198] (1/4) Epoch 17, batch 3050, loss[loss=0.2317, ctc_loss=0.1591, cr_loss=0.363, over 17003.00 frames. ], tot_loss[loss=0.2222, ctc_loss=0.1495, cr_loss=0.3633, over 3347061.95 frames. ], batch size: 53, lr: 7.02e-03, grad_scale: 8.0
2024-09-23 17:05:05,225 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.17 vs. limit=22.5
2024-09-23 17:05:09,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=305134.6666666667, ans=0.1
2024-09-23 17:05:12,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=305134.6666666667, ans=0.125
2024-09-23 17:05:36,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=305228.0, ans=0.125
2024-09-23 17:05:48,941 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.82 vs. limit=15.0
2024-09-23 17:05:56,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=305274.6666666667, ans=0.0
2024-09-23 17:06:07,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.20 vs. limit=22.5
2024-09-23 17:06:09,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=305321.3333333333, ans=0.0
2024-09-23 17:06:20,562 INFO [train.py:1198] (1/4) Epoch 17, batch 3100, loss[loss=0.2171, ctc_loss=0.1445, cr_loss=0.3633, over 17207.00 frames. ], tot_loss[loss=0.222, ctc_loss=0.1491, cr_loss=0.3641, over 3351161.38 frames. ], batch size: 47, lr: 7.02e-03, grad_scale: 8.0
2024-09-23 17:06:22,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=305368.0, ans=0.1
2024-09-23 17:06:59,870 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.25 vs. limit=22.5
2024-09-23 17:07:19,080 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.257e+02 1.347e+02 1.447e+02 2.070e+02, threshold=2.694e+02, percent-clipped=0.0
2024-09-23 17:07:25,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=305554.6666666667, ans=0.0
2024-09-23 17:07:30,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=305554.6666666667, ans=0.0
2024-09-23 17:07:39,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=305601.3333333333, ans=0.2
2024-09-23 17:07:41,116 INFO [train.py:1198] (1/4) Epoch 17, batch 3150, loss[loss=0.2009, ctc_loss=0.1325, cr_loss=0.3424, over 17095.00 frames. ], tot_loss[loss=0.2212, ctc_loss=0.1487, cr_loss=0.3626, over 3349731.63 frames. ], batch size: 46, lr: 7.02e-03, grad_scale: 8.0
2024-09-23 17:07:44,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=305601.3333333333, ans=0.0
2024-09-23 17:08:00,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=305648.0, ans=0.0
2024-09-23 17:08:42,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=305741.3333333333, ans=0.1
2024-09-23 17:08:45,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=305788.0, ans=0.125
2024-09-23 17:08:49,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=305788.0, ans=0.1
2024-09-23 17:09:00,534 INFO [train.py:1198] (1/4) Epoch 17, batch 3200, loss[loss=0.1886, ctc_loss=0.1245, cr_loss=0.3204, over 17032.00 frames. ], tot_loss[loss=0.2188, ctc_loss=0.1468, cr_loss=0.3595, over 3355292.57 frames. ], batch size: 39, lr: 7.01e-03, grad_scale: 16.0
2024-09-23 17:09:25,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.76 vs. limit=15.0
2024-09-23 17:09:48,190 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.68 vs. limit=15.0
2024-09-23 17:09:56,501 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.224e+02 1.300e+02 1.391e+02 2.057e+02, threshold=2.599e+02, percent-clipped=0.0
2024-09-23 17:10:18,312 INFO [train.py:1198] (1/4) Epoch 17, batch 3250, loss[loss=0.2447, ctc_loss=0.1688, cr_loss=0.3793, over 14905.00 frames. ], tot_loss[loss=0.221, ctc_loss=0.1484, cr_loss=0.3627, over 3358502.39 frames. ], batch size: 88, lr: 7.01e-03, grad_scale: 16.0
2024-09-23 17:10:24,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=306068.0, ans=0.125
2024-09-23 17:10:52,633 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.73 vs. limit=15.0
2024-09-23 17:11:05,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=306208.0, ans=0.125
2024-09-23 17:11:07,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=306208.0, ans=0.0
2024-09-23 17:11:07,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=306208.0, ans=0.1
2024-09-23 17:11:22,367 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=9.16 vs. limit=15.0
2024-09-23 17:11:27,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=306254.6666666667, ans=0.1
2024-09-23 17:11:36,836 INFO [train.py:1198] (1/4) Epoch 17, batch 3300, loss[loss=0.2698, ctc_loss=0.1925, cr_loss=0.3867, over 12174.00 frames. ], tot_loss[loss=0.2226, ctc_loss=0.1497, cr_loss=0.3645, over 3350640.29 frames. ], batch size: 123, lr: 7.01e-03, grad_scale: 16.0
2024-09-23 17:12:18,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=306394.6666666667, ans=0.125
2024-09-23 17:12:34,842 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.271e+02 1.388e+02 1.557e+02 2.598e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-23 17:12:39,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=306488.0, ans=0.125
2024-09-23 17:12:40,112 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.91 vs. limit=15.0
2024-09-23 17:12:56,558 INFO [train.py:1198] (1/4) Epoch 17, batch 3350, loss[loss=0.2137, ctc_loss=0.1387, cr_loss=0.375, over 17348.00 frames. ], tot_loss[loss=0.2223, ctc_loss=0.1493, cr_loss=0.3651, over 3355988.85 frames. ], batch size: 48, lr: 7.01e-03, grad_scale: 16.0
2024-09-23 17:13:02,201 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.37 vs. limit=15.0
2024-09-23 17:13:20,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=306581.3333333333, ans=0.125
2024-09-23 17:14:03,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=306721.3333333333, ans=0.125
2024-09-23 17:14:14,500 INFO [train.py:1198] (1/4) Epoch 17, batch 3400, loss[loss=0.2157, ctc_loss=0.1444, cr_loss=0.3563, over 17261.00 frames. ], tot_loss[loss=0.2224, ctc_loss=0.1494, cr_loss=0.3649, over 3358844.26 frames. ], batch size: 44, lr: 7.00e-03, grad_scale: 16.0
2024-09-23 17:14:14,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=306768.0, ans=0.125
2024-09-23 17:14:24,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.05 vs. limit=15.0
2024-09-23 17:14:33,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=306814.6666666667, ans=0.1
2024-09-23 17:15:09,980 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.230e+02 1.305e+02 1.432e+02 2.019e+02, threshold=2.610e+02, percent-clipped=0.0
2024-09-23 17:15:10,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.63 vs. limit=15.0
2024-09-23 17:15:24,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=306954.6666666667, ans=0.125
2024-09-23 17:15:32,032 INFO [train.py:1198] (1/4) Epoch 17, batch 3450, loss[loss=0.2131, ctc_loss=0.1406, cr_loss=0.3624, over 17303.00 frames. ], tot_loss[loss=0.2211, ctc_loss=0.1483, cr_loss=0.3635, over 3366552.99 frames. ], batch size: 49, lr: 7.00e-03, grad_scale: 16.0
2024-09-23 17:16:02,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=307094.6666666667, ans=0.2
2024-09-23 17:16:17,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=307141.3333333333, ans=0.125
2024-09-23 17:16:20,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.80 vs. limit=15.0
2024-09-23 17:16:22,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=307141.3333333333, ans=0.125
2024-09-23 17:16:53,994 INFO [train.py:1198] (1/4) Epoch 17, batch 3500, loss[loss=0.2263, ctc_loss=0.1525, cr_loss=0.369, over 17207.00 frames. ], tot_loss[loss=0.2212, ctc_loss=0.1485, cr_loss=0.3636, over 3365359.79 frames. ], batch size: 47, lr: 7.00e-03, grad_scale: 16.0
2024-09-23 17:17:05,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=307234.6666666667, ans=0.125
2024-09-23 17:17:10,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=307281.3333333333, ans=0.1
2024-09-23 17:17:25,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=307328.0, ans=0.0
2024-09-23 17:17:33,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=307328.0, ans=0.125
2024-09-23 17:17:43,404 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.49 vs. limit=15.0
2024-09-23 17:17:47,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=307374.6666666667, ans=0.2
2024-09-23 17:17:50,323 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.294e+02 1.368e+02 1.478e+02 3.708e+02, threshold=2.737e+02, percent-clipped=1.0
2024-09-23 17:18:02,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=307421.3333333333, ans=0.125
2024-09-23 17:18:12,054 INFO [train.py:1198] (1/4) Epoch 17, batch 3550, loss[loss=0.2678, ctc_loss=0.1861, cr_loss=0.4088, over 17028.00 frames. ], tot_loss[loss=0.2224, ctc_loss=0.1494, cr_loss=0.3648, over 3353936.94 frames. ], batch size: 56, lr: 7.00e-03, grad_scale: 16.0
2024-09-23 17:18:21,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=307468.0, ans=0.125
2024-09-23 17:18:28,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=307514.6666666667, ans=0.125
2024-09-23 17:18:51,929 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.93 vs. limit=15.0
2024-09-23 17:18:56,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=307561.3333333333, ans=0.125
2024-09-23 17:18:59,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=307608.0, ans=0.2
2024-09-23 17:19:00,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=307608.0, ans=0.025
2024-09-23 17:19:08,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=307608.0, ans=0.125
2024-09-23 17:19:15,213 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.90 vs. limit=10.0
2024-09-23 17:19:21,470 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.85 vs. limit=15.0
2024-09-23 17:19:31,529 INFO [train.py:1198] (1/4) Epoch 17, batch 3600, loss[loss=0.2533, ctc_loss=0.1693, cr_loss=0.4198, over 16755.00 frames. ], tot_loss[loss=0.2224, ctc_loss=0.1495, cr_loss=0.3648, over 3352519.98 frames. ], batch size: 61, lr: 6.99e-03, grad_scale: 16.0
2024-09-23 17:19:42,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=307701.3333333333, ans=0.125
2024-09-23 17:20:15,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=307794.6666666667, ans=0.09899494936611666
2024-09-23 17:20:27,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=307841.3333333333, ans=0.0
2024-09-23 17:20:29,251 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.229e+02 1.313e+02 1.436e+02 1.870e+02, threshold=2.625e+02, percent-clipped=0.0
2024-09-23 17:20:46,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=307888.0, ans=0.0
2024-09-23 17:20:48,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=307934.6666666667, ans=0.125
2024-09-23 17:20:49,540 INFO [train.py:1198] (1/4) Epoch 17, batch 3650, loss[loss=0.2059, ctc_loss=0.1344, cr_loss=0.3575, over 17183.00 frames. ], tot_loss[loss=0.2219, ctc_loss=0.149, cr_loss=0.3644, over 3357357.02 frames. ], batch size: 41, lr: 6.99e-03, grad_scale: 16.0
2024-09-23 17:21:03,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=307981.3333333333, ans=0.0
2024-09-23 17:21:05,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=307981.3333333333, ans=10.0
2024-09-23 17:21:41,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=308074.6666666667, ans=0.125
2024-09-23 17:21:44,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=308074.6666666667, ans=0.0
2024-09-23 17:21:53,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=308121.3333333333, ans=0.0
2024-09-23 17:21:53,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=308121.3333333333, ans=0.2
2024-09-23 17:21:57,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=308121.3333333333, ans=0.95
2024-09-23 17:22:01,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=308121.3333333333, ans=0.025
2024-09-23 17:22:10,729 INFO [train.py:1198] (1/4) Epoch 17, batch 3700, loss[loss=0.2413, ctc_loss=0.1623, cr_loss=0.3946, over 17003.00 frames. ], tot_loss[loss=0.2227, ctc_loss=0.1497, cr_loss=0.3653, over 3340600.49 frames. ], batch size: 56, lr: 6.99e-03, grad_scale: 16.0
2024-09-23 17:22:22,844 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.38 vs. limit=22.5
2024-09-23 17:22:34,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=308214.6666666667, ans=0.125
2024-09-23 17:22:34,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=308214.6666666667, ans=0.125
2024-09-23 17:22:34,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=308214.6666666667, ans=0.125
2024-09-23 17:22:45,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=308261.3333333333, ans=0.025
2024-09-23 17:23:05,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=308308.0, ans=0.0
2024-09-23 17:23:07,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=308308.0, ans=0.0
2024-09-23 17:23:08,758 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.332e+02 1.441e+02 1.620e+02 2.318e+02, threshold=2.882e+02, percent-clipped=0.0
2024-09-23 17:23:28,817 INFO [train.py:1198] (1/4) Epoch 17, batch 3750, loss[loss=0.2379, ctc_loss=0.1612, cr_loss=0.3837, over 17023.00 frames. ], tot_loss[loss=0.2219, ctc_loss=0.1491, cr_loss=0.364, over 3340365.14 frames. ], batch size: 51, lr: 6.99e-03, grad_scale: 16.0
2024-09-23 17:23:38,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=308401.3333333333, ans=0.125
2024-09-23 17:23:54,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=308448.0, ans=0.0
2024-09-23 17:23:57,491 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:24:27,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=308541.3333333333, ans=0.0
2024-09-23 17:24:35,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=308588.0, ans=0.125
2024-09-23 17:24:44,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=308588.0, ans=0.2
2024-09-23 17:24:47,306 INFO [train.py:1198] (1/4) Epoch 17, batch 3800, loss[loss=0.2592, ctc_loss=0.1772, cr_loss=0.4098, over 16994.00 frames. ], tot_loss[loss=0.2221, ctc_loss=0.1493, cr_loss=0.3638, over 3330990.49 frames. ], batch size: 53, lr: 6.98e-03, grad_scale: 16.0
2024-09-23 17:24:47,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=308634.6666666667, ans=0.125
2024-09-23 17:24:47,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=308634.6666666667, ans=0.025
2024-09-23 17:25:05,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.92 vs. limit=22.5
2024-09-23 17:25:38,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=308774.6666666667, ans=0.0
2024-09-23 17:25:39,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=308774.6666666667, ans=0.0
2024-09-23 17:25:41,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=308774.6666666667, ans=0.125
2024-09-23 17:25:45,744 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.280e+02 1.358e+02 1.516e+02 2.710e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-23 17:26:05,959 INFO [train.py:1198] (1/4) Epoch 17, batch 3850, loss[loss=0.2715, ctc_loss=0.1962, cr_loss=0.3764, over 12246.00 frames. ], tot_loss[loss=0.2238, ctc_loss=0.1508, cr_loss=0.3651, over 3307156.58 frames. ], batch size: 125, lr: 6.98e-03, grad_scale: 16.0
2024-09-23 17:26:41,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=308961.3333333333, ans=0.2
2024-09-23 17:26:47,530 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:26:56,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=309008.0, ans=0.125
2024-09-23 17:27:11,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=309054.6666666667, ans=0.125
2024-09-23 17:27:13,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=309054.6666666667, ans=0.1
2024-09-23 17:28:06,782 INFO [train.py:1198] (1/4) Epoch 18, batch 0, loss[loss=0.2341, ctc_loss=0.1536, cr_loss=0.4025, over 17105.00 frames. ], tot_loss[loss=0.2341, ctc_loss=0.1536, cr_loss=0.4025, over 17105.00 frames. ], batch size: 49, lr: 6.78e-03, grad_scale: 32.0
2024-09-23 17:28:06,783 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 17:28:21,935 INFO [train.py:1230] (1/4) Epoch 18, validation: loss=0.03994, ctc_loss=0.03994, cr_loss=8.27e-15, over 944034.00 frames. 
2024-09-23 17:28:21,936 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 17:28:26,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=309082.6666666667, ans=0.1
2024-09-23 17:29:09,166 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.47 vs. limit=15.0
2024-09-23 17:29:09,376 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.55 vs. limit=15.0
2024-09-23 17:29:30,231 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.283e+02 1.487e+02 1.642e+02 2.774e+02, threshold=2.974e+02, percent-clipped=1.0
2024-09-23 17:29:30,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=309269.3333333333, ans=0.0
2024-09-23 17:29:37,484 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.43 vs. limit=10.0
2024-09-23 17:29:44,550 INFO [train.py:1198] (1/4) Epoch 18, batch 50, loss[loss=0.1924, ctc_loss=0.1276, cr_loss=0.3242, over 17044.00 frames. ], tot_loss[loss=0.2173, ctc_loss=0.1457, cr_loss=0.3582, over 761801.81 frames. ], batch size: 39, lr: 6.78e-03, grad_scale: 32.0
2024-09-23 17:29:59,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.47 vs. limit=22.5
2024-09-23 17:30:13,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=309362.6666666667, ans=0.125
2024-09-23 17:30:31,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=309456.0, ans=0.09899494936611666
2024-09-23 17:30:42,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=309456.0, ans=0.5
2024-09-23 17:30:53,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=309502.6666666667, ans=0.1
2024-09-23 17:31:04,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=309502.6666666667, ans=0.125
2024-09-23 17:31:06,898 INFO [train.py:1198] (1/4) Epoch 18, batch 100, loss[loss=0.2325, ctc_loss=0.157, cr_loss=0.3775, over 17018.00 frames. ], tot_loss[loss=0.2194, ctc_loss=0.1472, cr_loss=0.3612, over 1344364.27 frames. ], batch size: 51, lr: 6.77e-03, grad_scale: 16.0
2024-09-23 17:31:12,765 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.85 vs. limit=15.0
2024-09-23 17:31:15,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=309549.3333333333, ans=0.0
2024-09-23 17:31:50,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=309642.6666666667, ans=0.0
2024-09-23 17:32:04,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=309689.3333333333, ans=0.125
2024-09-23 17:32:12,960 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.18 vs. limit=15.0
2024-09-23 17:32:13,774 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.252e+02 1.337e+02 1.407e+02 3.310e+02, threshold=2.674e+02, percent-clipped=1.0
2024-09-23 17:32:15,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=309736.0, ans=0.2
2024-09-23 17:32:25,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=309736.0, ans=0.2
2024-09-23 17:32:28,257 INFO [train.py:1198] (1/4) Epoch 18, batch 150, loss[loss=0.2459, ctc_loss=0.1676, cr_loss=0.3915, over 16054.00 frames. ], tot_loss[loss=0.2211, ctc_loss=0.1483, cr_loss=0.3639, over 1788928.52 frames. ], batch size: 74, lr: 6.77e-03, grad_scale: 16.0
2024-09-23 17:32:28,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=309782.6666666667, ans=0.09899494936611666
2024-09-23 17:32:44,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=309829.3333333333, ans=0.2
2024-09-23 17:32:45,041 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.57 vs. limit=15.0
2024-09-23 17:32:50,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=309829.3333333333, ans=0.09899494936611666
2024-09-23 17:33:15,043 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.60 vs. limit=6.0
2024-09-23 17:33:29,202 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.99 vs. limit=10.0
2024-09-23 17:33:44,577 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:33:46,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=309969.3333333333, ans=0.0
2024-09-23 17:33:50,778 INFO [train.py:1198] (1/4) Epoch 18, batch 200, loss[loss=0.2262, ctc_loss=0.1506, cr_loss=0.3778, over 17329.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.1471, cr_loss=0.3622, over 2145103.42 frames. ], batch size: 46, lr: 6.77e-03, grad_scale: 16.0
2024-09-23 17:33:52,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=310016.0, ans=0.0
2024-09-23 17:34:07,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=9.50 vs. limit=15.0
2024-09-23 17:34:37,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=310109.3333333333, ans=0.125
2024-09-23 17:34:37,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.43 vs. limit=15.0
2024-09-23 17:35:00,728 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.233e+02 1.340e+02 1.521e+02 2.141e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-23 17:35:07,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=310202.6666666667, ans=0.0
2024-09-23 17:35:07,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=310202.6666666667, ans=0.125
2024-09-23 17:35:10,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=310202.6666666667, ans=0.0
2024-09-23 17:35:13,460 INFO [train.py:1198] (1/4) Epoch 18, batch 250, loss[loss=0.2413, ctc_loss=0.1627, cr_loss=0.3927, over 16899.00 frames. ], tot_loss[loss=0.2194, ctc_loss=0.147, cr_loss=0.3619, over 2415238.78 frames. ], batch size: 58, lr: 6.77e-03, grad_scale: 16.0
2024-09-23 17:35:33,574 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=11.57 vs. limit=15.0
2024-09-23 17:35:47,855 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.19 vs. limit=12.0
2024-09-23 17:36:20,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=310436.0, ans=0.0
2024-09-23 17:36:21,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=310436.0, ans=0.125
2024-09-23 17:36:33,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=310436.0, ans=0.2
2024-09-23 17:36:36,380 INFO [train.py:1198] (1/4) Epoch 18, batch 300, loss[loss=0.2098, ctc_loss=0.1389, cr_loss=0.3546, over 17077.00 frames. ], tot_loss[loss=0.2198, ctc_loss=0.1472, cr_loss=0.3631, over 2623893.15 frames. ], batch size: 43, lr: 6.76e-03, grad_scale: 16.0
2024-09-23 17:36:53,389 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=11.19 vs. limit=15.0
2024-09-23 17:37:22,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=310576.0, ans=0.125
2024-09-23 17:37:28,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=310622.6666666667, ans=0.0
2024-09-23 17:37:32,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=310622.6666666667, ans=0.125
2024-09-23 17:37:46,270 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.225e+02 1.352e+02 1.557e+02 2.949e+02, threshold=2.705e+02, percent-clipped=1.0
2024-09-23 17:38:01,323 INFO [train.py:1198] (1/4) Epoch 18, batch 350, loss[loss=0.1848, ctc_loss=0.1214, cr_loss=0.3173, over 17142.00 frames. ], tot_loss[loss=0.2212, ctc_loss=0.1483, cr_loss=0.3648, over 2795384.22 frames. ], batch size: 40, lr: 6.76e-03, grad_scale: 16.0
2024-09-23 17:38:03,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=310716.0, ans=0.0
2024-09-23 17:38:39,693 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.10 vs. limit=15.0
2024-09-23 17:38:45,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=310809.3333333333, ans=0.125
2024-09-23 17:38:55,314 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.27 vs. limit=15.0
2024-09-23 17:39:01,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=310856.0, ans=0.125
2024-09-23 17:39:24,258 INFO [train.py:1198] (1/4) Epoch 18, batch 400, loss[loss=0.2142, ctc_loss=0.1424, cr_loss=0.3591, over 17082.00 frames. ], tot_loss[loss=0.2206, ctc_loss=0.1478, cr_loss=0.3636, over 2920409.81 frames. ], batch size: 46, lr: 6.76e-03, grad_scale: 32.0
2024-09-23 17:39:30,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=310949.3333333333, ans=0.2
2024-09-23 17:39:45,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=310996.0, ans=0.2
2024-09-23 17:39:50,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=310996.0, ans=0.1
2024-09-23 17:39:50,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=310996.0, ans=0.0
2024-09-23 17:40:21,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=311089.3333333333, ans=0.125
2024-09-23 17:40:26,983 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.58 vs. limit=22.5
2024-09-23 17:40:28,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=311136.0, ans=0.125
2024-09-23 17:40:31,100 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.269e+02 1.393e+02 1.575e+02 2.470e+02, threshold=2.786e+02, percent-clipped=0.0
2024-09-23 17:40:43,672 INFO [train.py:1198] (1/4) Epoch 18, batch 450, loss[loss=0.2242, ctc_loss=0.15, cr_loss=0.3713, over 17016.00 frames. ], tot_loss[loss=0.2206, ctc_loss=0.1479, cr_loss=0.3635, over 3023354.58 frames. ], batch size: 51, lr: 6.76e-03, grad_scale: 32.0
2024-09-23 17:40:50,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=311182.6666666667, ans=0.0
2024-09-23 17:41:11,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=311229.3333333333, ans=0.0
2024-09-23 17:41:16,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=311276.0, ans=0.0
2024-09-23 17:41:19,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=311276.0, ans=0.0
2024-09-23 17:41:25,379 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.29 vs. limit=12.0
2024-09-23 17:41:53,396 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.70 vs. limit=15.0
2024-09-23 17:41:55,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten.whitening_limit, batch_count=311369.3333333333, ans=15.0
2024-09-23 17:42:05,309 INFO [train.py:1198] (1/4) Epoch 18, batch 500, loss[loss=0.2178, ctc_loss=0.1441, cr_loss=0.3683, over 17362.00 frames. ], tot_loss[loss=0.2208, ctc_loss=0.148, cr_loss=0.3638, over 3097829.77 frames. ], batch size: 48, lr: 6.75e-03, grad_scale: 32.0
2024-09-23 17:42:34,456 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.68 vs. limit=15.0
2024-09-23 17:42:38,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=311509.3333333333, ans=0.125
2024-09-23 17:43:08,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=311556.0, ans=0.125
2024-09-23 17:43:17,645 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.265e+02 1.370e+02 1.572e+02 2.414e+02, threshold=2.740e+02, percent-clipped=0.0
2024-09-23 17:43:22,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=311602.6666666667, ans=0.125
2024-09-23 17:43:30,320 INFO [train.py:1198] (1/4) Epoch 18, batch 550, loss[loss=0.2229, ctc_loss=0.15, cr_loss=0.3646, over 16739.00 frames. ], tot_loss[loss=0.2203, ctc_loss=0.1478, cr_loss=0.3626, over 3159386.25 frames. ], batch size: 61, lr: 6.75e-03, grad_scale: 32.0
2024-09-23 17:43:30,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=311649.3333333333, ans=10.0
2024-09-23 17:44:09,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=311742.6666666667, ans=0.1
2024-09-23 17:44:24,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=311789.3333333333, ans=0.125
2024-09-23 17:44:27,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=311789.3333333333, ans=0.2
2024-09-23 17:44:32,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=311789.3333333333, ans=0.2
2024-09-23 17:44:43,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=311836.0, ans=0.125
2024-09-23 17:44:53,369 INFO [train.py:1198] (1/4) Epoch 18, batch 600, loss[loss=0.2004, ctc_loss=0.1338, cr_loss=0.3327, over 17219.00 frames. ], tot_loss[loss=0.2201, ctc_loss=0.1475, cr_loss=0.3627, over 3215834.00 frames. ], batch size: 41, lr: 6.75e-03, grad_scale: 32.0
2024-09-23 17:45:08,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=311929.3333333333, ans=0.2
2024-09-23 17:45:10,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=311929.3333333333, ans=6.0
2024-09-23 17:45:51,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=312022.6666666667, ans=0.0
2024-09-23 17:46:04,639 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.282e+02 1.384e+02 1.538e+02 2.458e+02, threshold=2.768e+02, percent-clipped=0.0
2024-09-23 17:46:11,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=312069.3333333333, ans=0.1
2024-09-23 17:46:12,874 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:46:16,003 INFO [train.py:1198] (1/4) Epoch 18, batch 650, loss[loss=0.179, ctc_loss=0.116, cr_loss=0.3146, over 15807.00 frames. ], tot_loss[loss=0.2193, ctc_loss=0.1471, cr_loss=0.3606, over 3245503.16 frames. ], batch size: 35, lr: 6.75e-03, grad_scale: 16.0
2024-09-23 17:46:25,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=312116.0, ans=0.125
2024-09-23 17:47:38,725 INFO [train.py:1198] (1/4) Epoch 18, batch 700, loss[loss=0.2615, ctc_loss=0.1787, cr_loss=0.4137, over 14744.00 frames. ], tot_loss[loss=0.2191, ctc_loss=0.147, cr_loss=0.3602, over 3257197.10 frames. ], batch size: 89, lr: 6.74e-03, grad_scale: 16.0
2024-09-23 17:47:49,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=312349.3333333333, ans=0.125
2024-09-23 17:47:51,170 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:48:16,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=312442.6666666667, ans=0.0
2024-09-23 17:48:49,978 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.252e+02 1.369e+02 1.593e+02 2.409e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-23 17:49:00,933 INFO [train.py:1198] (1/4) Epoch 18, batch 750, loss[loss=0.2153, ctc_loss=0.1435, cr_loss=0.3585, over 17151.00 frames. ], tot_loss[loss=0.2194, ctc_loss=0.1473, cr_loss=0.3605, over 3279683.45 frames. ], batch size: 45, lr: 6.74e-03, grad_scale: 16.0
2024-09-23 17:49:16,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=312582.6666666667, ans=0.1
2024-09-23 17:49:16,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=312582.6666666667, ans=0.04949747468305833
2024-09-23 17:49:18,062 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:49:35,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=312676.0, ans=0.07
2024-09-23 17:49:38,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=312676.0, ans=0.2
2024-09-23 17:49:51,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=312722.6666666667, ans=0.125
2024-09-23 17:49:55,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=312722.6666666667, ans=0.0
2024-09-23 17:50:18,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=312769.3333333333, ans=0.07
2024-09-23 17:50:23,308 INFO [train.py:1198] (1/4) Epoch 18, batch 800, loss[loss=0.2621, ctc_loss=0.1754, cr_loss=0.4336, over 16990.00 frames. ], tot_loss[loss=0.2203, ctc_loss=0.148, cr_loss=0.3616, over 3299853.00 frames. ], batch size: 53, lr: 6.74e-03, grad_scale: 32.0
2024-09-23 17:50:25,504 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.28 vs. limit=15.0
2024-09-23 17:50:55,256 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.57 vs. limit=10.0
2024-09-23 17:50:56,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=312909.3333333333, ans=0.0
2024-09-23 17:51:28,818 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.79 vs. limit=22.5
2024-09-23 17:51:34,555 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.268e+02 1.368e+02 1.483e+02 2.318e+02, threshold=2.737e+02, percent-clipped=0.0
2024-09-23 17:51:38,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=313002.6666666667, ans=0.125
2024-09-23 17:51:45,656 INFO [train.py:1198] (1/4) Epoch 18, batch 850, loss[loss=0.2295, ctc_loss=0.1489, cr_loss=0.4031, over 17058.00 frames. ], tot_loss[loss=0.2205, ctc_loss=0.148, cr_loss=0.3621, over 3323464.42 frames. ], batch size: 46, lr: 6.74e-03, grad_scale: 32.0
2024-09-23 17:51:53,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=313049.3333333333, ans=0.125
2024-09-23 17:52:03,011 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.63 vs. limit=15.0
2024-09-23 17:52:04,311 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 17:52:29,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=313142.6666666667, ans=0.2
2024-09-23 17:52:31,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=313142.6666666667, ans=0.2
2024-09-23 17:52:33,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.71 vs. limit=6.0
2024-09-23 17:52:43,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=313189.3333333333, ans=0.0
2024-09-23 17:53:04,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=313236.0, ans=0.2
2024-09-23 17:53:10,236 INFO [train.py:1198] (1/4) Epoch 18, batch 900, loss[loss=0.1973, ctc_loss=0.1343, cr_loss=0.3151, over 17091.00 frames. ], tot_loss[loss=0.2199, ctc_loss=0.1476, cr_loss=0.3615, over 3337300.40 frames. ], batch size: 40, lr: 6.73e-03, grad_scale: 32.0
2024-09-23 17:53:10,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=313282.6666666667, ans=0.2
2024-09-23 17:53:10,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=313282.6666666667, ans=0.1
2024-09-23 17:53:27,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=313329.3333333333, ans=0.125
2024-09-23 17:53:56,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=313422.6666666667, ans=0.125
2024-09-23 17:54:18,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.03 vs. limit=15.0
2024-09-23 17:54:21,419 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.286e+02 1.405e+02 1.600e+02 2.669e+02, threshold=2.810e+02, percent-clipped=0.0
2024-09-23 17:54:26,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=313469.3333333333, ans=0.125
2024-09-23 17:54:26,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=313469.3333333333, ans=0.125
2024-09-23 17:54:32,762 INFO [train.py:1198] (1/4) Epoch 18, batch 950, loss[loss=0.2163, ctc_loss=0.1458, cr_loss=0.3526, over 17005.00 frames. ], tot_loss[loss=0.22, ctc_loss=0.1476, cr_loss=0.3619, over 3341674.84 frames. ], batch size: 53, lr: 6.73e-03, grad_scale: 32.0
2024-09-23 17:54:44,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=313516.0, ans=0.0
2024-09-23 17:54:48,478 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.82 vs. limit=22.5
2024-09-23 17:55:13,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=313609.3333333333, ans=0.2
2024-09-23 17:55:22,550 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.88 vs. limit=10.0
2024-09-23 17:55:34,784 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.93 vs. limit=15.0
2024-09-23 17:55:55,788 INFO [train.py:1198] (1/4) Epoch 18, batch 1000, loss[loss=0.1861, ctc_loss=0.1233, cr_loss=0.3139, over 17028.00 frames. ], tot_loss[loss=0.2199, ctc_loss=0.1475, cr_loss=0.3622, over 3346298.86 frames. ], batch size: 39, lr: 6.73e-03, grad_scale: 32.0
2024-09-23 17:56:04,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=313749.3333333333, ans=0.1
2024-09-23 17:56:44,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=313889.3333333333, ans=0.0
2024-09-23 17:57:00,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=313936.0, ans=0.125
2024-09-23 17:57:06,776 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.230e+02 1.328e+02 1.431e+02 2.241e+02, threshold=2.657e+02, percent-clipped=0.0
2024-09-23 17:57:17,823 INFO [train.py:1198] (1/4) Epoch 18, batch 1050, loss[loss=0.2495, ctc_loss=0.1733, cr_loss=0.3811, over 17242.00 frames. ], tot_loss[loss=0.2206, ctc_loss=0.1481, cr_loss=0.3624, over 3336867.40 frames. ], batch size: 50, lr: 6.73e-03, grad_scale: 32.0
2024-09-23 17:57:41,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=314029.3333333333, ans=0.0
2024-09-23 17:57:45,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=314029.3333333333, ans=0.125
2024-09-23 17:58:01,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=314076.0, ans=0.0
2024-09-23 17:58:26,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=314169.3333333333, ans=0.1
2024-09-23 17:58:34,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=314169.3333333333, ans=10.0
2024-09-23 17:58:40,062 INFO [train.py:1198] (1/4) Epoch 18, batch 1100, loss[loss=0.2177, ctc_loss=0.1455, cr_loss=0.3612, over 17091.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.1468, cr_loss=0.3607, over 3349787.40 frames. ], batch size: 49, lr: 6.72e-03, grad_scale: 32.0
2024-09-23 17:58:46,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=314216.0, ans=0.0
2024-09-23 17:59:02,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=314262.6666666667, ans=0.035
2024-09-23 17:59:04,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=314262.6666666667, ans=0.125
2024-09-23 17:59:40,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=314356.0, ans=0.1
2024-09-23 17:59:51,144 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.280e+02 1.383e+02 1.491e+02 2.284e+02, threshold=2.766e+02, percent-clipped=0.0
2024-09-23 17:59:59,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=314402.6666666667, ans=0.0
2024-09-23 18:00:02,377 INFO [train.py:1198] (1/4) Epoch 18, batch 1150, loss[loss=0.2002, ctc_loss=0.1317, cr_loss=0.3424, over 17116.00 frames. ], tot_loss[loss=0.2206, ctc_loss=0.148, cr_loss=0.3633, over 3349840.37 frames. ], batch size: 40, lr: 6.72e-03, grad_scale: 32.0
2024-09-23 18:00:20,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=314496.0, ans=0.125
2024-09-23 18:00:34,315 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.50 vs. limit=6.0
2024-09-23 18:00:39,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=314542.6666666667, ans=0.025
2024-09-23 18:00:50,615 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.84 vs. limit=12.0
2024-09-23 18:01:25,140 INFO [train.py:1198] (1/4) Epoch 18, batch 1200, loss[loss=0.1948, ctc_loss=0.1288, cr_loss=0.3305, over 17208.00 frames. ], tot_loss[loss=0.2214, ctc_loss=0.1486, cr_loss=0.3639, over 3351372.23 frames. ], batch size: 47, lr: 6.72e-03, grad_scale: 32.0
2024-09-23 18:02:38,634 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.86 vs. limit=12.0
2024-09-23 18:02:38,964 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.291e+02 1.389e+02 1.552e+02 2.070e+02, threshold=2.777e+02, percent-clipped=0.0
2024-09-23 18:02:49,967 INFO [train.py:1198] (1/4) Epoch 18, batch 1250, loss[loss=0.1884, ctc_loss=0.1218, cr_loss=0.3329, over 16310.00 frames. ], tot_loss[loss=0.2219, ctc_loss=0.149, cr_loss=0.3647, over 3360775.37 frames. ], batch size: 36, lr: 6.72e-03, grad_scale: 32.0
2024-09-23 18:02:53,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=314916.0, ans=0.0
2024-09-23 18:03:04,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=314962.6666666667, ans=0.125
2024-09-23 18:03:13,340 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.21 vs. limit=6.0
2024-09-23 18:03:24,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=315009.3333333333, ans=0.125
2024-09-23 18:03:38,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=315056.0, ans=0.0
2024-09-23 18:03:44,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=315056.0, ans=0.1
2024-09-23 18:03:52,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=315102.6666666667, ans=0.0
2024-09-23 18:03:54,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=315102.6666666667, ans=0.0
2024-09-23 18:04:04,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=315102.6666666667, ans=0.125
2024-09-23 18:04:10,279 INFO [train.py:1198] (1/4) Epoch 18, batch 1300, loss[loss=0.2379, ctc_loss=0.1566, cr_loss=0.4069, over 17341.00 frames. ], tot_loss[loss=0.2214, ctc_loss=0.1485, cr_loss=0.3648, over 3368272.57 frames. ], batch size: 48, lr: 6.71e-03, grad_scale: 32.0
2024-09-23 18:04:12,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=315149.3333333333, ans=0.125
2024-09-23 18:04:17,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=315149.3333333333, ans=0.125
2024-09-23 18:04:19,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=315149.3333333333, ans=0.2
2024-09-23 18:04:49,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=315242.6666666667, ans=0.125
2024-09-23 18:05:17,648 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.34 vs. limit=22.5
2024-09-23 18:05:21,638 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.312e+02 1.416e+02 1.661e+02 2.501e+02, threshold=2.832e+02, percent-clipped=0.0
2024-09-23 18:05:28,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=315336.0, ans=0.015
2024-09-23 18:05:28,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=315336.0, ans=0.125
2024-09-23 18:05:32,770 INFO [train.py:1198] (1/4) Epoch 18, batch 1350, loss[loss=0.2469, ctc_loss=0.1651, cr_loss=0.4089, over 16513.00 frames. ], tot_loss[loss=0.22, ctc_loss=0.1475, cr_loss=0.3626, over 3379171.65 frames. ], batch size: 66, lr: 6.71e-03, grad_scale: 32.0
2024-09-23 18:05:37,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=315382.6666666667, ans=0.1
2024-09-23 18:05:45,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=315382.6666666667, ans=0.0
2024-09-23 18:06:15,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=315476.0, ans=0.125
2024-09-23 18:06:47,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=315569.3333333333, ans=0.5
2024-09-23 18:06:48,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=315569.3333333333, ans=0.1
2024-09-23 18:06:54,862 INFO [train.py:1198] (1/4) Epoch 18, batch 1400, loss[loss=0.2182, ctc_loss=0.1483, cr_loss=0.3495, over 16767.00 frames. ], tot_loss[loss=0.2184, ctc_loss=0.1464, cr_loss=0.3602, over 3381436.27 frames. ], batch size: 61, lr: 6.71e-03, grad_scale: 32.0
2024-09-23 18:07:04,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=315616.0, ans=0.125
2024-09-23 18:07:07,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=315616.0, ans=0.125
2024-09-23 18:07:50,416 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.00 vs. limit=22.5
2024-09-23 18:08:08,561 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.236e+02 1.313e+02 1.450e+02 2.376e+02, threshold=2.626e+02, percent-clipped=0.0
2024-09-23 18:08:19,846 INFO [train.py:1198] (1/4) Epoch 18, batch 1450, loss[loss=0.2198, ctc_loss=0.1468, cr_loss=0.3648, over 17070.00 frames. ], tot_loss[loss=0.2203, ctc_loss=0.1476, cr_loss=0.3632, over 3375410.96 frames. ], batch size: 46, lr: 6.71e-03, grad_scale: 32.0
2024-09-23 18:08:30,027 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.24 vs. limit=22.5
2024-09-23 18:09:30,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.45 vs. limit=15.0
2024-09-23 18:09:42,228 INFO [train.py:1198] (1/4) Epoch 18, batch 1500, loss[loss=0.1931, ctc_loss=0.1303, cr_loss=0.3141, over 16968.00 frames. ], tot_loss[loss=0.2198, ctc_loss=0.1471, cr_loss=0.3632, over 3380078.70 frames. ], batch size: 42, lr: 6.70e-03, grad_scale: 32.0
2024-09-23 18:10:14,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=316176.0, ans=0.125
2024-09-23 18:10:21,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=316176.0, ans=0.07
2024-09-23 18:10:21,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=316176.0, ans=0.125
2024-09-23 18:10:42,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=316222.6666666667, ans=0.125
2024-09-23 18:10:53,731 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.296e+02 1.405e+02 1.565e+02 2.957e+02, threshold=2.810e+02, percent-clipped=1.0
2024-09-23 18:11:05,019 INFO [train.py:1198] (1/4) Epoch 18, batch 1550, loss[loss=0.2683, ctc_loss=0.1867, cr_loss=0.4081, over 14890.00 frames. ], tot_loss[loss=0.2197, ctc_loss=0.1473, cr_loss=0.3619, over 3369846.93 frames. ], batch size: 89, lr: 6.70e-03, grad_scale: 32.0
2024-09-23 18:11:19,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=316362.6666666667, ans=0.125
2024-09-23 18:11:27,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=316362.6666666667, ans=0.125
2024-09-23 18:11:29,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=316362.6666666667, ans=0.2
2024-09-23 18:11:29,633 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=10.21 vs. limit=12.0
2024-09-23 18:11:34,512 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.46 vs. limit=22.5
2024-09-23 18:11:37,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=316409.3333333333, ans=0.125
2024-09-23 18:11:45,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=316409.3333333333, ans=0.1
2024-09-23 18:12:28,051 INFO [train.py:1198] (1/4) Epoch 18, batch 1600, loss[loss=0.1982, ctc_loss=0.1307, cr_loss=0.3372, over 17203.00 frames. ], tot_loss[loss=0.2209, ctc_loss=0.1483, cr_loss=0.3629, over 3349064.83 frames. ], batch size: 41, lr: 6.70e-03, grad_scale: 32.0
2024-09-23 18:12:36,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=316549.3333333333, ans=0.125
2024-09-23 18:12:52,041 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.46 vs. limit=15.0
2024-09-23 18:12:58,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=316596.0, ans=0.125
2024-09-23 18:13:22,773 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.79 vs. limit=6.0
2024-09-23 18:13:28,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=316689.3333333333, ans=0.04949747468305833
2024-09-23 18:13:31,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=316689.3333333333, ans=0.125
2024-09-23 18:13:39,405 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.280e+02 1.386e+02 1.539e+02 3.392e+02, threshold=2.772e+02, percent-clipped=1.0
2024-09-23 18:13:50,542 INFO [train.py:1198] (1/4) Epoch 18, batch 1650, loss[loss=0.2406, ctc_loss=0.1594, cr_loss=0.4059, over 17206.00 frames. ], tot_loss[loss=0.2201, ctc_loss=0.1478, cr_loss=0.3618, over 3344497.71 frames. ], batch size: 55, lr: 6.70e-03, grad_scale: 32.0
2024-09-23 18:13:54,169 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:14:15,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=316829.3333333333, ans=0.2
2024-09-23 18:14:19,353 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.50 vs. limit=22.5
2024-09-23 18:14:25,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer_ff2.min_abs, batch_count=316876.0, ans=0.1
2024-09-23 18:15:13,363 INFO [train.py:1198] (1/4) Epoch 18, batch 1700, loss[loss=0.2424, ctc_loss=0.1608, cr_loss=0.4078, over 16544.00 frames. ], tot_loss[loss=0.2208, ctc_loss=0.1483, cr_loss=0.3625, over 3342756.90 frames. ], batch size: 66, lr: 6.69e-03, grad_scale: 32.0
2024-09-23 18:15:13,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=317016.0, ans=0.0
2024-09-23 18:15:24,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=317016.0, ans=0.0
2024-09-23 18:15:43,210 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:16:02,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.68 vs. limit=15.0
2024-09-23 18:16:15,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=317156.0, ans=0.125
2024-09-23 18:16:24,388 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.272e+02 1.364e+02 1.549e+02 1.950e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-23 18:16:35,523 INFO [train.py:1198] (1/4) Epoch 18, batch 1750, loss[loss=0.221, ctc_loss=0.1473, cr_loss=0.3687, over 17069.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.1474, cr_loss=0.3608, over 3347664.10 frames. ], batch size: 46, lr: 6.69e-03, grad_scale: 32.0
2024-09-23 18:16:40,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=317249.3333333333, ans=0.125
2024-09-23 18:17:34,380 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.33 vs. limit=15.0
2024-09-23 18:17:58,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=317436.0, ans=0.125
2024-09-23 18:17:59,187 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.68 vs. limit=15.0
2024-09-23 18:18:02,969 INFO [train.py:1198] (1/4) Epoch 18, batch 1800, loss[loss=0.1852, ctc_loss=0.1228, cr_loss=0.3121, over 16290.00 frames. ], tot_loss[loss=0.2191, ctc_loss=0.147, cr_loss=0.3605, over 3349686.11 frames. ], batch size: 36, lr: 6.69e-03, grad_scale: 32.0
2024-09-23 18:18:03,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=317482.6666666667, ans=0.2
2024-09-23 18:18:05,363 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.51 vs. limit=6.0
2024-09-23 18:18:11,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=317482.6666666667, ans=0.0
2024-09-23 18:18:12,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=317482.6666666667, ans=0.0
2024-09-23 18:18:16,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=317482.6666666667, ans=0.125
2024-09-23 18:18:19,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=317529.3333333333, ans=0.025
2024-09-23 18:18:59,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=317622.6666666667, ans=0.2
2024-09-23 18:19:06,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=317669.3333333333, ans=0.125
2024-09-23 18:19:11,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=317669.3333333333, ans=0.125
2024-09-23 18:19:14,899 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.243e+02 1.320e+02 1.455e+02 2.029e+02, threshold=2.640e+02, percent-clipped=0.0
2024-09-23 18:19:18,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=317669.3333333333, ans=0.125
2024-09-23 18:19:26,149 INFO [train.py:1198] (1/4) Epoch 18, batch 1850, loss[loss=0.2433, ctc_loss=0.162, cr_loss=0.4064, over 17036.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.1472, cr_loss=0.3611, over 3344116.56 frames. ], batch size: 52, lr: 6.69e-03, grad_scale: 32.0
2024-09-23 18:19:33,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=317716.0, ans=0.125
2024-09-23 18:19:44,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=317762.6666666667, ans=0.2
2024-09-23 18:19:44,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=317762.6666666667, ans=0.125
2024-09-23 18:19:47,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=317762.6666666667, ans=0.0
2024-09-23 18:20:16,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=317856.0, ans=0.125
2024-09-23 18:20:24,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=317856.0, ans=0.0
2024-09-23 18:20:26,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=317856.0, ans=0.125
2024-09-23 18:20:49,117 INFO [train.py:1198] (1/4) Epoch 18, batch 1900, loss[loss=0.2595, ctc_loss=0.1813, cr_loss=0.391, over 14957.00 frames. ], tot_loss[loss=0.22, ctc_loss=0.1475, cr_loss=0.3626, over 3351904.54 frames. ], batch size: 89, lr: 6.68e-03, grad_scale: 32.0
2024-09-23 18:21:04,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=317996.0, ans=0.025
2024-09-23 18:21:13,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=317996.0, ans=0.0
2024-09-23 18:21:45,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=318089.3333333333, ans=0.125
2024-09-23 18:21:46,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.48 vs. limit=15.0
2024-09-23 18:21:58,144 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.235e+02 1.336e+02 1.423e+02 1.924e+02, threshold=2.671e+02, percent-clipped=0.0
2024-09-23 18:22:11,955 INFO [train.py:1198] (1/4) Epoch 18, batch 1950, loss[loss=0.1961, ctc_loss=0.1312, cr_loss=0.3248, over 17304.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.1467, cr_loss=0.3612, over 3357982.21 frames. ], batch size: 46, lr: 6.68e-03, grad_scale: 32.0
2024-09-23 18:22:25,433 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.65 vs. limit=15.0
2024-09-23 18:22:48,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=318276.0, ans=0.0
2024-09-23 18:23:06,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=318322.6666666667, ans=0.025
2024-09-23 18:23:34,523 INFO [train.py:1198] (1/4) Epoch 18, batch 2000, loss[loss=0.2287, ctc_loss=0.1549, cr_loss=0.369, over 17226.00 frames. ], tot_loss[loss=0.22, ctc_loss=0.1476, cr_loss=0.3619, over 3351090.05 frames. ], batch size: 50, lr: 6.68e-03, grad_scale: 32.0
2024-09-23 18:23:36,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=318416.0, ans=0.1
2024-09-23 18:23:46,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=318416.0, ans=0.0
2024-09-23 18:23:47,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=318416.0, ans=0.0
2024-09-23 18:24:10,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=318509.3333333333, ans=0.0
2024-09-23 18:24:21,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=318509.3333333333, ans=0.125
2024-09-23 18:24:46,013 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.249e+02 1.342e+02 1.462e+02 2.296e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-23 18:24:50,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=318602.6666666667, ans=0.0
2024-09-23 18:24:57,327 INFO [train.py:1198] (1/4) Epoch 18, batch 2050, loss[loss=0.1872, ctc_loss=0.1232, cr_loss=0.3201, over 17188.00 frames. ], tot_loss[loss=0.2187, ctc_loss=0.1466, cr_loss=0.3605, over 3355502.83 frames. ], batch size: 41, lr: 6.68e-03, grad_scale: 32.0
2024-09-23 18:25:03,062 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.60 vs. limit=6.0
2024-09-23 18:25:04,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=318649.3333333333, ans=0.125
2024-09-23 18:25:05,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=318649.3333333333, ans=0.125
2024-09-23 18:25:12,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=318696.0, ans=0.1
2024-09-23 18:25:16,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=318696.0, ans=0.125
2024-09-23 18:25:23,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.32 vs. limit=10.0
2024-09-23 18:25:56,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=318789.3333333333, ans=0.0
2024-09-23 18:25:58,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.44 vs. limit=15.0
2024-09-23 18:26:15,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=318836.0, ans=0.025
2024-09-23 18:26:19,581 INFO [train.py:1198] (1/4) Epoch 18, batch 2100, loss[loss=0.2217, ctc_loss=0.1474, cr_loss=0.3719, over 17173.00 frames. ], tot_loss[loss=0.2185, ctc_loss=0.1464, cr_loss=0.3602, over 3366287.52 frames. ], batch size: 45, lr: 6.67e-03, grad_scale: 32.0
2024-09-23 18:27:12,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=319022.6666666667, ans=0.125
2024-09-23 18:27:31,218 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.263e+02 1.339e+02 1.437e+02 2.082e+02, threshold=2.677e+02, percent-clipped=0.0
2024-09-23 18:27:44,929 INFO [train.py:1198] (1/4) Epoch 18, batch 2150, loss[loss=0.2039, ctc_loss=0.1346, cr_loss=0.3465, over 17314.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.1471, cr_loss=0.3621, over 3373364.25 frames. ], batch size: 49, lr: 6.67e-03, grad_scale: 32.0
2024-09-23 18:27:54,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=319116.0, ans=0.1
2024-09-23 18:27:59,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=319162.6666666667, ans=0.07
2024-09-23 18:28:36,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=319256.0, ans=0.04949747468305833
2024-09-23 18:28:38,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=319256.0, ans=0.125
2024-09-23 18:28:44,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=319256.0, ans=0.015
2024-09-23 18:29:05,237 INFO [train.py:1198] (1/4) Epoch 18, batch 2200, loss[loss=0.2457, ctc_loss=0.1677, cr_loss=0.3897, over 17027.00 frames. ], tot_loss[loss=0.2206, ctc_loss=0.1479, cr_loss=0.3636, over 3373100.77 frames. ], batch size: 52, lr: 6.67e-03, grad_scale: 16.0
2024-09-23 18:29:23,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=319396.0, ans=0.125
2024-09-23 18:29:36,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=319396.0, ans=0.125
2024-09-23 18:30:00,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=319489.3333333333, ans=0.1
2024-09-23 18:30:14,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=319536.0, ans=0.125
2024-09-23 18:30:17,667 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.244e+02 1.317e+02 1.469e+02 2.029e+02, threshold=2.633e+02, percent-clipped=0.0
2024-09-23 18:30:21,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=319536.0, ans=0.0
2024-09-23 18:30:27,418 INFO [train.py:1198] (1/4) Epoch 18, batch 2250, loss[loss=0.2448, ctc_loss=0.1652, cr_loss=0.398, over 17026.00 frames. ], tot_loss[loss=0.2211, ctc_loss=0.1482, cr_loss=0.3643, over 3367344.85 frames. ], batch size: 56, lr: 6.67e-03, grad_scale: 16.0
2024-09-23 18:30:39,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=319582.6666666667, ans=0.125
2024-09-23 18:30:54,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=319629.3333333333, ans=0.125
2024-09-23 18:31:00,144 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.22 vs. limit=22.5
2024-09-23 18:31:04,718 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.11 vs. limit=22.5
2024-09-23 18:31:17,719 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.72 vs. limit=15.0
2024-09-23 18:31:50,360 INFO [train.py:1198] (1/4) Epoch 18, batch 2300, loss[loss=0.2065, ctc_loss=0.1335, cr_loss=0.3649, over 16960.00 frames. ], tot_loss[loss=0.2209, ctc_loss=0.1481, cr_loss=0.3639, over 3366033.25 frames. ], batch size: 42, lr: 6.67e-03, grad_scale: 16.0
2024-09-23 18:32:04,811 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.30 vs. limit=15.0
2024-09-23 18:32:46,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=319956.0, ans=0.125
2024-09-23 18:32:54,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=319956.0, ans=0.0
2024-09-23 18:33:05,708 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.272e+02 1.367e+02 1.553e+02 2.225e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-23 18:33:14,708 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.35 vs. limit=15.0
2024-09-23 18:33:15,257 INFO [train.py:1198] (1/4) Epoch 18, batch 2350, loss[loss=0.2392, ctc_loss=0.1612, cr_loss=0.3902, over 17163.00 frames. ], tot_loss[loss=0.2207, ctc_loss=0.148, cr_loss=0.3638, over 3360719.12 frames. ], batch size: 48, lr: 6.66e-03, grad_scale: 16.0
2024-09-23 18:33:21,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=320049.3333333333, ans=0.125
2024-09-23 18:33:48,274 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.33 vs. limit=15.0
2024-09-23 18:33:49,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=320142.6666666667, ans=0.2
2024-09-23 18:33:57,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=320142.6666666667, ans=0.025
2024-09-23 18:34:23,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=320236.0, ans=0.125
2024-09-23 18:34:23,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=320236.0, ans=0.0
2024-09-23 18:34:33,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=320236.0, ans=0.2
2024-09-23 18:34:33,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=320236.0, ans=0.125
2024-09-23 18:34:37,843 INFO [train.py:1198] (1/4) Epoch 18, batch 2400, loss[loss=0.1891, ctc_loss=0.1245, cr_loss=0.3233, over 17040.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.147, cr_loss=0.3625, over 3363179.87 frames. ], batch size: 39, lr: 6.66e-03, grad_scale: 32.0
2024-09-23 18:34:46,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=320282.6666666667, ans=0.0
2024-09-23 18:35:13,485 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:35:24,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=320422.6666666667, ans=0.0
2024-09-23 18:35:50,936 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.273e+02 1.349e+02 1.474e+02 2.344e+02, threshold=2.697e+02, percent-clipped=0.0
2024-09-23 18:35:56,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=320469.3333333333, ans=0.125
2024-09-23 18:36:00,512 INFO [train.py:1198] (1/4) Epoch 18, batch 2450, loss[loss=0.1982, ctc_loss=0.1309, cr_loss=0.3362, over 16740.00 frames. ], tot_loss[loss=0.2196, ctc_loss=0.1472, cr_loss=0.3619, over 3357777.79 frames. ], batch size: 37, lr: 6.66e-03, grad_scale: 32.0
2024-09-23 18:36:29,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=320562.6666666667, ans=0.1
2024-09-23 18:36:54,203 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=7.89 vs. limit=15.0
2024-09-23 18:37:09,370 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.60 vs. limit=10.0
2024-09-23 18:37:22,755 INFO [train.py:1198] (1/4) Epoch 18, batch 2500, loss[loss=0.2219, ctc_loss=0.147, cr_loss=0.3747, over 16996.00 frames. ], tot_loss[loss=0.2202, ctc_loss=0.1477, cr_loss=0.3623, over 3342965.25 frames. ], batch size: 56, lr: 6.66e-03, grad_scale: 32.0
2024-09-23 18:37:23,222 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:37:56,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=320842.6666666667, ans=0.125
2024-09-23 18:37:59,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=320842.6666666667, ans=0.125
2024-09-23 18:38:28,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=320936.0, ans=0.125
2024-09-23 18:38:31,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=320936.0, ans=0.1
2024-09-23 18:38:37,300 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.301e+02 1.381e+02 1.501e+02 2.403e+02, threshold=2.762e+02, percent-clipped=0.0
2024-09-23 18:38:45,213 INFO [train.py:1198] (1/4) Epoch 18, batch 2550, loss[loss=0.2167, ctc_loss=0.1447, cr_loss=0.36, over 17293.00 frames. ], tot_loss[loss=0.2198, ctc_loss=0.1473, cr_loss=0.3622, over 3353930.46 frames. ], batch size: 51, lr: 6.65e-03, grad_scale: 16.0
2024-09-23 18:39:04,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=321029.3333333333, ans=0.125
2024-09-23 18:39:16,585 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=8.80 vs. limit=15.0
2024-09-23 18:39:36,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=321122.6666666667, ans=0.0
2024-09-23 18:39:41,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=321122.6666666667, ans=0.125
2024-09-23 18:39:44,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=321122.6666666667, ans=0.125
2024-09-23 18:40:07,839 INFO [train.py:1198] (1/4) Epoch 18, batch 2600, loss[loss=0.1886, ctc_loss=0.1235, cr_loss=0.3258, over 17129.00 frames. ], tot_loss[loss=0.2193, ctc_loss=0.1469, cr_loss=0.3617, over 3353577.51 frames. ], batch size: 40, lr: 6.65e-03, grad_scale: 16.0
2024-09-23 18:40:08,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=321216.0, ans=0.025
2024-09-23 18:40:49,329 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.50 vs. limit=15.0
2024-09-23 18:41:13,564 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.88 vs. limit=15.0
2024-09-23 18:41:17,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=321402.6666666667, ans=0.0
2024-09-23 18:41:17,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=321402.6666666667, ans=0.0
2024-09-23 18:41:22,320 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.312e+02 1.421e+02 1.594e+02 2.386e+02, threshold=2.841e+02, percent-clipped=0.0
2024-09-23 18:41:30,434 INFO [train.py:1198] (1/4) Epoch 18, batch 2650, loss[loss=0.2053, ctc_loss=0.136, cr_loss=0.3465, over 17021.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.1467, cr_loss=0.3614, over 3361661.70 frames. ], batch size: 44, lr: 6.65e-03, grad_scale: 16.0
2024-09-23 18:42:01,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=321496.0, ans=0.0
2024-09-23 18:42:49,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=321636.0, ans=0.035
2024-09-23 18:42:54,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=321682.6666666667, ans=0.1
2024-09-23 18:42:55,427 INFO [train.py:1198] (1/4) Epoch 18, batch 2700, loss[loss=0.2478, ctc_loss=0.1683, cr_loss=0.3973, over 17036.00 frames. ], tot_loss[loss=0.2185, ctc_loss=0.1464, cr_loss=0.3609, over 3364618.95 frames. ], batch size: 52, lr: 6.65e-03, grad_scale: 16.0
2024-09-23 18:42:55,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=321682.6666666667, ans=0.125
2024-09-23 18:43:16,669 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.78 vs. limit=15.0
2024-09-23 18:43:37,273 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:44:09,951 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.267e+02 1.364e+02 1.525e+02 2.538e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-23 18:44:15,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=321869.3333333333, ans=0.09899494936611666
2024-09-23 18:44:17,772 INFO [train.py:1198] (1/4) Epoch 18, batch 2750, loss[loss=0.24, ctc_loss=0.1633, cr_loss=0.3836, over 16781.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.1472, cr_loss=0.3615, over 3349990.53 frames. ], batch size: 61, lr: 6.64e-03, grad_scale: 16.0
2024-09-23 18:44:22,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=321916.0, ans=0.05
2024-09-23 18:45:15,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=322056.0, ans=0.0
2024-09-23 18:45:20,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=322102.6666666667, ans=0.125
2024-09-23 18:45:36,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=322102.6666666667, ans=0.2
2024-09-23 18:45:40,662 INFO [train.py:1198] (1/4) Epoch 18, batch 2800, loss[loss=0.2174, ctc_loss=0.1454, cr_loss=0.3601, over 17190.00 frames. ], tot_loss[loss=0.2184, ctc_loss=0.1463, cr_loss=0.3604, over 3352142.11 frames. ], batch size: 41, lr: 6.64e-03, grad_scale: 32.0
2024-09-23 18:46:22,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=322242.6666666667, ans=0.0
2024-09-23 18:46:25,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=322242.6666666667, ans=0.0
2024-09-23 18:46:39,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=322289.3333333333, ans=0.125
2024-09-23 18:46:55,162 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.235e+02 1.345e+02 1.484e+02 2.490e+02, threshold=2.689e+02, percent-clipped=0.0
2024-09-23 18:47:00,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=322336.0, ans=0.0
2024-09-23 18:47:03,239 INFO [train.py:1198] (1/4) Epoch 18, batch 2850, loss[loss=0.2136, ctc_loss=0.1425, cr_loss=0.3555, over 17137.00 frames. ], tot_loss[loss=0.2193, ctc_loss=0.147, cr_loss=0.3617, over 3358737.98 frames. ], batch size: 48, lr: 6.64e-03, grad_scale: 32.0
2024-09-23 18:47:11,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=322382.6666666667, ans=0.025
2024-09-23 18:47:35,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=322429.3333333333, ans=0.1
2024-09-23 18:48:03,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=322522.6666666667, ans=0.125
2024-09-23 18:48:26,237 INFO [train.py:1198] (1/4) Epoch 18, batch 2900, loss[loss=0.2159, ctc_loss=0.144, cr_loss=0.3596, over 16962.00 frames. ], tot_loss[loss=0.2189, ctc_loss=0.1465, cr_loss=0.3616, over 3372691.47 frames. ], batch size: 58, lr: 6.64e-03, grad_scale: 32.0
2024-09-23 18:48:34,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=322616.0, ans=0.125
2024-09-23 18:48:38,302 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.85 vs. limit=22.5
2024-09-23 18:48:56,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=322662.6666666667, ans=0.1
2024-09-23 18:49:00,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=322709.3333333333, ans=0.125
2024-09-23 18:49:06,563 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=29.00 vs. limit=22.5
2024-09-23 18:49:08,013 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.77 vs. limit=12.0
2024-09-23 18:49:09,384 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=6.64 vs. limit=15.0
2024-09-23 18:49:13,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=322709.3333333333, ans=0.125
2024-09-23 18:49:29,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=322756.0, ans=0.0
2024-09-23 18:49:40,536 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.288e+02 1.376e+02 1.532e+02 2.384e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-23 18:49:48,607 INFO [train.py:1198] (1/4) Epoch 18, batch 2950, loss[loss=0.2087, ctc_loss=0.1374, cr_loss=0.3567, over 17236.00 frames. ], tot_loss[loss=0.2201, ctc_loss=0.1475, cr_loss=0.3629, over 3363723.15 frames. ], batch size: 50, lr: 6.63e-03, grad_scale: 32.0
2024-09-23 18:49:58,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=322849.3333333333, ans=0.125
2024-09-23 18:50:13,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=322896.0, ans=0.0
2024-09-23 18:50:32,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=322942.6666666667, ans=0.125
2024-09-23 18:50:32,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.75 vs. limit=15.0
2024-09-23 18:50:43,300 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:51:01,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.29 vs. limit=15.0
2024-09-23 18:51:07,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=323036.0, ans=0.0
2024-09-23 18:51:08,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=323036.0, ans=0.0
2024-09-23 18:51:11,407 INFO [train.py:1198] (1/4) Epoch 18, batch 3000, loss[loss=0.2467, ctc_loss=0.1719, cr_loss=0.3736, over 14944.00 frames. ], tot_loss[loss=0.2198, ctc_loss=0.1473, cr_loss=0.3627, over 3363467.00 frames. ], batch size: 89, lr: 6.63e-03, grad_scale: 32.0
2024-09-23 18:51:11,408 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 18:51:24,464 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.2.self_attn_weights, attn_weights_entropy = tensor([1.9762, 2.3204, 2.3522, 2.3997, 2.2226, 2.1761, 2.4744, 2.3837],
       device='cuda:1')
2024-09-23 18:51:26,835 INFO [train.py:1230] (1/4) Epoch 18, validation: loss=0.04062, ctc_loss=0.04062, cr_loss=7.511e-15, over 944034.00 frames. 
2024-09-23 18:51:26,836 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 18:51:53,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=323129.3333333333, ans=0.125
2024-09-23 18:52:18,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=323222.6666666667, ans=0.0
2024-09-23 18:52:34,478 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.67 vs. limit=15.0
2024-09-23 18:52:36,009 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.25 vs. limit=22.5
2024-09-23 18:52:39,991 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.237e+02 1.313e+02 1.391e+02 1.785e+02, threshold=2.627e+02, percent-clipped=0.0
2024-09-23 18:52:47,793 INFO [train.py:1198] (1/4) Epoch 18, batch 3050, loss[loss=0.2638, ctc_loss=0.1889, cr_loss=0.3745, over 11588.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.147, cr_loss=0.362, over 3361455.99 frames. ], batch size: 123, lr: 6.63e-03, grad_scale: 32.0
2024-09-23 18:52:54,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=323316.0, ans=0.0
2024-09-23 18:52:54,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=323316.0, ans=22.5
2024-09-23 18:53:16,150 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=15.10 vs. limit=15.0
2024-09-23 18:53:17,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=323362.6666666667, ans=0.2
2024-09-23 18:53:17,306 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:53:43,835 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 18:53:48,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=323456.0, ans=0.125
2024-09-23 18:54:08,393 INFO [train.py:1198] (1/4) Epoch 18, batch 3100, loss[loss=0.2268, ctc_loss=0.1511, cr_loss=0.3788, over 17017.00 frames. ], tot_loss[loss=0.2199, ctc_loss=0.1474, cr_loss=0.3627, over 3360644.44 frames. ], batch size: 56, lr: 6.63e-03, grad_scale: 32.0
2024-09-23 18:54:44,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=323642.6666666667, ans=0.5
2024-09-23 18:54:53,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=323689.3333333333, ans=0.0
2024-09-23 18:55:00,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=323689.3333333333, ans=0.2
2024-09-23 18:55:10,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=323736.0, ans=0.125
2024-09-23 18:55:13,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=323736.0, ans=0.0
2024-09-23 18:55:19,121 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.021e+02 1.256e+02 1.342e+02 1.439e+02 3.475e+02, threshold=2.684e+02, percent-clipped=1.0
2024-09-23 18:55:26,828 INFO [train.py:1198] (1/4) Epoch 18, batch 3150, loss[loss=0.2487, ctc_loss=0.1678, cr_loss=0.4047, over 16694.00 frames. ], tot_loss[loss=0.2188, ctc_loss=0.1465, cr_loss=0.3615, over 3372236.50 frames. ], batch size: 61, lr: 6.62e-03, grad_scale: 32.0
2024-09-23 18:55:30,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=323782.6666666667, ans=0.125
2024-09-23 18:55:39,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=323782.6666666667, ans=0.125
2024-09-23 18:55:45,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=323829.3333333333, ans=0.1
2024-09-23 18:56:16,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=323922.6666666667, ans=0.125
2024-09-23 18:56:39,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=323969.3333333333, ans=0.1
2024-09-23 18:56:41,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.25 vs. limit=15.0
2024-09-23 18:56:45,843 INFO [train.py:1198] (1/4) Epoch 18, batch 3200, loss[loss=0.1941, ctc_loss=0.1275, cr_loss=0.3328, over 17105.00 frames. ], tot_loss[loss=0.218, ctc_loss=0.1459, cr_loss=0.3603, over 3374326.97 frames. ], batch size: 49, lr: 6.62e-03, grad_scale: 32.0
2024-09-23 18:57:04,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=324062.6666666667, ans=0.125
2024-09-23 18:57:11,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=324062.6666666667, ans=0.0
2024-09-23 18:57:14,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.55 vs. limit=15.0
2024-09-23 18:57:22,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=324109.3333333333, ans=0.0
2024-09-23 18:57:27,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=324109.3333333333, ans=0.1
2024-09-23 18:57:50,167 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.51 vs. limit=6.0
2024-09-23 18:57:51,502 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.72 vs. limit=15.0
2024-09-23 18:57:54,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=324202.6666666667, ans=0.125
2024-09-23 18:57:58,435 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.249e+02 1.354e+02 1.459e+02 2.306e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-23 18:58:06,267 INFO [train.py:1198] (1/4) Epoch 18, batch 3250, loss[loss=0.2194, ctc_loss=0.146, cr_loss=0.3669, over 17068.00 frames. ], tot_loss[loss=0.2199, ctc_loss=0.1476, cr_loss=0.3617, over 3348664.61 frames. ], batch size: 46, lr: 6.62e-03, grad_scale: 32.0
2024-09-23 18:58:31,094 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=10.93 vs. limit=22.5
2024-09-23 18:58:42,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=324342.6666666667, ans=0.125
2024-09-23 18:59:02,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=324389.3333333333, ans=0.025
2024-09-23 18:59:24,401 INFO [train.py:1198] (1/4) Epoch 18, batch 3300, loss[loss=0.223, ctc_loss=0.151, cr_loss=0.3598, over 17013.00 frames. ], tot_loss[loss=0.2201, ctc_loss=0.1476, cr_loss=0.3621, over 3352198.57 frames. ], batch size: 51, lr: 6.62e-03, grad_scale: 32.0
2024-09-23 18:59:44,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=324529.3333333333, ans=0.125
2024-09-23 19:00:30,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=324669.3333333333, ans=0.125
2024-09-23 19:00:30,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=324669.3333333333, ans=0.2
2024-09-23 19:00:36,180 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.318e+02 1.406e+02 1.543e+02 2.189e+02, threshold=2.811e+02, percent-clipped=0.0
2024-09-23 19:00:37,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=324669.3333333333, ans=0.2
2024-09-23 19:00:39,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=324669.3333333333, ans=0.125
2024-09-23 19:00:44,001 INFO [train.py:1198] (1/4) Epoch 18, batch 3350, loss[loss=0.2449, ctc_loss=0.1667, cr_loss=0.3913, over 15899.00 frames. ], tot_loss[loss=0.2204, ctc_loss=0.1478, cr_loss=0.3629, over 3351841.03 frames. ], batch size: 74, lr: 6.62e-03, grad_scale: 32.0
2024-09-23 19:01:01,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer_ff3.min_abs, batch_count=324762.6666666667, ans=0.2
2024-09-23 19:02:04,709 INFO [train.py:1198] (1/4) Epoch 18, batch 3400, loss[loss=0.2151, ctc_loss=0.1453, cr_loss=0.349, over 17012.00 frames. ], tot_loss[loss=0.2211, ctc_loss=0.1483, cr_loss=0.3642, over 3357980.97 frames. ], batch size: 44, lr: 6.61e-03, grad_scale: 32.0
2024-09-23 19:02:05,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=324949.3333333333, ans=0.025
2024-09-23 19:02:09,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=324949.3333333333, ans=0.0
2024-09-23 19:02:12,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=324949.3333333333, ans=0.125
2024-09-23 19:03:14,634 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.235e+02 1.326e+02 1.505e+02 2.372e+02, threshold=2.653e+02, percent-clipped=0.0
2024-09-23 19:03:22,389 INFO [train.py:1198] (1/4) Epoch 18, batch 3450, loss[loss=0.2263, ctc_loss=0.1503, cr_loss=0.3801, over 17219.00 frames. ], tot_loss[loss=0.2203, ctc_loss=0.1476, cr_loss=0.3636, over 3359579.51 frames. ], batch size: 55, lr: 6.61e-03, grad_scale: 32.0
2024-09-23 19:03:29,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=325182.6666666667, ans=0.125
2024-09-23 19:03:30,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=325182.6666666667, ans=0.0
2024-09-23 19:03:48,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=325229.3333333333, ans=0.2
2024-09-23 19:03:52,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=325229.3333333333, ans=0.125
2024-09-23 19:04:02,848 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.93 vs. limit=15.0
2024-09-23 19:04:07,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=325276.0, ans=0.0
2024-09-23 19:04:24,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=325322.6666666667, ans=0.0
2024-09-23 19:04:38,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=325369.3333333333, ans=0.125
2024-09-23 19:04:41,725 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=5.57 vs. limit=12.0
2024-09-23 19:04:42,581 INFO [train.py:1198] (1/4) Epoch 18, batch 3500, loss[loss=0.2419, ctc_loss=0.1643, cr_loss=0.3882, over 17044.00 frames. ], tot_loss[loss=0.2203, ctc_loss=0.1476, cr_loss=0.3637, over 3361752.45 frames. ], batch size: 56, lr: 6.61e-03, grad_scale: 32.0
2024-09-23 19:04:49,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=325416.0, ans=0.0
2024-09-23 19:05:01,311 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.65 vs. limit=5.0
2024-09-23 19:05:03,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=325462.6666666667, ans=0.0
2024-09-23 19:05:22,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=325509.3333333333, ans=0.125
2024-09-23 19:05:42,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=325556.0, ans=0.125
2024-09-23 19:05:53,067 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.293e+02 1.389e+02 1.509e+02 2.092e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-23 19:05:56,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=325602.6666666667, ans=0.125
2024-09-23 19:06:00,905 INFO [train.py:1198] (1/4) Epoch 18, batch 3550, loss[loss=0.2087, ctc_loss=0.1393, cr_loss=0.3469, over 17202.00 frames. ], tot_loss[loss=0.2201, ctc_loss=0.1473, cr_loss=0.3639, over 3357227.54 frames. ], batch size: 41, lr: 6.61e-03, grad_scale: 32.0
2024-09-23 19:06:17,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=325696.0, ans=0.125
2024-09-23 19:06:36,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=325742.6666666667, ans=0.125
2024-09-23 19:06:40,455 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.05 vs. limit=15.0
2024-09-23 19:06:41,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=325742.6666666667, ans=0.125
2024-09-23 19:07:03,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=325789.3333333333, ans=0.125
2024-09-23 19:07:09,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=325836.0, ans=0.2
2024-09-23 19:07:17,734 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.66 vs. limit=15.0
2024-09-23 19:07:21,663 INFO [train.py:1198] (1/4) Epoch 18, batch 3600, loss[loss=0.2124, ctc_loss=0.1407, cr_loss=0.3583, over 17259.00 frames. ], tot_loss[loss=0.2206, ctc_loss=0.1477, cr_loss=0.3645, over 3351270.82 frames. ], batch size: 42, lr: 6.60e-03, grad_scale: 32.0
2024-09-23 19:07:40,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=325929.3333333333, ans=0.2
2024-09-23 19:07:46,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=325929.3333333333, ans=0.125
2024-09-23 19:08:31,458 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.271e+02 1.404e+02 1.560e+02 2.337e+02, threshold=2.809e+02, percent-clipped=0.0
2024-09-23 19:08:39,251 INFO [train.py:1198] (1/4) Epoch 18, batch 3650, loss[loss=0.2138, ctc_loss=0.1411, cr_loss=0.3633, over 17153.00 frames. ], tot_loss[loss=0.2199, ctc_loss=0.1471, cr_loss=0.3641, over 3355044.47 frames. ], batch size: 48, lr: 6.60e-03, grad_scale: 32.0
2024-09-23 19:08:53,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=326162.6666666667, ans=0.125
2024-09-23 19:09:11,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=326209.3333333333, ans=0.05
2024-09-23 19:09:11,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=326209.3333333333, ans=0.0
2024-09-23 19:09:20,081 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.06 vs. limit=12.0
2024-09-23 19:09:20,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=326209.3333333333, ans=0.125
2024-09-23 19:09:39,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=326256.0, ans=0.0
2024-09-23 19:09:54,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=326302.6666666667, ans=0.125
2024-09-23 19:10:00,632 INFO [train.py:1198] (1/4) Epoch 18, batch 3700, loss[loss=0.2172, ctc_loss=0.146, cr_loss=0.356, over 17133.00 frames. ], tot_loss[loss=0.2209, ctc_loss=0.1479, cr_loss=0.365, over 3354654.85 frames. ], batch size: 48, lr: 6.60e-03, grad_scale: 32.0
2024-09-23 19:10:06,039 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.88 vs. limit=10.0
2024-09-23 19:11:01,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=326536.0, ans=0.125
2024-09-23 19:11:03,706 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.56 vs. limit=10.0
2024-09-23 19:11:10,729 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.270e+02 1.376e+02 1.485e+02 2.172e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-23 19:11:12,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=326536.0, ans=0.025
2024-09-23 19:11:18,531 INFO [train.py:1198] (1/4) Epoch 18, batch 3750, loss[loss=0.1939, ctc_loss=0.1261, cr_loss=0.339, over 17171.00 frames. ], tot_loss[loss=0.222, ctc_loss=0.1489, cr_loss=0.3653, over 3333272.12 frames. ], batch size: 45, lr: 6.60e-03, grad_scale: 32.0
2024-09-23 19:11:21,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=326582.6666666667, ans=0.125
2024-09-23 19:11:26,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=326582.6666666667, ans=0.0
2024-09-23 19:11:50,407 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.46 vs. limit=15.0
2024-09-23 19:12:16,821 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.80 vs. limit=15.0
2024-09-23 19:12:17,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=326722.6666666667, ans=0.125
2024-09-23 19:12:22,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=326769.3333333333, ans=0.07
2024-09-23 19:12:37,968 INFO [train.py:1198] (1/4) Epoch 18, batch 3800, loss[loss=0.1886, ctc_loss=0.1217, cr_loss=0.3343, over 16940.00 frames. ], tot_loss[loss=0.2209, ctc_loss=0.1482, cr_loss=0.3635, over 3321886.80 frames. ], batch size: 42, lr: 6.59e-03, grad_scale: 32.0
2024-09-23 19:12:52,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=326862.6666666667, ans=0.05
2024-09-23 19:13:17,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=326909.3333333333, ans=0.0
2024-09-23 19:13:24,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=326956.0, ans=0.1
2024-09-23 19:13:36,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=326956.0, ans=0.1
2024-09-23 19:13:47,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=327002.6666666667, ans=0.1
2024-09-23 19:13:49,015 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.286e+02 1.423e+02 1.561e+02 2.075e+02, threshold=2.847e+02, percent-clipped=0.0
2024-09-23 19:13:56,833 INFO [train.py:1198] (1/4) Epoch 18, batch 3850, loss[loss=0.2052, ctc_loss=0.1357, cr_loss=0.3473, over 16997.00 frames. ], tot_loss[loss=0.2212, ctc_loss=0.1487, cr_loss=0.3625, over 3279307.37 frames. ], batch size: 39, lr: 6.59e-03, grad_scale: 32.0
2024-09-23 19:14:12,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=327096.0, ans=0.0
2024-09-23 19:14:28,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.min_positive, batch_count=327142.6666666667, ans=0.025
2024-09-23 19:15:59,117 INFO [train.py:1198] (1/4) Epoch 19, batch 0, loss[loss=0.2769, ctc_loss=0.1896, cr_loss=0.4366, over 16465.00 frames. ], tot_loss[loss=0.2769, ctc_loss=0.1896, cr_loss=0.4366, over 16465.00 frames. ], batch size: 66, lr: 6.41e-03, grad_scale: 32.0
2024-09-23 19:15:59,118 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 19:16:14,371 INFO [train.py:1230] (1/4) Epoch 19, validation: loss=0.03972, ctc_loss=0.03972, cr_loss=8.025e-15, over 944034.00 frames. 
2024-09-23 19:16:14,372 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 19:16:16,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=327264.0, ans=0.5
2024-09-23 19:16:19,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=327264.0, ans=0.0
2024-09-23 19:16:40,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=327310.6666666667, ans=0.5
2024-09-23 19:17:02,156 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:17:27,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=327450.6666666667, ans=0.1
2024-09-23 19:17:36,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=327450.6666666667, ans=0.125
2024-09-23 19:17:38,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=327450.6666666667, ans=0.2
2024-09-23 19:17:38,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=327450.6666666667, ans=0.025
2024-09-23 19:17:39,622 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.290e+02 1.473e+02 1.789e+02 2.583e+02, threshold=2.945e+02, percent-clipped=0.0
2024-09-23 19:17:41,331 INFO [train.py:1198] (1/4) Epoch 19, batch 50, loss[loss=0.2254, ctc_loss=0.1541, cr_loss=0.3562, over 17097.00 frames. ], tot_loss[loss=0.2198, ctc_loss=0.147, cr_loss=0.3639, over 760900.41 frames. ], batch size: 49, lr: 6.41e-03, grad_scale: 32.0
2024-09-23 19:17:49,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=327497.3333333333, ans=0.0
2024-09-23 19:18:05,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=327544.0, ans=0.125
2024-09-23 19:18:10,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=327544.0, ans=0.025
2024-09-23 19:18:12,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=327590.6666666667, ans=0.0
2024-09-23 19:18:36,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=327637.3333333333, ans=0.125
2024-09-23 19:19:03,588 INFO [train.py:1198] (1/4) Epoch 19, batch 100, loss[loss=0.1751, ctc_loss=0.1144, cr_loss=0.3035, over 16329.00 frames. ], tot_loss[loss=0.2164, ctc_loss=0.1445, cr_loss=0.3595, over 1340508.22 frames. ], batch size: 36, lr: 6.41e-03, grad_scale: 32.0
2024-09-23 19:19:06,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=327730.6666666667, ans=0.09899494936611666
2024-09-23 19:19:21,292 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:19:22,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=327777.3333333333, ans=0.1
2024-09-23 19:19:24,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=327777.3333333333, ans=0.125
2024-09-23 19:19:24,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=327777.3333333333, ans=0.2
2024-09-23 19:19:47,396 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.03 vs. limit=6.0
2024-09-23 19:20:13,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=327917.3333333333, ans=0.125
2024-09-23 19:20:21,266 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.272e+02 1.373e+02 1.513e+02 2.086e+02, threshold=2.746e+02, percent-clipped=0.0
2024-09-23 19:20:22,822 INFO [train.py:1198] (1/4) Epoch 19, batch 150, loss[loss=0.1711, ctc_loss=0.1119, cr_loss=0.2959, over 17119.00 frames. ], tot_loss[loss=0.2181, ctc_loss=0.1458, cr_loss=0.3614, over 1793777.89 frames. ], batch size: 40, lr: 6.40e-03, grad_scale: 32.0
2024-09-23 19:20:34,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=327964.0, ans=0.0
2024-09-23 19:20:39,168 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.27 vs. limit=15.0
2024-09-23 19:20:40,512 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:20:46,080 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.08 vs. limit=12.0
2024-09-23 19:20:46,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=328010.6666666667, ans=0.2
2024-09-23 19:20:52,519 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=15.43 vs. limit=22.5
2024-09-23 19:20:55,458 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.37 vs. limit=10.0
2024-09-23 19:21:01,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=328057.3333333333, ans=0.125
2024-09-23 19:21:19,461 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.41 vs. limit=12.0
2024-09-23 19:21:30,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=328150.6666666667, ans=0.125
2024-09-23 19:21:33,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=328150.6666666667, ans=0.2
2024-09-23 19:21:34,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=328150.6666666667, ans=0.2
2024-09-23 19:21:40,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=328150.6666666667, ans=0.1
2024-09-23 19:21:42,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=328150.6666666667, ans=0.1
2024-09-23 19:21:48,166 INFO [train.py:1198] (1/4) Epoch 19, batch 200, loss[loss=0.2412, ctc_loss=0.1606, cr_loss=0.4034, over 16730.00 frames. ], tot_loss[loss=0.2182, ctc_loss=0.1458, cr_loss=0.3623, over 2145954.77 frames. ], batch size: 61, lr: 6.40e-03, grad_scale: 32.0
2024-09-23 19:21:56,450 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:22:06,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=328244.0, ans=0.125
2024-09-23 19:22:17,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=328244.0, ans=0.0
2024-09-23 19:22:47,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=328337.3333333333, ans=0.07
2024-09-23 19:23:01,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=328384.0, ans=0.125
2024-09-23 19:23:09,156 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.256e+02 1.366e+02 1.552e+02 2.193e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-23 19:23:10,719 INFO [train.py:1198] (1/4) Epoch 19, batch 250, loss[loss=0.186, ctc_loss=0.1257, cr_loss=0.3018, over 17078.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.1464, cr_loss=0.3629, over 2412546.46 frames. ], batch size: 43, lr: 6.40e-03, grad_scale: 32.0
2024-09-23 19:23:20,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=328430.6666666667, ans=0.025
2024-09-23 19:23:57,063 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.35 vs. limit=15.0
2024-09-23 19:24:09,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=328570.6666666667, ans=0.125
2024-09-23 19:24:09,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=328570.6666666667, ans=0.125
2024-09-23 19:24:18,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=328617.3333333333, ans=0.0
2024-09-23 19:24:32,610 INFO [train.py:1198] (1/4) Epoch 19, batch 300, loss[loss=0.1885, ctc_loss=0.1244, cr_loss=0.3204, over 17075.00 frames. ], tot_loss[loss=0.2179, ctc_loss=0.1456, cr_loss=0.3616, over 2627084.95 frames. ], batch size: 43, lr: 6.40e-03, grad_scale: 32.0
2024-09-23 19:25:16,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=328757.3333333333, ans=0.2
2024-09-23 19:25:35,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=328850.6666666667, ans=0.0
2024-09-23 19:25:37,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1.whitening_limit, batch_count=328850.6666666667, ans=10.0
2024-09-23 19:25:46,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=328850.6666666667, ans=0.125
2024-09-23 19:25:51,166 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.306e+02 1.376e+02 1.560e+02 2.317e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-23 19:25:52,831 INFO [train.py:1198] (1/4) Epoch 19, batch 350, loss[loss=0.2091, ctc_loss=0.1382, cr_loss=0.3542, over 17296.00 frames. ], tot_loss[loss=0.2204, ctc_loss=0.1475, cr_loss=0.3647, over 2780119.34 frames. ], batch size: 49, lr: 6.40e-03, grad_scale: 32.0
2024-09-23 19:26:07,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=328944.0, ans=0.125
2024-09-23 19:26:16,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=328944.0, ans=0.1
2024-09-23 19:26:24,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=328990.6666666667, ans=0.125
2024-09-23 19:26:44,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=329037.3333333333, ans=0.07
2024-09-23 19:27:08,141 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:27:16,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=329130.6666666667, ans=0.1
2024-09-23 19:27:17,482 INFO [train.py:1198] (1/4) Epoch 19, batch 400, loss[loss=0.2036, ctc_loss=0.1347, cr_loss=0.3445, over 17084.00 frames. ], tot_loss[loss=0.2198, ctc_loss=0.1469, cr_loss=0.3642, over 2914666.02 frames. ], batch size: 49, lr: 6.39e-03, grad_scale: 32.0
2024-09-23 19:27:35,436 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.10 vs. limit=10.0
2024-09-23 19:27:52,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=329224.0, ans=0.1
2024-09-23 19:27:52,828 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.19 vs. limit=22.5
2024-09-23 19:28:05,975 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=6.95 vs. limit=15.0
2024-09-23 19:28:32,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=329317.3333333333, ans=0.025
2024-09-23 19:28:41,423 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.241e+02 1.311e+02 1.431e+02 1.671e+02, threshold=2.622e+02, percent-clipped=0.0
2024-09-23 19:28:41,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=329364.0, ans=0.1
2024-09-23 19:28:43,032 INFO [train.py:1198] (1/4) Epoch 19, batch 450, loss[loss=0.2149, ctc_loss=0.1442, cr_loss=0.3534, over 17339.00 frames. ], tot_loss[loss=0.2202, ctc_loss=0.1472, cr_loss=0.3648, over 3001940.47 frames. ], batch size: 48, lr: 6.39e-03, grad_scale: 32.0
2024-09-23 19:28:47,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=329364.0, ans=0.1
2024-09-23 19:28:52,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=329364.0, ans=0.125
2024-09-23 19:29:03,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=329410.6666666667, ans=0.2
2024-09-23 19:29:03,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=329410.6666666667, ans=0.0
2024-09-23 19:29:36,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.85 vs. limit=22.5
2024-09-23 19:29:47,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=329550.6666666667, ans=0.125
2024-09-23 19:29:51,190 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.35 vs. limit=15.0
2024-09-23 19:29:52,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=329550.6666666667, ans=0.2
2024-09-23 19:30:02,506 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.95 vs. limit=10.0
2024-09-23 19:30:03,311 INFO [train.py:1198] (1/4) Epoch 19, batch 500, loss[loss=0.2243, ctc_loss=0.1491, cr_loss=0.376, over 16898.00 frames. ], tot_loss[loss=0.2204, ctc_loss=0.1475, cr_loss=0.3647, over 3084189.64 frames. ], batch size: 58, lr: 6.39e-03, grad_scale: 32.0
2024-09-23 19:30:05,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.max_positive, batch_count=329597.3333333333, ans=0.95
2024-09-23 19:30:13,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=329597.3333333333, ans=0.1
2024-09-23 19:30:23,459 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.51 vs. limit=6.0
2024-09-23 19:30:43,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=329690.6666666667, ans=0.0
2024-09-23 19:30:45,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=329690.6666666667, ans=0.125
2024-09-23 19:31:21,602 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.237e+02 1.337e+02 1.500e+02 1.948e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-23 19:31:22,441 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.79 vs. limit=12.0
2024-09-23 19:31:23,262 INFO [train.py:1198] (1/4) Epoch 19, batch 550, loss[loss=0.2124, ctc_loss=0.141, cr_loss=0.3571, over 17013.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.1463, cr_loss=0.3634, over 3151436.54 frames. ], batch size: 39, lr: 6.39e-03, grad_scale: 32.0
2024-09-23 19:32:37,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=330017.3333333333, ans=0.1
2024-09-23 19:32:46,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=330017.3333333333, ans=0.125
2024-09-23 19:32:51,189 INFO [train.py:1198] (1/4) Epoch 19, batch 600, loss[loss=0.2101, ctc_loss=0.1397, cr_loss=0.3518, over 17305.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.1467, cr_loss=0.3637, over 3198873.02 frames. ], batch size: 46, lr: 6.38e-03, grad_scale: 32.0
2024-09-23 19:33:14,128 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.37 vs. limit=15.0
2024-09-23 19:33:26,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=330157.3333333333, ans=0.0
2024-09-23 19:33:32,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=330157.3333333333, ans=0.125
2024-09-23 19:33:35,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=330157.3333333333, ans=0.1
2024-09-23 19:33:53,867 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:33:53,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=330204.0, ans=0.1
2024-09-23 19:33:58,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=330250.6666666667, ans=0.2
2024-09-23 19:34:12,589 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.274e+02 1.356e+02 1.568e+02 2.511e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-23 19:34:14,267 INFO [train.py:1198] (1/4) Epoch 19, batch 650, loss[loss=0.1859, ctc_loss=0.1234, cr_loss=0.3126, over 16708.00 frames. ], tot_loss[loss=0.218, ctc_loss=0.1456, cr_loss=0.362, over 3234490.69 frames. ], batch size: 37, lr: 6.38e-03, grad_scale: 64.0
2024-09-23 19:34:16,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=330297.3333333333, ans=0.125
2024-09-23 19:35:18,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=330484.0, ans=0.2
2024-09-23 19:35:22,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=330484.0, ans=0.1
2024-09-23 19:35:25,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=330484.0, ans=0.125
2024-09-23 19:35:34,529 INFO [train.py:1198] (1/4) Epoch 19, batch 700, loss[loss=0.2401, ctc_loss=0.159, cr_loss=0.4058, over 16614.00 frames. ], tot_loss[loss=0.2178, ctc_loss=0.1455, cr_loss=0.3617, over 3271004.04 frames. ], batch size: 66, lr: 6.38e-03, grad_scale: 64.0
2024-09-23 19:35:46,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=330530.6666666667, ans=0.125
2024-09-23 19:36:18,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=330624.0, ans=0.0
2024-09-23 19:36:24,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=330670.6666666667, ans=0.0
2024-09-23 19:36:58,032 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.250e+02 1.371e+02 1.487e+02 1.794e+02, threshold=2.742e+02, percent-clipped=0.0
2024-09-23 19:36:59,667 INFO [train.py:1198] (1/4) Epoch 19, batch 750, loss[loss=0.2469, ctc_loss=0.1677, cr_loss=0.396, over 16066.00 frames. ], tot_loss[loss=0.2174, ctc_loss=0.1452, cr_loss=0.3609, over 3300781.07 frames. ], batch size: 74, lr: 6.38e-03, grad_scale: 64.0
2024-09-23 19:36:59,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=330764.0, ans=0.0
2024-09-23 19:37:00,824 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.49 vs. limit=15.0
2024-09-23 19:37:03,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=330764.0, ans=0.0
2024-09-23 19:37:06,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=330764.0, ans=0.0
2024-09-23 19:37:15,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.whiten.whitening_limit, batch_count=330810.6666666667, ans=15.0
2024-09-23 19:37:18,099 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.23 vs. limit=22.5
2024-09-23 19:37:20,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=330810.6666666667, ans=0.04949747468305833
2024-09-23 19:37:20,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=330810.6666666667, ans=0.125
2024-09-23 19:37:20,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=330810.6666666667, ans=0.0
2024-09-23 19:37:36,759 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.25 vs. limit=15.0
2024-09-23 19:38:17,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=330950.6666666667, ans=0.125
2024-09-23 19:38:25,055 INFO [train.py:1198] (1/4) Epoch 19, batch 800, loss[loss=0.2236, ctc_loss=0.1463, cr_loss=0.3865, over 17015.00 frames. ], tot_loss[loss=0.2178, ctc_loss=0.1455, cr_loss=0.3618, over 3322633.46 frames. ], batch size: 51, lr: 6.38e-03, grad_scale: 32.0
2024-09-23 19:38:57,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=331090.6666666667, ans=0.125
2024-09-23 19:39:08,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=331090.6666666667, ans=0.0
2024-09-23 19:39:44,567 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.257e+02 1.307e+02 1.431e+02 3.261e+02, threshold=2.613e+02, percent-clipped=1.0
2024-09-23 19:39:44,591 INFO [train.py:1198] (1/4) Epoch 19, batch 850, loss[loss=0.1975, ctc_loss=0.1341, cr_loss=0.317, over 16944.00 frames. ], tot_loss[loss=0.2184, ctc_loss=0.1459, cr_loss=0.3625, over 3341434.91 frames. ], batch size: 42, lr: 6.37e-03, grad_scale: 32.0
2024-09-23 19:40:13,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=331277.3333333333, ans=0.125
2024-09-23 19:40:24,915 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:40:31,527 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.97 vs. limit=15.0
2024-09-23 19:40:52,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=331417.3333333333, ans=0.1
2024-09-23 19:41:04,316 INFO [train.py:1198] (1/4) Epoch 19, batch 900, loss[loss=0.2501, ctc_loss=0.1707, cr_loss=0.397, over 15378.00 frames. ], tot_loss[loss=0.2188, ctc_loss=0.1462, cr_loss=0.3629, over 3335232.51 frames. ], batch size: 89, lr: 6.37e-03, grad_scale: 16.0
2024-09-23 19:41:12,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=331464.0, ans=0.0
2024-09-23 19:42:24,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=331650.6666666667, ans=0.125
2024-09-23 19:42:24,657 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.73 vs. limit=15.0
2024-09-23 19:42:31,026 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.63 vs. limit=15.0
2024-09-23 19:42:31,886 INFO [train.py:1198] (1/4) Epoch 19, batch 950, loss[loss=0.2021, ctc_loss=0.1337, cr_loss=0.342, over 17021.00 frames. ], tot_loss[loss=0.2191, ctc_loss=0.1465, cr_loss=0.3628, over 3332961.23 frames. ], batch size: 51, lr: 6.37e-03, grad_scale: 16.0
2024-09-23 19:42:33,536 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.264e+02 1.377e+02 1.501e+02 1.833e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-23 19:43:24,891 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.51 vs. limit=15.0
2024-09-23 19:43:34,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=331837.3333333333, ans=0.0
2024-09-23 19:43:48,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=331884.0, ans=0.125
2024-09-23 19:43:54,920 INFO [train.py:1198] (1/4) Epoch 19, batch 1000, loss[loss=0.2354, ctc_loss=0.1585, cr_loss=0.3847, over 17109.00 frames. ], tot_loss[loss=0.2193, ctc_loss=0.1468, cr_loss=0.3624, over 3326300.98 frames. ], batch size: 49, lr: 6.37e-03, grad_scale: 16.0
2024-09-23 19:44:12,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.39 vs. limit=15.0
2024-09-23 19:44:22,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=331977.3333333333, ans=0.125
2024-09-23 19:45:09,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=332117.3333333333, ans=0.0
2024-09-23 19:45:14,335 INFO [train.py:1198] (1/4) Epoch 19, batch 1050, loss[loss=0.2228, ctc_loss=0.1499, cr_loss=0.3646, over 17093.00 frames. ], tot_loss[loss=0.2194, ctc_loss=0.1469, cr_loss=0.3626, over 3330177.97 frames. ], batch size: 49, lr: 6.36e-03, grad_scale: 16.0
2024-09-23 19:45:14,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=332164.0, ans=0.2
2024-09-23 19:45:15,994 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.292e+02 1.398e+02 1.505e+02 1.868e+02, threshold=2.795e+02, percent-clipped=0.0
2024-09-23 19:45:22,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=332164.0, ans=0.07
2024-09-23 19:45:24,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.87 vs. limit=15.0
2024-09-23 19:46:07,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.93 vs. limit=6.0
2024-09-23 19:46:15,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=332304.0, ans=0.125
2024-09-23 19:46:39,719 INFO [train.py:1198] (1/4) Epoch 19, batch 1100, loss[loss=0.1895, ctc_loss=0.1255, cr_loss=0.32, over 17080.00 frames. ], tot_loss[loss=0.2205, ctc_loss=0.1476, cr_loss=0.3642, over 3338653.76 frames. ], batch size: 43, lr: 6.36e-03, grad_scale: 16.0
2024-09-23 19:47:07,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=332444.0, ans=0.2
2024-09-23 19:47:22,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=332490.6666666667, ans=0.125
2024-09-23 19:47:46,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=332584.0, ans=0.0
2024-09-23 19:47:59,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=332584.0, ans=0.125
2024-09-23 19:48:01,782 INFO [train.py:1198] (1/4) Epoch 19, batch 1150, loss[loss=0.2219, ctc_loss=0.1483, cr_loss=0.3678, over 17206.00 frames. ], tot_loss[loss=0.22, ctc_loss=0.1472, cr_loss=0.3638, over 3341933.14 frames. ], batch size: 47, lr: 6.36e-03, grad_scale: 16.0
2024-09-23 19:48:06,032 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.288e+02 1.429e+02 1.556e+02 3.211e+02, threshold=2.859e+02, percent-clipped=1.0
2024-09-23 19:48:07,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=332630.6666666667, ans=0.1
2024-09-23 19:48:18,354 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=8.46 vs. limit=22.5
2024-09-23 19:48:31,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.02 vs. limit=10.0
2024-09-23 19:48:59,507 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=2.695e-03
2024-09-23 19:48:59,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=332770.6666666667, ans=0.1
2024-09-23 19:49:20,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=332817.3333333333, ans=0.125
2024-09-23 19:49:24,949 INFO [train.py:1198] (1/4) Epoch 19, batch 1200, loss[loss=0.181, ctc_loss=0.1195, cr_loss=0.3075, over 17167.00 frames. ], tot_loss[loss=0.2195, ctc_loss=0.1469, cr_loss=0.3631, over 3347209.64 frames. ], batch size: 41, lr: 6.36e-03, grad_scale: 32.0
2024-09-23 19:49:28,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=332864.0, ans=0.125
2024-09-23 19:49:43,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=332910.6666666667, ans=0.125
2024-09-23 19:50:08,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=332957.3333333333, ans=0.2
2024-09-23 19:50:11,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=333004.0, ans=0.0
2024-09-23 19:50:12,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=333004.0, ans=0.2
2024-09-23 19:50:12,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=333004.0, ans=0.125
2024-09-23 19:50:40,966 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 19:50:41,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=333050.6666666667, ans=0.125
2024-09-23 19:50:45,363 INFO [train.py:1198] (1/4) Epoch 19, batch 1250, loss[loss=0.1904, ctc_loss=0.1274, cr_loss=0.3148, over 17273.00 frames. ], tot_loss[loss=0.2194, ctc_loss=0.1469, cr_loss=0.3626, over 3344669.02 frames. ], batch size: 42, lr: 6.36e-03, grad_scale: 32.0
2024-09-23 19:50:46,878 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.324e+02 1.446e+02 1.574e+02 2.086e+02, threshold=2.891e+02, percent-clipped=0.0
2024-09-23 19:50:59,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=333144.0, ans=0.125
2024-09-23 19:51:12,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=333144.0, ans=0.125
2024-09-23 19:51:23,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=333190.6666666667, ans=0.0
2024-09-23 19:52:08,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=333284.0, ans=0.125
2024-09-23 19:52:12,988 INFO [train.py:1198] (1/4) Epoch 19, batch 1300, loss[loss=0.2003, ctc_loss=0.1316, cr_loss=0.3433, over 17080.00 frames. ], tot_loss[loss=0.2186, ctc_loss=0.1463, cr_loss=0.3613, over 3347140.87 frames. ], batch size: 43, lr: 6.35e-03, grad_scale: 16.0
2024-09-23 19:52:48,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=333424.0, ans=0.2
2024-09-23 19:53:01,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=333470.6666666667, ans=0.125
2024-09-23 19:53:35,916 INFO [train.py:1198] (1/4) Epoch 19, batch 1350, loss[loss=0.2086, ctc_loss=0.1367, cr_loss=0.3594, over 17292.00 frames. ], tot_loss[loss=0.2176, ctc_loss=0.1457, cr_loss=0.3598, over 3340349.39 frames. ], batch size: 49, lr: 6.35e-03, grad_scale: 16.0
2024-09-23 19:53:39,093 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.258e+02 1.337e+02 1.461e+02 2.024e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-23 19:53:54,554 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.35 vs. limit=15.0
2024-09-23 19:54:13,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=333657.3333333333, ans=0.125
2024-09-23 19:54:24,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=333704.0, ans=0.125
2024-09-23 19:54:56,452 INFO [train.py:1198] (1/4) Epoch 19, batch 1400, loss[loss=0.244, ctc_loss=0.168, cr_loss=0.3796, over 16569.00 frames. ], tot_loss[loss=0.2183, ctc_loss=0.1461, cr_loss=0.361, over 3344764.84 frames. ], batch size: 66, lr: 6.35e-03, grad_scale: 16.0
2024-09-23 19:55:01,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=333797.3333333333, ans=0.125
2024-09-23 19:55:03,798 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.03 vs. limit=10.0
2024-09-23 19:55:24,249 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.43 vs. limit=15.0
2024-09-23 19:55:27,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=333890.6666666667, ans=0.0
2024-09-23 19:55:49,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=333937.3333333333, ans=0.0
2024-09-23 19:56:05,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=333984.0, ans=0.125
2024-09-23 19:56:15,964 INFO [train.py:1198] (1/4) Epoch 19, batch 1450, loss[loss=0.2237, ctc_loss=0.149, cr_loss=0.3732, over 17235.00 frames. ], tot_loss[loss=0.217, ctc_loss=0.145, cr_loss=0.3602, over 3353953.31 frames. ], batch size: 50, lr: 6.35e-03, grad_scale: 16.0
2024-09-23 19:56:21,632 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.283e+02 1.375e+02 1.499e+02 2.283e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-23 19:56:45,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=334077.3333333333, ans=0.125
2024-09-23 19:56:51,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=334124.0, ans=0.125
2024-09-23 19:57:04,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=334124.0, ans=0.1
2024-09-23 19:57:16,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=334170.6666666667, ans=0.125
2024-09-23 19:57:39,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=334217.3333333333, ans=0.1
2024-09-23 19:57:43,769 INFO [train.py:1198] (1/4) Epoch 19, batch 1500, loss[loss=0.2338, ctc_loss=0.1567, cr_loss=0.3854, over 16517.00 frames. ], tot_loss[loss=0.2176, ctc_loss=0.1454, cr_loss=0.361, over 3351926.20 frames. ], batch size: 66, lr: 6.34e-03, grad_scale: 16.0
2024-09-23 19:57:56,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=334264.0, ans=0.025
2024-09-23 19:58:16,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=334357.3333333333, ans=0.1
2024-09-23 19:58:25,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=334357.3333333333, ans=0.125
2024-09-23 19:58:34,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=334404.0, ans=0.0
2024-09-23 19:58:58,867 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.52 vs. limit=22.5
2024-09-23 19:59:06,198 INFO [train.py:1198] (1/4) Epoch 19, batch 1550, loss[loss=0.2167, ctc_loss=0.1457, cr_loss=0.3547, over 17230.00 frames. ], tot_loss[loss=0.2177, ctc_loss=0.1456, cr_loss=0.3605, over 3355561.71 frames. ], batch size: 55, lr: 6.34e-03, grad_scale: 16.0
2024-09-23 19:59:09,456 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.039e+02 1.278e+02 1.387e+02 1.518e+02 2.007e+02, threshold=2.775e+02, percent-clipped=0.0
2024-09-23 20:00:26,299 INFO [train.py:1198] (1/4) Epoch 19, batch 1600, loss[loss=0.2053, ctc_loss=0.1336, cr_loss=0.3581, over 17045.00 frames. ], tot_loss[loss=0.2179, ctc_loss=0.1457, cr_loss=0.3608, over 3358917.07 frames. ], batch size: 39, lr: 6.34e-03, grad_scale: 32.0
2024-09-23 20:00:42,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=334777.3333333333, ans=0.0
2024-09-23 20:00:50,526 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 20:01:17,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=334870.6666666667, ans=0.125
2024-09-23 20:01:46,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=334917.3333333333, ans=0.2
2024-09-23 20:01:50,761 INFO [train.py:1198] (1/4) Epoch 19, batch 1650, loss[loss=0.26, ctc_loss=0.1823, cr_loss=0.3887, over 11799.00 frames. ], tot_loss[loss=0.2198, ctc_loss=0.1472, cr_loss=0.3633, over 3346901.86 frames. ], batch size: 123, lr: 6.34e-03, grad_scale: 32.0
2024-09-23 20:01:53,946 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.297e+02 1.356e+02 1.480e+02 2.099e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-23 20:02:56,778 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.38 vs. limit=15.0
2024-09-23 20:03:16,143 INFO [train.py:1198] (1/4) Epoch 19, batch 1700, loss[loss=0.2337, ctc_loss=0.1547, cr_loss=0.3954, over 17007.00 frames. ], tot_loss[loss=0.2187, ctc_loss=0.1463, cr_loss=0.3621, over 3353048.54 frames. ], batch size: 53, lr: 6.34e-03, grad_scale: 32.0
2024-09-23 20:03:17,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=335197.3333333333, ans=0.125
2024-09-23 20:03:37,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=335244.0, ans=0.125
2024-09-23 20:03:46,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=335290.6666666667, ans=0.125
2024-09-23 20:03:51,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=335290.6666666667, ans=0.125
2024-09-23 20:03:56,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=335290.6666666667, ans=0.2
2024-09-23 20:04:12,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=335337.3333333333, ans=0.125
2024-09-23 20:04:15,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer_na.min_abs, batch_count=335337.3333333333, ans=0.02
2024-09-23 20:04:28,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=335384.0, ans=0.125
2024-09-23 20:04:36,069 INFO [train.py:1198] (1/4) Epoch 19, batch 1750, loss[loss=0.1793, ctc_loss=0.1178, cr_loss=0.3078, over 15889.00 frames. ], tot_loss[loss=0.2169, ctc_loss=0.1449, cr_loss=0.3599, over 3354541.99 frames. ], batch size: 35, lr: 6.33e-03, grad_scale: 32.0
2024-09-23 20:04:39,273 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.243e+02 1.345e+02 1.441e+02 1.973e+02, threshold=2.691e+02, percent-clipped=0.0
2024-09-23 20:05:38,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=335617.3333333333, ans=0.0
2024-09-23 20:05:44,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=335617.3333333333, ans=0.125
2024-09-23 20:05:55,468 INFO [train.py:1198] (1/4) Epoch 19, batch 1800, loss[loss=0.2082, ctc_loss=0.1377, cr_loss=0.3524, over 17309.00 frames. ], tot_loss[loss=0.216, ctc_loss=0.1441, cr_loss=0.3594, over 3359232.85 frames. ], batch size: 49, lr: 6.33e-03, grad_scale: 32.0
2024-09-23 20:06:10,640 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.53 vs. limit=22.5
2024-09-23 20:06:18,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=335710.6666666667, ans=0.0
2024-09-23 20:06:39,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=335757.3333333333, ans=0.125
2024-09-23 20:06:48,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=335804.0, ans=0.125
2024-09-23 20:06:59,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=335804.0, ans=0.0
2024-09-23 20:07:04,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=335804.0, ans=0.5
2024-09-23 20:07:21,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=335897.3333333333, ans=0.125
2024-09-23 20:07:22,992 INFO [train.py:1198] (1/4) Epoch 19, batch 1850, loss[loss=0.1869, ctc_loss=0.1294, cr_loss=0.2873, over 16727.00 frames. ], tot_loss[loss=0.2165, ctc_loss=0.1445, cr_loss=0.3599, over 3362828.94 frames. ], batch size: 37, lr: 6.33e-03, grad_scale: 32.0
2024-09-23 20:07:23,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=335897.3333333333, ans=0.125
2024-09-23 20:07:26,227 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.276e+02 1.381e+02 1.510e+02 2.241e+02, threshold=2.761e+02, percent-clipped=0.0
2024-09-23 20:07:34,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=335897.3333333333, ans=0.025
2024-09-23 20:07:55,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=335990.6666666667, ans=0.1
2024-09-23 20:08:46,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=336130.6666666667, ans=0.125
2024-09-23 20:08:48,088 INFO [train.py:1198] (1/4) Epoch 19, batch 1900, loss[loss=0.2387, ctc_loss=0.1607, cr_loss=0.3896, over 16960.00 frames. ], tot_loss[loss=0.2158, ctc_loss=0.144, cr_loss=0.3591, over 3369915.93 frames. ], batch size: 58, lr: 6.33e-03, grad_scale: 32.0
2024-09-23 20:08:51,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=336130.6666666667, ans=0.09899494936611666
2024-09-23 20:09:07,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=336177.3333333333, ans=0.125
2024-09-23 20:09:31,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=336224.0, ans=0.2
2024-09-23 20:09:53,967 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.91 vs. limit=10.0
2024-09-23 20:09:58,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=336317.3333333333, ans=0.125
2024-09-23 20:10:06,856 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.19 vs. limit=15.0
2024-09-23 20:10:07,581 INFO [train.py:1198] (1/4) Epoch 19, batch 1950, loss[loss=0.2484, ctc_loss=0.1682, cr_loss=0.4012, over 17246.00 frames. ], tot_loss[loss=0.2161, ctc_loss=0.1442, cr_loss=0.3596, over 3367300.64 frames. ], batch size: 55, lr: 6.32e-03, grad_scale: 32.0
2024-09-23 20:10:10,843 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.232e+02 1.325e+02 1.435e+02 2.417e+02, threshold=2.650e+02, percent-clipped=0.0
2024-09-23 20:10:15,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=336364.0, ans=0.125
2024-09-23 20:10:56,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=336504.0, ans=0.0
2024-09-23 20:11:15,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=336550.6666666667, ans=0.1
2024-09-23 20:11:33,081 INFO [train.py:1198] (1/4) Epoch 19, batch 2000, loss[loss=0.1816, ctc_loss=0.1213, cr_loss=0.3014, over 17073.00 frames. ], tot_loss[loss=0.2167, ctc_loss=0.1447, cr_loss=0.3601, over 3362116.45 frames. ], batch size: 43, lr: 6.32e-03, grad_scale: 32.0
2024-09-23 20:11:40,130 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.21 vs. limit=12.0
2024-09-23 20:11:43,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten.whitening_limit, batch_count=336597.3333333333, ans=15.0
2024-09-23 20:11:52,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=336644.0, ans=0.125
2024-09-23 20:11:58,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=336644.0, ans=0.1
2024-09-23 20:12:15,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=336690.6666666667, ans=0.1
2024-09-23 20:12:18,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=336690.6666666667, ans=0.125
2024-09-23 20:12:26,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=336737.3333333333, ans=0.025
2024-09-23 20:12:32,205 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.09 vs. limit=6.0
2024-09-23 20:12:54,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=336830.6666666667, ans=0.2
2024-09-23 20:12:55,370 INFO [train.py:1198] (1/4) Epoch 19, batch 2050, loss[loss=0.2521, ctc_loss=0.1754, cr_loss=0.3835, over 14904.00 frames. ], tot_loss[loss=0.2173, ctc_loss=0.1452, cr_loss=0.3604, over 3356227.81 frames. ], batch size: 89, lr: 6.32e-03, grad_scale: 32.0
2024-09-23 20:12:57,650 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.95 vs. limit=10.0
2024-09-23 20:12:58,517 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.236e+02 1.316e+02 1.436e+02 2.046e+02, threshold=2.631e+02, percent-clipped=0.0
2024-09-23 20:13:04,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=336830.6666666667, ans=0.035
2024-09-23 20:13:06,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=336830.6666666667, ans=0.1
2024-09-23 20:13:17,923 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.85 vs. limit=15.0
2024-09-23 20:13:56,778 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.52 vs. limit=15.0
2024-09-23 20:13:59,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=336970.6666666667, ans=0.125
2024-09-23 20:14:12,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.88 vs. limit=15.0
2024-09-23 20:14:18,067 INFO [train.py:1198] (1/4) Epoch 19, batch 2100, loss[loss=0.2057, ctc_loss=0.1352, cr_loss=0.3524, over 17180.00 frames. ], tot_loss[loss=0.2164, ctc_loss=0.1445, cr_loss=0.3595, over 3360664.41 frames. ], batch size: 47, lr: 6.32e-03, grad_scale: 16.0
2024-09-23 20:15:06,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=337204.0, ans=0.125
2024-09-23 20:15:37,918 INFO [train.py:1198] (1/4) Epoch 19, batch 2150, loss[loss=0.2347, ctc_loss=0.1601, cr_loss=0.3728, over 17047.00 frames. ], tot_loss[loss=0.2177, ctc_loss=0.1456, cr_loss=0.3605, over 3348003.71 frames. ], batch size: 56, lr: 6.32e-03, grad_scale: 8.0
2024-09-23 20:15:44,256 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.267e+02 1.340e+02 1.517e+02 2.263e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-23 20:15:49,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=337297.3333333333, ans=0.05
2024-09-23 20:16:08,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=337390.6666666667, ans=0.125
2024-09-23 20:16:13,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=337390.6666666667, ans=0.2
2024-09-23 20:16:13,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=337390.6666666667, ans=0.1
2024-09-23 20:16:20,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=337390.6666666667, ans=0.1
2024-09-23 20:17:05,771 INFO [train.py:1198] (1/4) Epoch 19, batch 2200, loss[loss=0.2122, ctc_loss=0.145, cr_loss=0.336, over 16771.00 frames. ], tot_loss[loss=0.218, ctc_loss=0.1457, cr_loss=0.3611, over 3344518.21 frames. ], batch size: 61, lr: 6.31e-03, grad_scale: 8.0
2024-09-23 20:17:07,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=337530.6666666667, ans=0.1
2024-09-23 20:17:11,320 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.40 vs. limit=15.0
2024-09-23 20:17:12,817 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.48 vs. limit=12.0
2024-09-23 20:17:25,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=337577.3333333333, ans=0.1
2024-09-23 20:17:28,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=337577.3333333333, ans=0.025
2024-09-23 20:17:37,371 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.89 vs. limit=22.5
2024-09-23 20:17:55,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=337670.6666666667, ans=0.0
2024-09-23 20:18:26,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=337764.0, ans=0.125
2024-09-23 20:18:28,142 INFO [train.py:1198] (1/4) Epoch 19, batch 2250, loss[loss=0.2371, ctc_loss=0.1576, cr_loss=0.3975, over 16898.00 frames. ], tot_loss[loss=0.2177, ctc_loss=0.1455, cr_loss=0.3611, over 3352378.65 frames. ], batch size: 58, lr: 6.31e-03, grad_scale: 8.0
2024-09-23 20:18:34,521 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.290e+02 1.402e+02 1.494e+02 5.352e+02, threshold=2.803e+02, percent-clipped=1.0
2024-09-23 20:18:36,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=337764.0, ans=0.125
2024-09-23 20:18:38,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=337764.0, ans=0.2
2024-09-23 20:18:41,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=337764.0, ans=0.1
2024-09-23 20:18:41,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=337764.0, ans=0.035
2024-09-23 20:18:58,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=337857.3333333333, ans=0.1
2024-09-23 20:19:08,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=337857.3333333333, ans=0.0
2024-09-23 20:19:14,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=337904.0, ans=0.125
2024-09-23 20:19:30,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=337950.6666666667, ans=0.125
2024-09-23 20:19:34,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=337950.6666666667, ans=0.125
2024-09-23 20:19:45,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.08 vs. limit=15.0
2024-09-23 20:19:48,404 INFO [train.py:1198] (1/4) Epoch 19, batch 2300, loss[loss=0.2189, ctc_loss=0.1474, cr_loss=0.3576, over 17340.00 frames. ], tot_loss[loss=0.2176, ctc_loss=0.1454, cr_loss=0.3612, over 3364443.64 frames. ], batch size: 48, lr: 6.31e-03, grad_scale: 8.0
2024-09-23 20:19:49,174 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=7.06 vs. limit=15.0
2024-09-23 20:19:51,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=337997.3333333333, ans=0.125
2024-09-23 20:20:37,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=338137.3333333333, ans=10.0
2024-09-23 20:20:48,894 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.96 vs. limit=10.0
2024-09-23 20:21:09,227 INFO [train.py:1198] (1/4) Epoch 19, batch 2350, loss[loss=0.23, ctc_loss=0.1571, cr_loss=0.3648, over 16402.00 frames. ], tot_loss[loss=0.2187, ctc_loss=0.1461, cr_loss=0.3629, over 3361237.70 frames. ], batch size: 66, lr: 6.31e-03, grad_scale: 8.0
2024-09-23 20:21:11,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=338230.6666666667, ans=0.125
2024-09-23 20:21:18,094 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.269e+02 1.353e+02 1.497e+02 2.252e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-23 20:21:23,279 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.46 vs. limit=15.0
2024-09-23 20:21:32,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=338277.3333333333, ans=0.09899494936611666
2024-09-23 20:21:56,970 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 20:22:11,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=338370.6666666667, ans=0.125
2024-09-23 20:22:11,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=338370.6666666667, ans=0.125
2024-09-23 20:22:19,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=338417.3333333333, ans=0.07
2024-09-23 20:22:36,779 INFO [train.py:1198] (1/4) Epoch 19, batch 2400, loss[loss=0.1853, ctc_loss=0.1194, cr_loss=0.3292, over 16965.00 frames. ], tot_loss[loss=0.2194, ctc_loss=0.1467, cr_loss=0.3635, over 3355833.17 frames. ], batch size: 42, lr: 6.31e-03, grad_scale: 16.0
2024-09-23 20:22:40,751 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.88 vs. limit=15.0
2024-09-23 20:22:42,040 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.63 vs. limit=22.5
2024-09-23 20:22:53,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=338510.6666666667, ans=0.2
2024-09-23 20:22:54,242 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.75 vs. limit=15.0
2024-09-23 20:23:08,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=338510.6666666667, ans=0.125
2024-09-23 20:23:40,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=338604.0, ans=0.1
2024-09-23 20:23:45,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=338650.6666666667, ans=0.035
2024-09-23 20:23:59,488 INFO [train.py:1198] (1/4) Epoch 19, batch 2450, loss[loss=0.2092, ctc_loss=0.1389, cr_loss=0.3513, over 17009.00 frames. ], tot_loss[loss=0.2185, ctc_loss=0.146, cr_loss=0.3623, over 3357887.40 frames. ], batch size: 39, lr: 6.30e-03, grad_scale: 16.0
2024-09-23 20:24:05,811 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.266e+02 1.363e+02 1.497e+02 1.978e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-23 20:24:28,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=338744.0, ans=0.025
2024-09-23 20:24:38,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=338790.6666666667, ans=0.1
2024-09-23 20:24:58,486 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 20:25:19,141 INFO [train.py:1198] (1/4) Epoch 19, batch 2500, loss[loss=0.2151, ctc_loss=0.1434, cr_loss=0.3583, over 17203.00 frames. ], tot_loss[loss=0.2182, ctc_loss=0.1457, cr_loss=0.3622, over 3368024.93 frames. ], batch size: 50, lr: 6.30e-03, grad_scale: 16.0
2024-09-23 20:25:19,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=338930.6666666667, ans=0.0
2024-09-23 20:25:43,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=338977.3333333333, ans=0.04949747468305833
2024-09-23 20:25:49,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=339024.0, ans=0.0
2024-09-23 20:25:49,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=339024.0, ans=0.1
2024-09-23 20:26:10,214 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 20:26:30,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=339117.3333333333, ans=0.125
2024-09-23 20:26:36,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=339117.3333333333, ans=0.0
2024-09-23 20:26:42,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=339164.0, ans=0.07
2024-09-23 20:26:43,706 INFO [train.py:1198] (1/4) Epoch 19, batch 2550, loss[loss=0.2435, ctc_loss=0.1657, cr_loss=0.3887, over 17004.00 frames. ], tot_loss[loss=0.2181, ctc_loss=0.1456, cr_loss=0.3626, over 3374951.37 frames. ], batch size: 51, lr: 6.30e-03, grad_scale: 16.0
2024-09-23 20:26:51,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=339164.0, ans=15.0
2024-09-23 20:26:52,585 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.236e+02 1.315e+02 1.425e+02 2.139e+02, threshold=2.630e+02, percent-clipped=0.0
2024-09-23 20:27:04,440 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=13.14 vs. limit=15.0
2024-09-23 20:27:05,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=339210.6666666667, ans=0.125
2024-09-23 20:27:06,244 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=6.55 vs. limit=15.0
2024-09-23 20:27:14,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.27 vs. limit=15.0
2024-09-23 20:27:23,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=339257.3333333333, ans=0.1
2024-09-23 20:27:34,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=339304.0, ans=0.0
2024-09-23 20:27:34,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=339304.0, ans=0.0
2024-09-23 20:27:35,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=339304.0, ans=0.0
2024-09-23 20:27:38,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=339304.0, ans=0.125
2024-09-23 20:28:03,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=339350.6666666667, ans=0.125
2024-09-23 20:28:08,358 INFO [train.py:1198] (1/4) Epoch 19, batch 2600, loss[loss=0.232, ctc_loss=0.158, cr_loss=0.3697, over 17051.00 frames. ], tot_loss[loss=0.2191, ctc_loss=0.1464, cr_loss=0.3635, over 3372295.07 frames. ], batch size: 52, lr: 6.30e-03, grad_scale: 16.0
2024-09-23 20:28:15,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=339397.3333333333, ans=0.125
2024-09-23 20:28:18,599 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.65 vs. limit=15.0
2024-09-23 20:28:21,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=339397.3333333333, ans=0.125
2024-09-23 20:28:57,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.38 vs. limit=10.0
2024-09-23 20:29:09,219 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=339537.3333333333, ans=0.0
2024-09-23 20:29:10,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=339584.0, ans=0.2
2024-09-23 20:29:16,139 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.82 vs. limit=15.0
2024-09-23 20:29:28,190 INFO [train.py:1198] (1/4) Epoch 19, batch 2650, loss[loss=0.2282, ctc_loss=0.1549, cr_loss=0.3665, over 17221.00 frames. ], tot_loss[loss=0.2196, ctc_loss=0.1468, cr_loss=0.3638, over 3368576.64 frames. ], batch size: 47, lr: 6.29e-03, grad_scale: 16.0
2024-09-23 20:29:30,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=339630.6666666667, ans=0.125
2024-09-23 20:29:34,367 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.315e+02 1.464e+02 1.614e+02 2.211e+02, threshold=2.927e+02, percent-clipped=0.0
2024-09-23 20:29:39,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=339630.6666666667, ans=0.125
2024-09-23 20:30:30,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=339817.3333333333, ans=0.125
2024-09-23 20:30:48,090 INFO [train.py:1198] (1/4) Epoch 19, batch 2700, loss[loss=0.2104, ctc_loss=0.1377, cr_loss=0.3636, over 17101.00 frames. ], tot_loss[loss=0.2191, ctc_loss=0.1463, cr_loss=0.3639, over 3376813.42 frames. ], batch size: 43, lr: 6.29e-03, grad_scale: 16.0
2024-09-23 20:30:50,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=339864.0, ans=0.0
2024-09-23 20:30:50,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=339864.0, ans=0.125
2024-09-23 20:31:05,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=339910.6666666667, ans=0.025
2024-09-23 20:31:41,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=340004.0, ans=0.1
2024-09-23 20:32:00,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=340050.6666666667, ans=0.125
2024-09-23 20:32:02,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=340050.6666666667, ans=0.125
2024-09-23 20:32:05,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=340050.6666666667, ans=0.125
2024-09-23 20:32:16,072 INFO [train.py:1198] (1/4) Epoch 19, batch 2750, loss[loss=0.2303, ctc_loss=0.1536, cr_loss=0.3836, over 17206.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.1462, cr_loss=0.3636, over 3371794.61 frames. ], batch size: 55, lr: 6.29e-03, grad_scale: 16.0
2024-09-23 20:32:16,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=340097.3333333333, ans=0.125
2024-09-23 20:32:17,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=340097.3333333333, ans=0.125
2024-09-23 20:32:22,227 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.266e+02 1.341e+02 1.484e+02 3.814e+02, threshold=2.681e+02, percent-clipped=1.0
2024-09-23 20:32:22,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=340097.3333333333, ans=0.1
2024-09-23 20:32:30,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=340144.0, ans=0.0
2024-09-23 20:32:40,699 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.88 vs. limit=15.0
2024-09-23 20:32:46,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=340190.6666666667, ans=0.0
2024-09-23 20:33:13,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=340237.3333333333, ans=0.1
2024-09-23 20:33:32,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=340284.0, ans=0.04949747468305833
2024-09-23 20:33:38,450 INFO [train.py:1198] (1/4) Epoch 19, batch 2800, loss[loss=0.2178, ctc_loss=0.1456, cr_loss=0.3609, over 17021.00 frames. ], tot_loss[loss=0.2186, ctc_loss=0.146, cr_loss=0.3626, over 3363620.94 frames. ], batch size: 56, lr: 6.29e-03, grad_scale: 32.0
2024-09-23 20:33:46,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=340330.6666666667, ans=0.0
2024-09-23 20:33:59,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=340377.3333333333, ans=0.0
2024-09-23 20:34:58,053 INFO [train.py:1198] (1/4) Epoch 19, batch 2850, loss[loss=0.2558, ctc_loss=0.1732, cr_loss=0.413, over 17223.00 frames. ], tot_loss[loss=0.2197, ctc_loss=0.1469, cr_loss=0.364, over 3350490.73 frames. ], batch size: 55, lr: 6.29e-03, grad_scale: 32.0
2024-09-23 20:35:01,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=340564.0, ans=0.0
2024-09-23 20:35:04,575 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.252e+02 1.382e+02 1.524e+02 2.298e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-23 20:35:26,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=340610.6666666667, ans=0.0
2024-09-23 20:35:35,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=340657.3333333333, ans=0.1
2024-09-23 20:36:01,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=340704.0, ans=0.0
2024-09-23 20:36:13,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=340750.6666666667, ans=0.0
2024-09-23 20:36:18,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=340750.6666666667, ans=0.2
2024-09-23 20:36:22,620 INFO [train.py:1198] (1/4) Epoch 19, batch 2900, loss[loss=0.2177, ctc_loss=0.1458, cr_loss=0.3596, over 17019.00 frames. ], tot_loss[loss=0.2194, ctc_loss=0.1467, cr_loss=0.3634, over 3344415.13 frames. ], batch size: 52, lr: 6.28e-03, grad_scale: 16.0
2024-09-23 20:36:30,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=340797.3333333333, ans=0.0
2024-09-23 20:36:35,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=340797.3333333333, ans=0.125
2024-09-23 20:37:11,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=340937.3333333333, ans=0.125
2024-09-23 20:37:29,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=340984.0, ans=0.2
2024-09-23 20:37:39,304 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.81 vs. limit=15.0
2024-09-23 20:37:47,690 INFO [train.py:1198] (1/4) Epoch 19, batch 2950, loss[loss=0.2057, ctc_loss=0.1378, cr_loss=0.3397, over 16960.00 frames. ], tot_loss[loss=0.2192, ctc_loss=0.1465, cr_loss=0.3631, over 3351942.11 frames. ], batch size: 42, lr: 6.28e-03, grad_scale: 16.0
2024-09-23 20:37:55,446 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.260e+02 1.402e+02 1.500e+02 2.241e+02, threshold=2.804e+02, percent-clipped=0.0
2024-09-23 20:37:55,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=341030.6666666667, ans=0.0
2024-09-23 20:38:24,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=341124.0, ans=0.0
2024-09-23 20:38:53,376 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.83 vs. limit=10.0
2024-09-23 20:38:54,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=341217.3333333333, ans=0.125
2024-09-23 20:39:06,540 INFO [train.py:1198] (1/4) Epoch 19, batch 3000, loss[loss=0.2174, ctc_loss=0.1456, cr_loss=0.3591, over 17151.00 frames. ], tot_loss[loss=0.2196, ctc_loss=0.1469, cr_loss=0.3638, over 3356009.13 frames. ], batch size: 48, lr: 6.28e-03, grad_scale: 16.0
2024-09-23 20:39:06,540 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 20:39:21,854 INFO [train.py:1230] (1/4) Epoch 19, validation: loss=0.03984, ctc_loss=0.03984, cr_loss=8.01e-15, over 944034.00 frames. 
2024-09-23 20:39:21,855 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 20:39:26,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=341264.0, ans=0.125
2024-09-23 20:39:58,779 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.90 vs. limit=15.0
2024-09-23 20:40:24,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=341450.6666666667, ans=0.125
2024-09-23 20:40:40,304 INFO [train.py:1198] (1/4) Epoch 19, batch 3050, loss[loss=0.1932, ctc_loss=0.1262, cr_loss=0.3349, over 17153.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.1464, cr_loss=0.3631, over 3352659.38 frames. ], batch size: 45, lr: 6.28e-03, grad_scale: 16.0
2024-09-23 20:40:47,386 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=18.27 vs. limit=22.5
2024-09-23 20:40:48,102 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.265e+02 1.361e+02 1.501e+02 2.045e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-23 20:41:43,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=341684.0, ans=0.125
2024-09-23 20:41:59,531 INFO [train.py:1198] (1/4) Epoch 19, batch 3100, loss[loss=0.2419, ctc_loss=0.1613, cr_loss=0.403, over 17010.00 frames. ], tot_loss[loss=0.2186, ctc_loss=0.1461, cr_loss=0.3622, over 3353024.52 frames. ], batch size: 56, lr: 6.28e-03, grad_scale: 16.0
2024-09-23 20:42:07,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=341730.6666666667, ans=0.125
2024-09-23 20:42:18,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=341777.3333333333, ans=0.125
2024-09-23 20:42:36,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=341824.0, ans=0.0
2024-09-23 20:42:48,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=341870.6666666667, ans=0.2
2024-09-23 20:43:04,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=341917.3333333333, ans=0.125
2024-09-23 20:43:04,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=341917.3333333333, ans=0.125
2024-09-23 20:43:18,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=341964.0, ans=0.2
2024-09-23 20:43:20,310 INFO [train.py:1198] (1/4) Epoch 19, batch 3150, loss[loss=0.2121, ctc_loss=0.1417, cr_loss=0.3516, over 17205.00 frames. ], tot_loss[loss=0.2175, ctc_loss=0.1453, cr_loss=0.3611, over 3353263.87 frames. ], batch size: 50, lr: 6.27e-03, grad_scale: 16.0
2024-09-23 20:43:30,465 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.288e+02 1.385e+02 1.538e+02 2.696e+02, threshold=2.771e+02, percent-clipped=0.0
2024-09-23 20:43:43,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=342010.6666666667, ans=0.0
2024-09-23 20:44:16,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=342104.0, ans=0.125
2024-09-23 20:44:44,044 INFO [train.py:1198] (1/4) Epoch 19, batch 3200, loss[loss=0.2397, ctc_loss=0.1598, cr_loss=0.3993, over 17029.00 frames. ], tot_loss[loss=0.2179, ctc_loss=0.1456, cr_loss=0.3615, over 3359012.08 frames. ], batch size: 52, lr: 6.27e-03, grad_scale: 32.0
2024-09-23 20:44:50,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=342197.3333333333, ans=0.2
2024-09-23 20:45:34,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=342337.3333333333, ans=0.0
2024-09-23 20:45:35,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=342337.3333333333, ans=0.125
2024-09-23 20:46:01,866 INFO [train.py:1198] (1/4) Epoch 19, batch 3250, loss[loss=0.2032, ctc_loss=0.1348, cr_loss=0.342, over 17221.00 frames. ], tot_loss[loss=0.2168, ctc_loss=0.1448, cr_loss=0.3602, over 3362997.80 frames. ], batch size: 47, lr: 6.27e-03, grad_scale: 16.0
2024-09-23 20:46:11,249 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.296e+02 1.376e+02 1.501e+02 2.697e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-23 20:46:17,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=342477.3333333333, ans=0.09899494936611666
2024-09-23 20:46:30,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=342477.3333333333, ans=0.2
2024-09-23 20:46:36,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=342524.0, ans=0.1
2024-09-23 20:46:38,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=342524.0, ans=0.2
2024-09-23 20:46:46,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=342524.0, ans=0.125
2024-09-23 20:47:22,260 INFO [train.py:1198] (1/4) Epoch 19, batch 3300, loss[loss=0.221, ctc_loss=0.1464, cr_loss=0.373, over 17096.00 frames. ], tot_loss[loss=0.2171, ctc_loss=0.1451, cr_loss=0.3599, over 3363468.04 frames. ], batch size: 49, lr: 6.27e-03, grad_scale: 16.0
2024-09-23 20:47:46,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=342710.6666666667, ans=0.125
2024-09-23 20:47:47,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=342710.6666666667, ans=0.125
2024-09-23 20:47:56,208 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.68 vs. limit=10.0
2024-09-23 20:48:05,916 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.56 vs. limit=15.0
2024-09-23 20:48:14,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=342804.0, ans=0.07
2024-09-23 20:48:33,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=342850.6666666667, ans=0.0
2024-09-23 20:48:41,089 INFO [train.py:1198] (1/4) Epoch 19, batch 3350, loss[loss=0.1845, ctc_loss=0.1204, cr_loss=0.3203, over 17090.00 frames. ], tot_loss[loss=0.2171, ctc_loss=0.1451, cr_loss=0.3598, over 3364017.50 frames. ], batch size: 40, lr: 6.26e-03, grad_scale: 16.0
2024-09-23 20:48:49,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=342897.3333333333, ans=0.0
2024-09-23 20:48:50,524 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.298e+02 1.381e+02 1.515e+02 2.027e+02, threshold=2.761e+02, percent-clipped=0.0
2024-09-23 20:48:55,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=342944.0, ans=0.0
2024-09-23 20:49:33,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=343037.3333333333, ans=0.125
2024-09-23 20:49:39,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=343037.3333333333, ans=10.0
2024-09-23 20:49:41,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=343037.3333333333, ans=0.1
2024-09-23 20:49:58,907 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.09 vs. limit=6.0
2024-09-23 20:49:59,654 INFO [train.py:1198] (1/4) Epoch 19, batch 3400, loss[loss=0.2156, ctc_loss=0.1412, cr_loss=0.3721, over 17150.00 frames. ], tot_loss[loss=0.2169, ctc_loss=0.1448, cr_loss=0.36, over 3371871.44 frames. ], batch size: 45, lr: 6.26e-03, grad_scale: 16.0
2024-09-23 20:50:27,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=343177.3333333333, ans=0.025
2024-09-23 20:51:00,348 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.64 vs. limit=15.0
2024-09-23 20:51:12,701 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.69 vs. limit=10.0
2024-09-23 20:51:16,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=343364.0, ans=0.1
2024-09-23 20:51:17,914 INFO [train.py:1198] (1/4) Epoch 19, batch 3450, loss[loss=0.2412, ctc_loss=0.1628, cr_loss=0.3921, over 15022.00 frames. ], tot_loss[loss=0.217, ctc_loss=0.1449, cr_loss=0.3606, over 3364871.11 frames. ], batch size: 89, lr: 6.26e-03, grad_scale: 16.0
2024-09-23 20:51:26,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.88 vs. limit=10.0
2024-09-23 20:51:27,572 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.246e+02 1.361e+02 1.474e+02 2.072e+02, threshold=2.723e+02, percent-clipped=0.0
2024-09-23 20:51:33,078 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.66 vs. limit=22.5
2024-09-23 20:51:34,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=343410.6666666667, ans=0.125
2024-09-23 20:51:40,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=343410.6666666667, ans=0.0
2024-09-23 20:51:48,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=343457.3333333333, ans=0.025
2024-09-23 20:52:36,230 INFO [train.py:1198] (1/4) Epoch 19, batch 3500, loss[loss=0.17, ctc_loss=0.1115, cr_loss=0.2925, over 17115.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1438, cr_loss=0.3587, over 3367492.22 frames. ], batch size: 40, lr: 6.26e-03, grad_scale: 16.0
2024-09-23 20:52:39,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=343597.3333333333, ans=0.125
2024-09-23 20:52:42,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=343597.3333333333, ans=0.125
2024-09-23 20:52:44,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=343597.3333333333, ans=10.0
2024-09-23 20:52:45,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=343597.3333333333, ans=0.0
2024-09-23 20:52:56,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=343644.0, ans=0.04949747468305833
2024-09-23 20:53:11,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.min_positive, batch_count=343690.6666666667, ans=0.05
2024-09-23 20:53:20,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=343690.6666666667, ans=0.025
2024-09-23 20:53:35,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=343737.3333333333, ans=0.1
2024-09-23 20:53:57,647 INFO [train.py:1198] (1/4) Epoch 19, batch 3550, loss[loss=0.229, ctc_loss=0.1528, cr_loss=0.3813, over 17064.00 frames. ], tot_loss[loss=0.2175, ctc_loss=0.1451, cr_loss=0.3618, over 3367806.90 frames. ], batch size: 46, lr: 6.26e-03, grad_scale: 16.0
2024-09-23 20:54:01,199 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 20:54:04,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=343830.6666666667, ans=0.2
2024-09-23 20:54:07,040 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.247e+02 1.326e+02 1.445e+02 2.020e+02, threshold=2.653e+02, percent-clipped=0.0
2024-09-23 20:54:13,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=343877.3333333333, ans=0.0
2024-09-23 20:54:23,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=343877.3333333333, ans=0.1
2024-09-23 20:54:31,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=343924.0, ans=0.125
2024-09-23 20:55:17,372 INFO [train.py:1198] (1/4) Epoch 19, batch 3600, loss[loss=0.1751, ctc_loss=0.1148, cr_loss=0.3016, over 16977.00 frames. ], tot_loss[loss=0.2147, ctc_loss=0.1431, cr_loss=0.358, over 3369925.00 frames. ], batch size: 42, lr: 6.25e-03, grad_scale: 32.0
2024-09-23 20:55:28,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=344064.0, ans=0.0
2024-09-23 20:55:41,231 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.12 vs. limit=22.5
2024-09-23 20:55:54,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=344157.3333333333, ans=0.025
2024-09-23 20:55:56,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=344157.3333333333, ans=0.125
2024-09-23 20:55:56,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=344157.3333333333, ans=0.125
2024-09-23 20:55:59,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=344157.3333333333, ans=0.125
2024-09-23 20:56:00,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=344157.3333333333, ans=0.125
2024-09-23 20:56:02,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=344204.0, ans=0.125
2024-09-23 20:56:22,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.64 vs. limit=15.0
2024-09-23 20:56:34,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=344250.6666666667, ans=0.0
2024-09-23 20:56:36,797 INFO [train.py:1198] (1/4) Epoch 19, batch 3650, loss[loss=0.2336, ctc_loss=0.1576, cr_loss=0.3802, over 15864.00 frames. ], tot_loss[loss=0.2152, ctc_loss=0.1436, cr_loss=0.3581, over 3356872.01 frames. ], batch size: 74, lr: 6.25e-03, grad_scale: 32.0
2024-09-23 20:56:47,598 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.302e+02 1.379e+02 1.507e+02 2.534e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-23 20:57:02,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=344344.0, ans=0.0
2024-09-23 20:57:40,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=344484.0, ans=0.025
2024-09-23 20:57:50,650 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.68 vs. limit=6.0
2024-09-23 20:57:56,273 INFO [train.py:1198] (1/4) Epoch 19, batch 3700, loss[loss=0.2225, ctc_loss=0.1454, cr_loss=0.3856, over 17150.00 frames. ], tot_loss[loss=0.2164, ctc_loss=0.1445, cr_loss=0.3592, over 3347374.76 frames. ], batch size: 48, lr: 6.25e-03, grad_scale: 16.0
2024-09-23 20:57:56,693 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 20:58:10,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=344577.3333333333, ans=0.125
2024-09-23 20:58:10,763 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 20:58:21,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=344577.3333333333, ans=0.0
2024-09-23 20:58:21,897 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.42 vs. limit=6.0
2024-09-23 20:58:27,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=344624.0, ans=0.125
2024-09-23 20:58:32,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=344624.0, ans=0.125
2024-09-23 20:58:58,550 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.36 vs. limit=22.5
2024-09-23 20:59:02,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=344717.3333333333, ans=0.125
2024-09-23 20:59:05,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=344717.3333333333, ans=22.5
2024-09-23 20:59:14,535 INFO [train.py:1198] (1/4) Epoch 19, batch 3750, loss[loss=0.1814, ctc_loss=0.1186, cr_loss=0.3139, over 17026.00 frames. ], tot_loss[loss=0.2166, ctc_loss=0.1447, cr_loss=0.3597, over 3350223.71 frames. ], batch size: 39, lr: 6.25e-03, grad_scale: 16.0
2024-09-23 20:59:25,487 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.321e+02 1.412e+02 1.562e+02 2.185e+02, threshold=2.824e+02, percent-clipped=0.0
2024-09-23 20:59:38,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=344810.6666666667, ans=0.125
2024-09-23 20:59:54,149 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.93 vs. limit=15.0
2024-09-23 21:00:07,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=344904.0, ans=0.125
2024-09-23 21:00:14,744 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.84 vs. limit=15.0
2024-09-23 21:00:19,206 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.12 vs. limit=6.0
2024-09-23 21:00:32,421 INFO [train.py:1198] (1/4) Epoch 19, batch 3800, loss[loss=0.2188, ctc_loss=0.1467, cr_loss=0.3609, over 17311.00 frames. ], tot_loss[loss=0.2177, ctc_loss=0.1456, cr_loss=0.3603, over 3324530.18 frames. ], batch size: 51, lr: 6.25e-03, grad_scale: 16.0
2024-09-23 21:00:40,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=344997.3333333333, ans=0.125
2024-09-23 21:00:46,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=345044.0, ans=0.025
2024-09-23 21:01:21,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=345137.3333333333, ans=0.025
2024-09-23 21:01:30,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=345137.3333333333, ans=0.125
2024-09-23 21:01:35,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.18 vs. limit=15.0
2024-09-23 21:01:41,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=345184.0, ans=0.0
2024-09-23 21:01:49,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=345230.6666666667, ans=0.0
2024-09-23 21:01:50,702 INFO [train.py:1198] (1/4) Epoch 19, batch 3850, loss[loss=0.2502, ctc_loss=0.1721, cr_loss=0.3905, over 16850.00 frames. ], tot_loss[loss=0.2185, ctc_loss=0.1464, cr_loss=0.3607, over 3290838.99 frames. ], batch size: 58, lr: 6.24e-03, grad_scale: 16.0
2024-09-23 21:02:00,345 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=5.283e-03
2024-09-23 21:02:01,517 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.310e+02 1.460e+02 1.598e+02 2.355e+02, threshold=2.920e+02, percent-clipped=0.0
2024-09-23 21:02:01,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=345230.6666666667, ans=0.0
2024-09-23 21:02:36,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=345370.6666666667, ans=0.125
2024-09-23 21:02:37,621 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=345370.6666666667, ans=0.125
2024-09-23 21:03:53,557 INFO [train.py:1198] (1/4) Epoch 20, batch 0, loss[loss=0.2772, ctc_loss=0.1895, cr_loss=0.4385, over 15137.00 frames. ], tot_loss[loss=0.2772, ctc_loss=0.1895, cr_loss=0.4385, over 15137.00 frames. ], batch size: 89, lr: 6.08e-03, grad_scale: 32.0
2024-09-23 21:03:53,557 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 21:04:07,358 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.1.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([5.0173, 4.8625, 4.2903, 4.7824], device='cuda:1')
2024-09-23 21:04:08,661 INFO [train.py:1230] (1/4) Epoch 20, validation: loss=0.03935, ctc_loss=0.03935, cr_loss=7.664e-15, over 944034.00 frames. 
2024-09-23 21:04:08,662 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 21:04:26,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=345492.0, ans=0.0
2024-09-23 21:04:27,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=345492.0, ans=0.1
2024-09-23 21:05:04,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=345585.3333333333, ans=0.1
2024-09-23 21:05:07,623 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.68 vs. limit=22.5
2024-09-23 21:05:15,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=345585.3333333333, ans=0.2
2024-09-23 21:05:30,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=345632.0, ans=0.125
2024-09-23 21:05:33,920 INFO [train.py:1198] (1/4) Epoch 20, batch 50, loss[loss=0.1958, ctc_loss=0.1267, cr_loss=0.3454, over 17269.00 frames. ], tot_loss[loss=0.2204, ctc_loss=0.1476, cr_loss=0.3641, over 758464.14 frames. ], batch size: 42, lr: 6.08e-03, grad_scale: 32.0
2024-09-23 21:05:35,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=345678.6666666667, ans=0.125
2024-09-23 21:05:51,260 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.262e+02 1.457e+02 1.636e+02 2.185e+02, threshold=2.915e+02, percent-clipped=0.0
2024-09-23 21:06:30,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=345818.6666666667, ans=0.025
2024-09-23 21:06:44,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=345865.3333333333, ans=0.035
2024-09-23 21:06:51,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=345865.3333333333, ans=0.04949747468305833
2024-09-23 21:06:53,842 INFO [train.py:1198] (1/4) Epoch 20, batch 100, loss[loss=0.2399, ctc_loss=0.1629, cr_loss=0.3852, over 16898.00 frames. ], tot_loss[loss=0.221, ctc_loss=0.1478, cr_loss=0.3659, over 1336857.90 frames. ], batch size: 58, lr: 6.08e-03, grad_scale: 32.0
2024-09-23 21:07:16,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=345958.6666666667, ans=0.5
2024-09-23 21:07:19,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=345958.6666666667, ans=0.125
2024-09-23 21:08:02,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=346098.6666666667, ans=0.0
2024-09-23 21:08:18,775 INFO [train.py:1198] (1/4) Epoch 20, batch 150, loss[loss=0.2519, ctc_loss=0.1694, cr_loss=0.4124, over 15909.00 frames. ], tot_loss[loss=0.2189, ctc_loss=0.1462, cr_loss=0.3632, over 1787919.86 frames. ], batch size: 74, lr: 6.07e-03, grad_scale: 16.0
2024-09-23 21:08:35,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=346192.0, ans=0.0
2024-09-23 21:08:38,025 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.021e+02 1.321e+02 1.429e+02 1.602e+02 2.448e+02, threshold=2.858e+02, percent-clipped=0.0
2024-09-23 21:09:17,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=346285.3333333333, ans=0.09899494936611666
2024-09-23 21:09:34,934 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.85 vs. limit=15.0
2024-09-23 21:09:40,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=346332.0, ans=0.1
2024-09-23 21:09:45,099 INFO [train.py:1198] (1/4) Epoch 20, batch 200, loss[loss=0.27, ctc_loss=0.1888, cr_loss=0.4063, over 11914.00 frames. ], tot_loss[loss=0.2153, ctc_loss=0.1436, cr_loss=0.3588, over 2129654.12 frames. ], batch size: 125, lr: 6.07e-03, grad_scale: 16.0
2024-09-23 21:09:50,219 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=346378.6666666667, ans=0.125
2024-09-23 21:10:20,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=346472.0, ans=0.125
2024-09-23 21:10:22,545 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.20 vs. limit=15.0
2024-09-23 21:10:36,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=346518.6666666667, ans=0.0
2024-09-23 21:10:41,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=346518.6666666667, ans=0.125
2024-09-23 21:10:46,071 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 21:10:46,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=346518.6666666667, ans=0.125
2024-09-23 21:11:04,972 INFO [train.py:1198] (1/4) Epoch 20, batch 250, loss[loss=0.2035, ctc_loss=0.1399, cr_loss=0.3183, over 15918.00 frames. ], tot_loss[loss=0.2154, ctc_loss=0.1437, cr_loss=0.3584, over 2402122.40 frames. ], batch size: 74, lr: 6.07e-03, grad_scale: 16.0
2024-09-23 21:11:23,903 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.297e+02 1.428e+02 1.617e+02 1.854e+02, threshold=2.857e+02, percent-clipped=0.0
2024-09-23 21:12:24,539 INFO [train.py:1198] (1/4) Epoch 20, batch 300, loss[loss=0.2008, ctc_loss=0.1336, cr_loss=0.3359, over 17041.00 frames. ], tot_loss[loss=0.2159, ctc_loss=0.1441, cr_loss=0.3591, over 2611058.49 frames. ], batch size: 39, lr: 6.07e-03, grad_scale: 16.0
2024-09-23 21:12:48,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=346892.0, ans=0.2
2024-09-23 21:12:54,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_positive, batch_count=346892.0, ans=0.05
2024-09-23 21:13:04,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=346938.6666666667, ans=0.125
2024-09-23 21:13:29,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=346985.3333333333, ans=0.025
2024-09-23 21:13:48,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=347078.6666666667, ans=0.125
2024-09-23 21:13:48,767 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.36 vs. limit=22.5
2024-09-23 21:13:49,665 INFO [train.py:1198] (1/4) Epoch 20, batch 350, loss[loss=0.2422, ctc_loss=0.1626, cr_loss=0.398, over 16990.00 frames. ], tot_loss[loss=0.2164, ctc_loss=0.1444, cr_loss=0.36, over 2776850.63 frames. ], batch size: 53, lr: 6.07e-03, grad_scale: 16.0
2024-09-23 21:13:50,585 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=8.44 vs. limit=15.0
2024-09-23 21:13:51,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=347078.6666666667, ans=0.2
2024-09-23 21:14:08,665 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.225e+02 1.312e+02 1.421e+02 1.795e+02, threshold=2.625e+02, percent-clipped=0.0
2024-09-23 21:14:18,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=8.44 vs. limit=12.0
2024-09-23 21:15:12,642 INFO [train.py:1198] (1/4) Epoch 20, batch 400, loss[loss=0.2269, ctc_loss=0.1534, cr_loss=0.3673, over 17138.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1437, cr_loss=0.3588, over 2912773.56 frames. ], batch size: 48, lr: 6.06e-03, grad_scale: 32.0
2024-09-23 21:15:22,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=347312.0, ans=0.2
2024-09-23 21:15:28,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=347358.6666666667, ans=0.125
2024-09-23 21:16:29,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=347498.6666666667, ans=0.0
2024-09-23 21:16:32,400 INFO [train.py:1198] (1/4) Epoch 20, batch 450, loss[loss=0.2158, ctc_loss=0.144, cr_loss=0.3587, over 17219.00 frames. ], tot_loss[loss=0.2146, ctc_loss=0.1431, cr_loss=0.3576, over 3010882.78 frames. ], batch size: 47, lr: 6.06e-03, grad_scale: 32.0
2024-09-23 21:16:34,833 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.74 vs. limit=22.5
2024-09-23 21:16:52,900 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.293e+02 1.421e+02 1.585e+02 2.168e+02, threshold=2.842e+02, percent-clipped=0.0
2024-09-23 21:17:10,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.min_positive, batch_count=347638.6666666667, ans=0.025
2024-09-23 21:17:44,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=347732.0, ans=0.0
2024-09-23 21:17:48,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=347732.0, ans=0.1
2024-09-23 21:17:49,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1.whitening_limit, batch_count=347732.0, ans=10.0
2024-09-23 21:17:54,852 INFO [train.py:1198] (1/4) Epoch 20, batch 500, loss[loss=0.2341, ctc_loss=0.1583, cr_loss=0.3791, over 17079.00 frames. ], tot_loss[loss=0.214, ctc_loss=0.1426, cr_loss=0.3569, over 3098797.48 frames. ], batch size: 49, lr: 6.06e-03, grad_scale: 16.0
2024-09-23 21:18:21,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=347825.3333333333, ans=0.5
2024-09-23 21:18:31,833 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.45 vs. limit=12.0
2024-09-23 21:18:53,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=347918.6666666667, ans=0.0
2024-09-23 21:19:17,192 INFO [train.py:1198] (1/4) Epoch 20, batch 550, loss[loss=0.1753, ctc_loss=0.1125, cr_loss=0.3143, over 17179.00 frames. ], tot_loss[loss=0.2133, ctc_loss=0.1419, cr_loss=0.3566, over 3153761.47 frames. ], batch size: 41, lr: 6.06e-03, grad_scale: 16.0
2024-09-23 21:19:42,496 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.238e+02 1.328e+02 1.432e+02 2.472e+02, threshold=2.656e+02, percent-clipped=0.0
2024-09-23 21:19:44,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.34 vs. limit=15.0
2024-09-23 21:19:54,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=348105.3333333333, ans=0.07
2024-09-23 21:20:16,722 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.81 vs. limit=22.5
2024-09-23 21:20:41,716 INFO [train.py:1198] (1/4) Epoch 20, batch 600, loss[loss=0.2062, ctc_loss=0.1337, cr_loss=0.3625, over 17144.00 frames. ], tot_loss[loss=0.2135, ctc_loss=0.1422, cr_loss=0.3566, over 3194581.05 frames. ], batch size: 48, lr: 6.06e-03, grad_scale: 16.0
2024-09-23 21:20:49,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=348245.3333333333, ans=0.125
2024-09-23 21:20:49,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=348245.3333333333, ans=0.125
2024-09-23 21:20:56,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=348292.0, ans=0.0
2024-09-23 21:21:02,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=348292.0, ans=0.125
2024-09-23 21:21:13,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=348338.6666666667, ans=0.125
2024-09-23 21:21:20,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=348338.6666666667, ans=0.125
2024-09-23 21:21:41,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten.whitening_limit, batch_count=348385.3333333333, ans=15.0
2024-09-23 21:21:45,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=348432.0, ans=0.125
2024-09-23 21:21:56,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=348432.0, ans=0.125
2024-09-23 21:22:01,009 INFO [train.py:1198] (1/4) Epoch 20, batch 650, loss[loss=0.1733, ctc_loss=0.1133, cr_loss=0.3, over 17090.00 frames. ], tot_loss[loss=0.2141, ctc_loss=0.1426, cr_loss=0.3573, over 3228198.53 frames. ], batch size: 43, lr: 6.05e-03, grad_scale: 16.0
2024-09-23 21:22:21,604 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.264e+02 1.348e+02 1.477e+02 2.156e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-23 21:22:31,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=348572.0, ans=0.2
2024-09-23 21:22:48,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=348572.0, ans=0.125
2024-09-23 21:22:49,774 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.34 vs. limit=8.0
2024-09-23 21:22:50,591 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.47 vs. limit=12.0
2024-09-23 21:23:25,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=348712.0, ans=0.0
2024-09-23 21:23:26,231 INFO [train.py:1198] (1/4) Epoch 20, batch 700, loss[loss=0.178, ctc_loss=0.113, cr_loss=0.3249, over 17112.00 frames. ], tot_loss[loss=0.2154, ctc_loss=0.1436, cr_loss=0.3588, over 3262701.57 frames. ], batch size: 40, lr: 6.05e-03, grad_scale: 16.0
2024-09-23 21:23:38,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=348712.0, ans=0.125
2024-09-23 21:24:01,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=348805.3333333333, ans=0.0
2024-09-23 21:24:17,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=348852.0, ans=0.125
2024-09-23 21:24:17,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=348852.0, ans=0.0
2024-09-23 21:24:51,549 INFO [train.py:1198] (1/4) Epoch 20, batch 750, loss[loss=0.1815, ctc_loss=0.1198, cr_loss=0.3085, over 17288.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1437, cr_loss=0.359, over 3284583.57 frames. ], batch size: 42, lr: 6.05e-03, grad_scale: 16.0
2024-09-23 21:24:57,202 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=11.44 vs. limit=15.0
2024-09-23 21:25:12,208 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.238e+02 1.344e+02 1.442e+02 2.138e+02, threshold=2.688e+02, percent-clipped=0.0
2024-09-23 21:25:16,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.32 vs. limit=6.0
2024-09-23 21:25:29,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=349038.6666666667, ans=0.2
2024-09-23 21:26:11,141 INFO [train.py:1198] (1/4) Epoch 20, batch 800, loss[loss=0.1911, ctc_loss=0.126, cr_loss=0.3257, over 17075.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1436, cr_loss=0.3595, over 3310366.67 frames. ], batch size: 43, lr: 6.05e-03, grad_scale: 32.0
2024-09-23 21:26:30,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=349225.3333333333, ans=0.0
2024-09-23 21:26:40,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=349225.3333333333, ans=0.1
2024-09-23 21:26:40,517 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=6.74 vs. limit=15.0
2024-09-23 21:26:41,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=349272.0, ans=0.125
2024-09-23 21:26:50,403 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.94 vs. limit=6.0
2024-09-23 21:26:59,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=349318.6666666667, ans=0.125
2024-09-23 21:27:10,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=349318.6666666667, ans=0.1
2024-09-23 21:27:29,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=349412.0, ans=0.025
2024-09-23 21:27:31,181 INFO [train.py:1198] (1/4) Epoch 20, batch 850, loss[loss=0.2108, ctc_loss=0.1399, cr_loss=0.3546, over 17200.00 frames. ], tot_loss[loss=0.2168, ctc_loss=0.1445, cr_loss=0.3617, over 3325442.88 frames. ], batch size: 47, lr: 6.05e-03, grad_scale: 32.0
2024-09-23 21:27:54,463 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.268e+02 1.363e+02 1.494e+02 2.147e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-23 21:28:20,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.59 vs. limit=15.0
2024-09-23 21:28:30,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=349552.0, ans=0.125
2024-09-23 21:28:56,003 INFO [train.py:1198] (1/4) Epoch 20, batch 900, loss[loss=0.205, ctc_loss=0.1333, cr_loss=0.3583, over 17268.00 frames. ], tot_loss[loss=0.217, ctc_loss=0.1446, cr_loss=0.362, over 3339935.08 frames. ], batch size: 44, lr: 6.04e-03, grad_scale: 32.0
2024-09-23 21:29:10,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=349692.0, ans=0.125
2024-09-23 21:29:16,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=349692.0, ans=0.1
2024-09-23 21:29:38,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=349738.6666666667, ans=0.0
2024-09-23 21:29:55,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=349785.3333333333, ans=0.1
2024-09-23 21:30:14,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.63 vs. limit=22.5
2024-09-23 21:30:21,167 INFO [train.py:1198] (1/4) Epoch 20, batch 950, loss[loss=0.2421, ctc_loss=0.1607, cr_loss=0.4071, over 17029.00 frames. ], tot_loss[loss=0.2177, ctc_loss=0.1452, cr_loss=0.3628, over 3341758.57 frames. ], batch size: 52, lr: 6.04e-03, grad_scale: 32.0
2024-09-23 21:30:42,122 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.279e+02 1.392e+02 1.519e+02 1.959e+02, threshold=2.784e+02, percent-clipped=0.0
2024-09-23 21:30:50,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=349925.3333333333, ans=0.0
2024-09-23 21:30:52,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=349972.0, ans=0.2
2024-09-23 21:31:20,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=350018.6666666667, ans=0.025
2024-09-23 21:31:33,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=350065.3333333333, ans=0.0
2024-09-23 21:31:41,386 INFO [train.py:1198] (1/4) Epoch 20, batch 1000, loss[loss=0.2841, ctc_loss=0.202, cr_loss=0.4108, over 11518.00 frames. ], tot_loss[loss=0.2186, ctc_loss=0.1459, cr_loss=0.3633, over 3342814.96 frames. ], batch size: 123, lr: 6.04e-03, grad_scale: 32.0
2024-09-23 21:31:44,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=350112.0, ans=0.125
2024-09-23 21:32:12,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=350205.3333333333, ans=0.2
2024-09-23 21:32:33,481 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.94 vs. limit=15.0
2024-09-23 21:32:39,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=350252.0, ans=0.025
2024-09-23 21:32:46,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.14 vs. limit=15.0
2024-09-23 21:32:49,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=350298.6666666667, ans=0.0
2024-09-23 21:33:04,840 INFO [train.py:1198] (1/4) Epoch 20, batch 1050, loss[loss=0.2147, ctc_loss=0.1425, cr_loss=0.3609, over 17226.00 frames. ], tot_loss[loss=0.2179, ctc_loss=0.1453, cr_loss=0.3628, over 3347329.43 frames. ], batch size: 50, lr: 6.04e-03, grad_scale: 32.0
2024-09-23 21:33:23,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=350392.0, ans=0.0
2024-09-23 21:33:27,842 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.295e+02 1.374e+02 1.534e+02 2.501e+02, threshold=2.748e+02, percent-clipped=0.0
2024-09-23 21:33:41,635 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.59 vs. limit=15.0
2024-09-23 21:33:44,646 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.00 vs. limit=6.0
2024-09-23 21:33:45,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=350438.6666666667, ans=0.125
2024-09-23 21:34:00,293 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.92 vs. limit=15.0
2024-09-23 21:34:10,313 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.22 vs. limit=22.5
2024-09-23 21:34:32,260 INFO [train.py:1198] (1/4) Epoch 20, batch 1100, loss[loss=0.2608, ctc_loss=0.1802, cr_loss=0.403, over 14820.00 frames. ], tot_loss[loss=0.2175, ctc_loss=0.1451, cr_loss=0.3617, over 3338058.63 frames. ], batch size: 89, lr: 6.04e-03, grad_scale: 16.0
2024-09-23 21:34:41,461 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.88 vs. limit=10.0
2024-09-23 21:34:51,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=350625.3333333333, ans=0.0
2024-09-23 21:35:01,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=350625.3333333333, ans=0.04949747468305833
2024-09-23 21:35:52,149 INFO [train.py:1198] (1/4) Epoch 20, batch 1150, loss[loss=0.2323, ctc_loss=0.1557, cr_loss=0.3827, over 16989.00 frames. ], tot_loss[loss=0.2169, ctc_loss=0.1446, cr_loss=0.3615, over 3347523.71 frames. ], batch size: 53, lr: 6.03e-03, grad_scale: 16.0
2024-09-23 21:36:14,527 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.209e+02 1.298e+02 1.420e+02 2.385e+02, threshold=2.595e+02, percent-clipped=0.0
2024-09-23 21:36:19,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=350858.6666666667, ans=0.0
2024-09-23 21:36:21,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=350858.6666666667, ans=0.0
2024-09-23 21:36:22,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=350905.3333333333, ans=0.125
2024-09-23 21:36:22,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=350905.3333333333, ans=0.0
2024-09-23 21:36:25,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=350905.3333333333, ans=0.125
2024-09-23 21:36:32,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=350905.3333333333, ans=0.125
2024-09-23 21:36:41,127 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.08 vs. limit=10.0
2024-09-23 21:36:42,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=350952.0, ans=0.2
2024-09-23 21:36:43,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=350952.0, ans=0.0
2024-09-23 21:36:55,615 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.98 vs. limit=12.0
2024-09-23 21:37:01,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=350998.6666666667, ans=0.0
2024-09-23 21:37:10,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=351045.3333333333, ans=0.125
2024-09-23 21:37:12,176 INFO [train.py:1198] (1/4) Epoch 20, batch 1200, loss[loss=0.2671, ctc_loss=0.1871, cr_loss=0.3997, over 12037.00 frames. ], tot_loss[loss=0.2167, ctc_loss=0.1446, cr_loss=0.3605, over 3344570.03 frames. ], batch size: 123, lr: 6.03e-03, grad_scale: 32.0
2024-09-23 21:37:45,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=351138.6666666667, ans=0.0
2024-09-23 21:38:19,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=351232.0, ans=0.125
2024-09-23 21:38:26,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=351232.0, ans=0.2
2024-09-23 21:38:28,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=351232.0, ans=0.125
2024-09-23 21:38:32,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=351232.0, ans=0.125
2024-09-23 21:38:37,551 INFO [train.py:1198] (1/4) Epoch 20, batch 1250, loss[loss=0.2213, ctc_loss=0.1479, cr_loss=0.3672, over 17001.00 frames. ], tot_loss[loss=0.2176, ctc_loss=0.1452, cr_loss=0.3619, over 3347984.49 frames. ], batch size: 53, lr: 6.03e-03, grad_scale: 32.0
2024-09-23 21:38:50,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=351278.6666666667, ans=0.125
2024-09-23 21:38:53,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=351325.3333333333, ans=0.125
2024-09-23 21:38:59,802 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.265e+02 1.387e+02 1.557e+02 1.898e+02, threshold=2.775e+02, percent-clipped=0.0
2024-09-23 21:39:21,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=351372.0, ans=0.1
2024-09-23 21:40:02,003 INFO [train.py:1198] (1/4) Epoch 20, batch 1300, loss[loss=0.1966, ctc_loss=0.1302, cr_loss=0.3323, over 17076.00 frames. ], tot_loss[loss=0.2169, ctc_loss=0.1448, cr_loss=0.3605, over 3350188.75 frames. ], batch size: 43, lr: 6.03e-03, grad_scale: 32.0
2024-09-23 21:40:06,130 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.79 vs. limit=22.5
2024-09-23 21:40:12,200 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.38 vs. limit=6.0
2024-09-23 21:40:14,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.63 vs. limit=15.0
2024-09-23 21:40:23,742 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.58 vs. limit=22.5
2024-09-23 21:40:35,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=351605.3333333333, ans=0.125
2024-09-23 21:40:40,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=351605.3333333333, ans=0.015
2024-09-23 21:40:40,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.98 vs. limit=6.0
2024-09-23 21:40:59,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=351652.0, ans=0.09899494936611666
2024-09-23 21:41:20,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=351745.3333333333, ans=0.125
2024-09-23 21:41:21,510 INFO [train.py:1198] (1/4) Epoch 20, batch 1350, loss[loss=0.2086, ctc_loss=0.1407, cr_loss=0.3396, over 17066.00 frames. ], tot_loss[loss=0.2161, ctc_loss=0.1442, cr_loss=0.3593, over 3341376.66 frames. ], batch size: 46, lr: 6.03e-03, grad_scale: 32.0
2024-09-23 21:41:23,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=351745.3333333333, ans=0.1
2024-09-23 21:41:25,030 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 21:41:29,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=351745.3333333333, ans=0.125
2024-09-23 21:41:40,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=351792.0, ans=0.125
2024-09-23 21:41:43,587 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.285e+02 1.356e+02 1.494e+02 3.061e+02, threshold=2.711e+02, percent-clipped=1.0
2024-09-23 21:41:48,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=351792.0, ans=0.1
2024-09-23 21:42:04,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=351838.6666666667, ans=0.125
2024-09-23 21:42:12,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=351885.3333333333, ans=0.125
2024-09-23 21:42:25,331 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.98 vs. limit=12.0
2024-09-23 21:42:28,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=351932.0, ans=0.125
2024-09-23 21:42:38,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=351932.0, ans=0.125
2024-09-23 21:42:43,486 INFO [train.py:1198] (1/4) Epoch 20, batch 1400, loss[loss=0.1657, ctc_loss=0.1079, cr_loss=0.2891, over 17027.00 frames. ], tot_loss[loss=0.2163, ctc_loss=0.1444, cr_loss=0.3597, over 3346744.60 frames. ], batch size: 39, lr: 6.02e-03, grad_scale: 32.0
2024-09-23 21:42:46,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=351978.6666666667, ans=0.125
2024-09-23 21:43:03,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=352025.3333333333, ans=0.1
2024-09-23 21:43:42,589 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.71 vs. limit=15.0
2024-09-23 21:43:49,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=352165.3333333333, ans=0.125
2024-09-23 21:43:51,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=352165.3333333333, ans=0.0
2024-09-23 21:43:51,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=352165.3333333333, ans=0.125
2024-09-23 21:43:56,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=352165.3333333333, ans=0.125
2024-09-23 21:44:00,747 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.80 vs. limit=10.0
2024-09-23 21:44:08,212 INFO [train.py:1198] (1/4) Epoch 20, batch 1450, loss[loss=0.2464, ctc_loss=0.167, cr_loss=0.3972, over 16901.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1438, cr_loss=0.3585, over 3350985.59 frames. ], batch size: 58, lr: 6.02e-03, grad_scale: 32.0
2024-09-23 21:44:17,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=352212.0, ans=0.1
2024-09-23 21:44:18,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=352212.0, ans=0.2
2024-09-23 21:44:22,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=352212.0, ans=0.1
2024-09-23 21:44:32,962 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.276e+02 1.364e+02 1.479e+02 2.768e+02, threshold=2.727e+02, percent-clipped=1.0
2024-09-23 21:44:39,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=352258.6666666667, ans=0.2
2024-09-23 21:44:58,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=352352.0, ans=0.125
2024-09-23 21:45:30,181 INFO [train.py:1198] (1/4) Epoch 20, batch 1500, loss[loss=0.2301, ctc_loss=0.1532, cr_loss=0.3845, over 17228.00 frames. ], tot_loss[loss=0.2167, ctc_loss=0.1448, cr_loss=0.3593, over 3331671.12 frames. ], batch size: 55, lr: 6.02e-03, grad_scale: 16.0
2024-09-23 21:45:32,723 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.00 vs. limit=22.5
2024-09-23 21:46:06,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten.whitening_limit, batch_count=352538.6666666667, ans=22.5
2024-09-23 21:46:10,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=352538.6666666667, ans=0.125
2024-09-23 21:46:12,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=352538.6666666667, ans=0.125
2024-09-23 21:46:21,179 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.57 vs. limit=15.0
2024-09-23 21:46:26,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=352585.3333333333, ans=0.125
2024-09-23 21:46:26,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=352585.3333333333, ans=0.09899494936611666
2024-09-23 21:46:51,135 INFO [train.py:1198] (1/4) Epoch 20, batch 1550, loss[loss=0.1922, ctc_loss=0.1271, cr_loss=0.3252, over 17082.00 frames. ], tot_loss[loss=0.2162, ctc_loss=0.1444, cr_loss=0.3588, over 3336003.17 frames. ], batch size: 40, lr: 6.02e-03, grad_scale: 16.0
2024-09-23 21:46:56,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=352678.6666666667, ans=0.1
2024-09-23 21:47:05,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=352725.3333333333, ans=0.125
2024-09-23 21:47:05,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=352725.3333333333, ans=0.05
2024-09-23 21:47:17,719 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.019e+02 1.276e+02 1.387e+02 1.513e+02 2.213e+02, threshold=2.774e+02, percent-clipped=0.0
2024-09-23 21:48:06,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.97 vs. limit=22.5
2024-09-23 21:48:07,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=352865.3333333333, ans=0.0
2024-09-23 21:48:16,279 INFO [train.py:1198] (1/4) Epoch 20, batch 1600, loss[loss=0.2521, ctc_loss=0.1709, cr_loss=0.4059, over 16012.00 frames. ], tot_loss[loss=0.2167, ctc_loss=0.1447, cr_loss=0.3599, over 3343673.99 frames. ], batch size: 74, lr: 6.02e-03, grad_scale: 32.0
2024-09-23 21:48:26,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=352912.0, ans=0.1
2024-09-23 21:48:37,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=352958.6666666667, ans=0.125
2024-09-23 21:48:42,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=352958.6666666667, ans=0.0
2024-09-23 21:48:45,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=352958.6666666667, ans=0.025
2024-09-23 21:49:15,782 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 21:49:38,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=353098.6666666667, ans=0.125
2024-09-23 21:49:41,003 INFO [train.py:1198] (1/4) Epoch 20, batch 1650, loss[loss=0.1762, ctc_loss=0.1175, cr_loss=0.2933, over 16798.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1439, cr_loss=0.3591, over 3359758.01 frames. ], batch size: 37, lr: 6.02e-03, grad_scale: 32.0
2024-09-23 21:49:59,154 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 21:50:05,308 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.288e+02 1.374e+02 1.505e+02 2.586e+02, threshold=2.748e+02, percent-clipped=0.0
2024-09-23 21:50:23,668 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.24 vs. limit=15.0
2024-09-23 21:50:28,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=353285.3333333333, ans=0.125
2024-09-23 21:50:35,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=353285.3333333333, ans=0.125
2024-09-23 21:50:43,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=353332.0, ans=0.95
2024-09-23 21:50:45,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=353332.0, ans=0.5
2024-09-23 21:50:50,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=353332.0, ans=0.125
2024-09-23 21:51:01,006 INFO [train.py:1198] (1/4) Epoch 20, batch 1700, loss[loss=0.2247, ctc_loss=0.1527, cr_loss=0.3599, over 16070.00 frames. ], tot_loss[loss=0.2179, ctc_loss=0.1455, cr_loss=0.3622, over 3361731.32 frames. ], batch size: 74, lr: 6.01e-03, grad_scale: 32.0
2024-09-23 21:51:11,128 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 21:51:20,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=353425.3333333333, ans=0.125
2024-09-23 21:52:00,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=353518.6666666667, ans=0.0
2024-09-23 21:52:11,851 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.49 vs. limit=10.0
2024-09-23 21:52:24,263 INFO [train.py:1198] (1/4) Epoch 20, batch 1750, loss[loss=0.2162, ctc_loss=0.1459, cr_loss=0.3513, over 16928.00 frames. ], tot_loss[loss=0.2173, ctc_loss=0.145, cr_loss=0.3614, over 3366955.12 frames. ], batch size: 58, lr: 6.01e-03, grad_scale: 16.0
2024-09-23 21:52:49,738 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.235e+02 1.321e+02 1.434e+02 2.366e+02, threshold=2.642e+02, percent-clipped=0.0
2024-09-23 21:52:51,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=353658.6666666667, ans=0.1
2024-09-23 21:52:56,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=353705.3333333333, ans=0.2
2024-09-23 21:53:47,292 INFO [train.py:1198] (1/4) Epoch 20, batch 1800, loss[loss=0.2448, ctc_loss=0.1646, cr_loss=0.4012, over 15022.00 frames. ], tot_loss[loss=0.2162, ctc_loss=0.1441, cr_loss=0.3602, over 3374078.06 frames. ], batch size: 89, lr: 6.01e-03, grad_scale: 16.0
2024-09-23 21:54:03,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=353845.3333333333, ans=0.125
2024-09-23 21:54:53,544 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 21:55:07,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=354032.0, ans=10.0
2024-09-23 21:55:12,166 INFO [train.py:1198] (1/4) Epoch 20, batch 1850, loss[loss=0.2439, ctc_loss=0.166, cr_loss=0.3896, over 15850.00 frames. ], tot_loss[loss=0.2163, ctc_loss=0.1443, cr_loss=0.3601, over 3371006.31 frames. ], batch size: 74, lr: 6.01e-03, grad_scale: 16.0
2024-09-23 21:55:13,058 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.34 vs. limit=15.0
2024-09-23 21:55:35,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=354125.3333333333, ans=0.125
2024-09-23 21:55:35,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=354125.3333333333, ans=0.0
2024-09-23 21:55:37,927 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.237e+02 1.317e+02 1.410e+02 2.955e+02, threshold=2.633e+02, percent-clipped=1.0
2024-09-23 21:55:38,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=354125.3333333333, ans=0.125
2024-09-23 21:55:52,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=354172.0, ans=0.125
2024-09-23 21:56:02,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=354218.6666666667, ans=0.125
2024-09-23 21:56:02,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.93 vs. limit=15.0
2024-09-23 21:56:10,388 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.21 vs. limit=15.0
2024-09-23 21:56:32,457 INFO [train.py:1198] (1/4) Epoch 20, batch 1900, loss[loss=0.231, ctc_loss=0.1525, cr_loss=0.3924, over 17032.00 frames. ], tot_loss[loss=0.2168, ctc_loss=0.1446, cr_loss=0.3609, over 3365886.05 frames. ], batch size: 56, lr: 6.01e-03, grad_scale: 16.0
2024-09-23 21:57:31,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=354452.0, ans=0.1
2024-09-23 21:57:42,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=354498.6666666667, ans=0.125
2024-09-23 21:57:55,486 INFO [train.py:1198] (1/4) Epoch 20, batch 1950, loss[loss=0.2336, ctc_loss=0.1573, cr_loss=0.3814, over 16498.00 frames. ], tot_loss[loss=0.2164, ctc_loss=0.1443, cr_loss=0.3606, over 3371996.99 frames. ], batch size: 66, lr: 6.00e-03, grad_scale: 16.0
2024-09-23 21:58:23,509 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.283e+02 1.409e+02 1.574e+02 2.318e+02, threshold=2.818e+02, percent-clipped=0.0
2024-09-23 21:58:37,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=354638.6666666667, ans=0.0
2024-09-23 21:58:52,704 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=6.30 vs. limit=12.0
2024-09-23 21:59:11,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=354732.0, ans=0.1
2024-09-23 21:59:25,720 INFO [train.py:1198] (1/4) Epoch 20, batch 2000, loss[loss=0.2205, ctc_loss=0.148, cr_loss=0.3625, over 16997.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1435, cr_loss=0.3598, over 3374138.51 frames. ], batch size: 53, lr: 6.00e-03, grad_scale: 32.0
2024-09-23 21:59:45,665 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.09 vs. limit=22.5
2024-09-23 21:59:48,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=354825.3333333333, ans=0.125
2024-09-23 22:00:01,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=354872.0, ans=0.125
2024-09-23 22:00:36,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=354965.3333333333, ans=0.125
2024-09-23 22:00:41,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=354965.3333333333, ans=0.125
2024-09-23 22:00:42,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=354965.3333333333, ans=0.125
2024-09-23 22:00:44,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=355012.0, ans=0.0
2024-09-23 22:00:45,850 INFO [train.py:1198] (1/4) Epoch 20, batch 2050, loss[loss=0.1803, ctc_loss=0.1187, cr_loss=0.3081, over 17230.00 frames. ], tot_loss[loss=0.2162, ctc_loss=0.1441, cr_loss=0.3605, over 3369715.98 frames. ], batch size: 47, lr: 6.00e-03, grad_scale: 32.0
2024-09-23 22:00:59,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=355012.0, ans=0.025
2024-09-23 22:01:04,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=355058.6666666667, ans=0.125
2024-09-23 22:01:08,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=355058.6666666667, ans=0.05
2024-09-23 22:01:10,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=355058.6666666667, ans=0.125
2024-09-23 22:01:11,577 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.282e+02 1.359e+02 1.463e+02 2.527e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-23 22:01:31,305 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=11.79 vs. limit=22.5
2024-09-23 22:01:54,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=355198.6666666667, ans=0.035
2024-09-23 22:02:04,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=355245.3333333333, ans=0.1
2024-09-23 22:02:05,991 INFO [train.py:1198] (1/4) Epoch 20, batch 2100, loss[loss=0.1957, ctc_loss=0.1294, cr_loss=0.3314, over 17173.00 frames. ], tot_loss[loss=0.2162, ctc_loss=0.144, cr_loss=0.3608, over 3372241.49 frames. ], batch size: 41, lr: 6.00e-03, grad_scale: 32.0
2024-09-23 22:02:06,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=355245.3333333333, ans=0.125
2024-09-23 22:02:48,887 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=9.21 vs. limit=15.0
2024-09-23 22:03:13,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=355432.0, ans=0.2
2024-09-23 22:03:16,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=355432.0, ans=0.125
2024-09-23 22:03:16,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=355432.0, ans=0.0
2024-09-23 22:03:19,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=355432.0, ans=0.125
2024-09-23 22:03:20,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=355432.0, ans=0.125
2024-09-23 22:03:20,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=355432.0, ans=0.2
2024-09-23 22:03:22,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=355432.0, ans=0.125
2024-09-23 22:03:30,233 INFO [train.py:1198] (1/4) Epoch 20, batch 2150, loss[loss=0.2603, ctc_loss=0.1801, cr_loss=0.4009, over 17021.00 frames. ], tot_loss[loss=0.2168, ctc_loss=0.1444, cr_loss=0.3621, over 3374353.91 frames. ], batch size: 52, lr: 6.00e-03, grad_scale: 32.0
2024-09-23 22:03:49,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=355525.3333333333, ans=0.0
2024-09-23 22:03:49,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=355525.3333333333, ans=0.125
2024-09-23 22:03:58,362 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.263e+02 1.377e+02 1.523e+02 2.016e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-23 22:03:58,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=355525.3333333333, ans=0.125
2024-09-23 22:04:00,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=355525.3333333333, ans=0.2
2024-09-23 22:04:55,855 INFO [train.py:1198] (1/4) Epoch 20, batch 2200, loss[loss=0.175, ctc_loss=0.1131, cr_loss=0.3093, over 17174.00 frames. ], tot_loss[loss=0.2153, ctc_loss=0.1433, cr_loss=0.3598, over 3365979.59 frames. ], batch size: 41, lr: 5.99e-03, grad_scale: 32.0
2024-09-23 22:05:02,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=355712.0, ans=0.04949747468305833
2024-09-23 22:05:09,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=355712.0, ans=0.0
2024-09-23 22:05:21,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=355758.6666666667, ans=0.0
2024-09-23 22:05:36,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=355805.3333333333, ans=0.0
2024-09-23 22:06:05,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=355898.6666666667, ans=0.125
2024-09-23 22:06:10,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=355898.6666666667, ans=0.125
2024-09-23 22:06:14,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=355945.3333333333, ans=0.125
2024-09-23 22:06:14,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=355945.3333333333, ans=0.025
2024-09-23 22:06:16,040 INFO [train.py:1198] (1/4) Epoch 20, batch 2250, loss[loss=0.2191, ctc_loss=0.1471, cr_loss=0.3597, over 16905.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1435, cr_loss=0.3601, over 3370246.78 frames. ], batch size: 58, lr: 5.99e-03, grad_scale: 32.0
2024-09-23 22:06:41,143 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.99 vs. limit=15.0
2024-09-23 22:06:41,714 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.265e+02 1.369e+02 1.505e+02 1.904e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-23 22:07:01,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=356038.6666666667, ans=0.125
2024-09-23 22:07:07,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=356085.3333333333, ans=0.125
2024-09-23 22:07:20,184 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.06 vs. limit=22.5
2024-09-23 22:07:24,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=356132.0, ans=0.1
2024-09-23 22:07:30,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=356132.0, ans=0.125
2024-09-23 22:07:38,556 INFO [train.py:1198] (1/4) Epoch 20, batch 2300, loss[loss=0.2411, ctc_loss=0.1608, cr_loss=0.4014, over 15949.00 frames. ], tot_loss[loss=0.214, ctc_loss=0.1424, cr_loss=0.3581, over 3371016.90 frames. ], batch size: 74, lr: 5.99e-03, grad_scale: 32.0
2024-09-23 22:07:51,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=356178.6666666667, ans=0.0
2024-09-23 22:07:53,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=356225.3333333333, ans=0.125
2024-09-23 22:07:54,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=356225.3333333333, ans=0.2
2024-09-23 22:08:00,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=356225.3333333333, ans=0.0
2024-09-23 22:08:07,064 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.59 vs. limit=15.0
2024-09-23 22:08:27,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=356318.6666666667, ans=0.125
2024-09-23 22:08:58,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=356365.3333333333, ans=0.2
2024-09-23 22:09:02,583 INFO [train.py:1198] (1/4) Epoch 20, batch 2350, loss[loss=0.2323, ctc_loss=0.1573, cr_loss=0.375, over 17292.00 frames. ], tot_loss[loss=0.215, ctc_loss=0.1432, cr_loss=0.359, over 3365337.40 frames. ], batch size: 46, lr: 5.99e-03, grad_scale: 32.0
2024-09-23 22:09:30,700 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.236e+02 1.335e+02 1.500e+02 2.118e+02, threshold=2.671e+02, percent-clipped=0.0
2024-09-23 22:09:35,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=356505.3333333333, ans=0.2
2024-09-23 22:09:35,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=356505.3333333333, ans=0.125
2024-09-23 22:10:20,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=356598.6666666667, ans=0.07
2024-09-23 22:10:24,947 INFO [train.py:1198] (1/4) Epoch 20, batch 2400, loss[loss=0.1697, ctc_loss=0.1096, cr_loss=0.3006, over 17192.00 frames. ], tot_loss[loss=0.216, ctc_loss=0.1441, cr_loss=0.3596, over 3360642.65 frames. ], batch size: 41, lr: 5.99e-03, grad_scale: 32.0
2024-09-23 22:10:25,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=356645.3333333333, ans=0.125
2024-09-23 22:10:33,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=356645.3333333333, ans=0.125
2024-09-23 22:10:33,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer_na.min_abs, batch_count=356645.3333333333, ans=0.02
2024-09-23 22:10:54,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=356692.0, ans=0.125
2024-09-23 22:11:23,315 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 22:11:28,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=356832.0, ans=0.2
2024-09-23 22:11:29,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=356832.0, ans=0.0
2024-09-23 22:11:34,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=356832.0, ans=10.0
2024-09-23 22:11:39,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=356832.0, ans=0.1
2024-09-23 22:11:45,255 INFO [train.py:1198] (1/4) Epoch 20, batch 2450, loss[loss=0.1934, ctc_loss=0.1275, cr_loss=0.3291, over 17019.00 frames. ], tot_loss[loss=0.216, ctc_loss=0.1441, cr_loss=0.3598, over 3366388.26 frames. ], batch size: 44, lr: 5.98e-03, grad_scale: 32.0
2024-09-23 22:11:51,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=356878.6666666667, ans=0.2
2024-09-23 22:12:13,315 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.240e+02 1.349e+02 1.469e+02 2.826e+02, threshold=2.697e+02, percent-clipped=1.0
2024-09-23 22:12:23,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=356972.0, ans=0.2
2024-09-23 22:12:32,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=356972.0, ans=0.125
2024-09-23 22:12:43,112 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.29 vs. limit=15.0
2024-09-23 22:12:56,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=357065.3333333333, ans=0.125
2024-09-23 22:13:04,822 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.91 vs. limit=15.0
2024-09-23 22:13:10,054 INFO [train.py:1198] (1/4) Epoch 20, batch 2500, loss[loss=0.181, ctc_loss=0.1174, cr_loss=0.3181, over 17274.00 frames. ], tot_loss[loss=0.2158, ctc_loss=0.1439, cr_loss=0.3594, over 3350999.31 frames. ], batch size: 44, lr: 5.98e-03, grad_scale: 32.0
2024-09-23 22:13:19,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=357112.0, ans=0.2
2024-09-23 22:14:34,665 INFO [train.py:1198] (1/4) Epoch 20, batch 2550, loss[loss=0.1773, ctc_loss=0.1159, cr_loss=0.3073, over 17099.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1438, cr_loss=0.3595, over 3356189.85 frames. ], batch size: 40, lr: 5.98e-03, grad_scale: 32.0
2024-09-23 22:14:41,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=357345.3333333333, ans=0.1
2024-09-23 22:14:55,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=357392.0, ans=0.2
2024-09-23 22:15:00,261 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.269e+02 1.372e+02 1.542e+02 2.100e+02, threshold=2.744e+02, percent-clipped=0.0
2024-09-23 22:15:18,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=357438.6666666667, ans=0.2
2024-09-23 22:15:20,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=357438.6666666667, ans=10.0
2024-09-23 22:15:24,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=357485.3333333333, ans=0.0
2024-09-23 22:15:31,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=357485.3333333333, ans=0.1
2024-09-23 22:15:34,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=357485.3333333333, ans=0.2
2024-09-23 22:15:37,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=357532.0, ans=0.0
2024-09-23 22:15:39,740 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.84 vs. limit=15.0
2024-09-23 22:15:54,811 INFO [train.py:1198] (1/4) Epoch 20, batch 2600, loss[loss=0.2274, ctc_loss=0.1506, cr_loss=0.3837, over 17020.00 frames. ], tot_loss[loss=0.2164, ctc_loss=0.1443, cr_loss=0.3602, over 3347318.99 frames. ], batch size: 51, lr: 5.98e-03, grad_scale: 32.0
2024-09-23 22:16:03,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=357578.6666666667, ans=0.2
2024-09-23 22:16:06,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=357578.6666666667, ans=0.125
2024-09-23 22:16:06,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=357578.6666666667, ans=0.0
2024-09-23 22:16:08,701 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=7.15 vs. limit=15.0
2024-09-23 22:16:23,852 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.14 vs. limit=8.0
2024-09-23 22:16:26,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=357672.0, ans=0.1
2024-09-23 22:16:30,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=357672.0, ans=0.04949747468305833
2024-09-23 22:16:40,812 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.66 vs. limit=15.0
2024-09-23 22:17:13,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=357765.3333333333, ans=0.125
2024-09-23 22:17:14,001 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.23 vs. limit=15.0
2024-09-23 22:17:17,931 INFO [train.py:1198] (1/4) Epoch 20, batch 2650, loss[loss=0.2026, ctc_loss=0.1372, cr_loss=0.3268, over 17297.00 frames. ], tot_loss[loss=0.2166, ctc_loss=0.1445, cr_loss=0.3608, over 3353348.71 frames. ], batch size: 51, lr: 5.98e-03, grad_scale: 32.0
2024-09-23 22:17:39,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=357858.6666666667, ans=0.125
2024-09-23 22:17:40,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=357858.6666666667, ans=0.1
2024-09-23 22:17:43,557 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.263e+02 1.370e+02 1.503e+02 2.130e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-23 22:18:05,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=357905.3333333333, ans=0.0
2024-09-23 22:18:15,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=357952.0, ans=0.2
2024-09-23 22:18:29,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=357998.6666666667, ans=0.2
2024-09-23 22:18:43,118 INFO [train.py:1198] (1/4) Epoch 20, batch 2700, loss[loss=0.2449, ctc_loss=0.1717, cr_loss=0.3659, over 11841.00 frames. ], tot_loss[loss=0.2169, ctc_loss=0.1448, cr_loss=0.3602, over 3330398.76 frames. ], batch size: 123, lr: 5.97e-03, grad_scale: 32.0
2024-09-23 22:18:53,329 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.18 vs. limit=15.0
2024-09-23 22:19:03,036 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 22:19:27,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2.whitening_limit, batch_count=358138.6666666667, ans=15.0
2024-09-23 22:20:05,466 INFO [train.py:1198] (1/4) Epoch 20, batch 2750, loss[loss=0.2278, ctc_loss=0.1532, cr_loss=0.373, over 17137.00 frames. ], tot_loss[loss=0.2159, ctc_loss=0.1441, cr_loss=0.359, over 3331644.43 frames. ], batch size: 48, lr: 5.97e-03, grad_scale: 32.0
2024-09-23 22:20:09,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=358278.6666666667, ans=0.1
2024-09-23 22:20:12,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=358278.6666666667, ans=0.0
2024-09-23 22:20:15,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=358278.6666666667, ans=0.0
2024-09-23 22:20:18,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=358278.6666666667, ans=0.0
2024-09-23 22:20:31,065 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.266e+02 1.360e+02 1.482e+02 2.958e+02, threshold=2.720e+02, percent-clipped=1.0
2024-09-23 22:20:36,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=358372.0, ans=0.0
2024-09-23 22:20:37,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=358372.0, ans=0.0
2024-09-23 22:20:53,212 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.93 vs. limit=10.0
2024-09-23 22:21:19,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=358465.3333333333, ans=0.125
2024-09-23 22:21:25,605 INFO [train.py:1198] (1/4) Epoch 20, batch 2800, loss[loss=0.2321, ctc_loss=0.1542, cr_loss=0.3893, over 16614.00 frames. ], tot_loss[loss=0.2167, ctc_loss=0.1445, cr_loss=0.3607, over 3335912.72 frames. ], batch size: 66, lr: 5.97e-03, grad_scale: 32.0
2024-09-23 22:22:01,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=358605.3333333333, ans=0.0
2024-09-23 22:22:20,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=358652.0, ans=0.2
2024-09-23 22:22:35,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=358698.6666666667, ans=0.0
2024-09-23 22:22:39,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=358698.6666666667, ans=0.125
2024-09-23 22:22:47,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=358698.6666666667, ans=0.1
2024-09-23 22:22:47,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=358698.6666666667, ans=0.05
2024-09-23 22:22:50,492 INFO [train.py:1198] (1/4) Epoch 20, batch 2850, loss[loss=0.1982, ctc_loss=0.1301, cr_loss=0.3406, over 17170.00 frames. ], tot_loss[loss=0.2145, ctc_loss=0.1428, cr_loss=0.3583, over 3347705.93 frames. ], batch size: 41, lr: 5.97e-03, grad_scale: 32.0
2024-09-23 22:22:54,811 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.66 vs. limit=15.0
2024-09-23 22:23:09,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=358792.0, ans=0.125
2024-09-23 22:23:15,984 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.246e+02 1.349e+02 1.409e+02 2.188e+02, threshold=2.698e+02, percent-clipped=0.0
2024-09-23 22:23:37,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=358838.6666666667, ans=0.125
2024-09-23 22:23:41,619 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.66 vs. limit=6.0
2024-09-23 22:23:55,920 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.00 vs. limit=6.0
2024-09-23 22:24:04,176 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.45 vs. limit=15.0
2024-09-23 22:24:15,304 INFO [train.py:1198] (1/4) Epoch 20, batch 2900, loss[loss=0.2187, ctc_loss=0.1464, cr_loss=0.3615, over 17345.00 frames. ], tot_loss[loss=0.2149, ctc_loss=0.1431, cr_loss=0.3593, over 3360082.74 frames. ], batch size: 48, lr: 5.97e-03, grad_scale: 32.0
2024-09-23 22:25:35,526 INFO [train.py:1198] (1/4) Epoch 20, batch 2950, loss[loss=0.1986, ctc_loss=0.1322, cr_loss=0.3321, over 17309.00 frames. ], tot_loss[loss=0.2145, ctc_loss=0.1427, cr_loss=0.3589, over 3366957.78 frames. ], batch size: 49, lr: 5.96e-03, grad_scale: 32.0
2024-09-23 22:25:57,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=359258.6666666667, ans=0.0
2024-09-23 22:26:00,732 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.258e+02 1.349e+02 1.512e+02 2.191e+02, threshold=2.699e+02, percent-clipped=0.0
2024-09-23 22:26:12,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=359305.3333333333, ans=0.125
2024-09-23 22:26:18,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=359305.3333333333, ans=0.125
2024-09-23 22:26:33,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=359352.0, ans=0.125
2024-09-23 22:26:54,936 INFO [train.py:1198] (1/4) Epoch 20, batch 3000, loss[loss=0.2226, ctc_loss=0.1485, cr_loss=0.3708, over 17281.00 frames. ], tot_loss[loss=0.2138, ctc_loss=0.1421, cr_loss=0.3584, over 3363012.88 frames. ], batch size: 51, lr: 5.96e-03, grad_scale: 32.0
2024-09-23 22:26:54,936 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 22:27:10,609 INFO [train.py:1230] (1/4) Epoch 20, validation: loss=0.03912, ctc_loss=0.03912, cr_loss=8.309e-15, over 944034.00 frames. 
2024-09-23 22:27:10,609 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 22:27:22,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=359445.3333333333, ans=0.025
2024-09-23 22:27:32,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=359492.0, ans=0.125
2024-09-23 22:28:31,108 INFO [train.py:1198] (1/4) Epoch 20, batch 3050, loss[loss=0.2058, ctc_loss=0.1332, cr_loss=0.363, over 17124.00 frames. ], tot_loss[loss=0.2144, ctc_loss=0.1426, cr_loss=0.3588, over 3360439.94 frames. ], batch size: 40, lr: 5.96e-03, grad_scale: 32.0
2024-09-23 22:28:44,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=359678.6666666667, ans=0.125
2024-09-23 22:28:56,199 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.297e+02 1.389e+02 1.525e+02 3.485e+02, threshold=2.778e+02, percent-clipped=1.0
2024-09-23 22:29:11,317 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.11 vs. limit=15.0
2024-09-23 22:29:32,598 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 22:29:49,412 INFO [train.py:1198] (1/4) Epoch 20, batch 3100, loss[loss=0.2003, ctc_loss=0.1327, cr_loss=0.3382, over 17060.00 frames. ], tot_loss[loss=0.2147, ctc_loss=0.1429, cr_loss=0.3589, over 3352077.45 frames. ], batch size: 46, lr: 5.96e-03, grad_scale: 32.0
2024-09-23 22:29:56,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys.whitening_limit, batch_count=359912.0, ans=6.0
2024-09-23 22:30:03,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=359958.6666666667, ans=0.125
2024-09-23 22:30:26,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=360005.3333333333, ans=0.025
2024-09-23 22:30:34,522 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.99 vs. limit=15.0
2024-09-23 22:31:12,470 INFO [train.py:1198] (1/4) Epoch 20, batch 3150, loss[loss=0.2142, ctc_loss=0.1389, cr_loss=0.3769, over 16974.00 frames. ], tot_loss[loss=0.2158, ctc_loss=0.1438, cr_loss=0.3603, over 3352204.76 frames. ], batch size: 53, lr: 5.96e-03, grad_scale: 32.0
2024-09-23 22:31:26,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=360192.0, ans=0.125
2024-09-23 22:31:28,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=360192.0, ans=0.125
2024-09-23 22:31:37,610 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.291e+02 1.400e+02 1.624e+02 2.307e+02, threshold=2.800e+02, percent-clipped=0.0
2024-09-23 22:31:58,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=360285.3333333333, ans=0.125
2024-09-23 22:32:31,335 INFO [train.py:1198] (1/4) Epoch 20, batch 3200, loss[loss=0.17, ctc_loss=0.1111, cr_loss=0.2943, over 17020.00 frames. ], tot_loss[loss=0.2148, ctc_loss=0.1431, cr_loss=0.3587, over 3345419.72 frames. ], batch size: 39, lr: 5.95e-03, grad_scale: 32.0
2024-09-23 22:32:38,386 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.89 vs. limit=15.0
2024-09-23 22:32:50,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=360425.3333333333, ans=0.0
2024-09-23 22:33:09,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=360472.0, ans=0.0
2024-09-23 22:33:12,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=360472.0, ans=0.0
2024-09-23 22:33:12,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=360472.0, ans=0.2
2024-09-23 22:33:20,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=360518.6666666667, ans=0.0
2024-09-23 22:33:49,550 INFO [train.py:1198] (1/4) Epoch 20, batch 3250, loss[loss=0.2173, ctc_loss=0.1452, cr_loss=0.3607, over 17209.00 frames. ], tot_loss[loss=0.2148, ctc_loss=0.143, cr_loss=0.359, over 3351401.01 frames. ], batch size: 55, lr: 5.95e-03, grad_scale: 32.0
2024-09-23 22:34:16,116 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.218e+02 1.307e+02 1.462e+02 2.065e+02, threshold=2.615e+02, percent-clipped=0.0
2024-09-23 22:34:39,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=360752.0, ans=0.2
2024-09-23 22:34:41,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=360752.0, ans=0.125
2024-09-23 22:35:01,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=360798.6666666667, ans=0.1
2024-09-23 22:35:08,070 INFO [train.py:1198] (1/4) Epoch 20, batch 3300, loss[loss=0.2042, ctc_loss=0.1337, cr_loss=0.3525, over 17289.00 frames. ], tot_loss[loss=0.2152, ctc_loss=0.1433, cr_loss=0.359, over 3355508.59 frames. ], batch size: 46, lr: 5.95e-03, grad_scale: 32.0
2024-09-23 22:35:36,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.15 vs. limit=15.0
2024-09-23 22:35:49,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=360938.6666666667, ans=0.125
2024-09-23 22:35:58,768 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.19 vs. limit=22.5
2024-09-23 22:36:02,159 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.80 vs. limit=22.5
2024-09-23 22:36:12,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=361032.0, ans=0.2
2024-09-23 22:36:26,306 INFO [train.py:1198] (1/4) Epoch 20, batch 3350, loss[loss=0.2037, ctc_loss=0.139, cr_loss=0.3233, over 17020.00 frames. ], tot_loss[loss=0.2144, ctc_loss=0.1427, cr_loss=0.3582, over 3363034.75 frames. ], batch size: 52, lr: 5.95e-03, grad_scale: 16.0
2024-09-23 22:36:31,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=361078.6666666667, ans=0.2
2024-09-23 22:36:56,430 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.040e+02 1.260e+02 1.354e+02 1.461e+02 2.333e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-23 22:37:19,322 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.10 vs. limit=12.0
2024-09-23 22:37:46,671 INFO [train.py:1198] (1/4) Epoch 20, batch 3400, loss[loss=0.2341, ctc_loss=0.1583, cr_loss=0.3787, over 17213.00 frames. ], tot_loss[loss=0.2149, ctc_loss=0.1431, cr_loss=0.3588, over 3356557.05 frames. ], batch size: 50, lr: 5.95e-03, grad_scale: 16.0
2024-09-23 22:38:19,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=361405.3333333333, ans=0.125
2024-09-23 22:38:24,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=361405.3333333333, ans=0.1
2024-09-23 22:38:30,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=361405.3333333333, ans=0.125
2024-09-23 22:38:53,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=361498.6666666667, ans=0.125
2024-09-23 22:38:58,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=361498.6666666667, ans=0.125
2024-09-23 22:38:58,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=361498.6666666667, ans=0.125
2024-09-23 22:39:06,006 INFO [train.py:1198] (1/4) Epoch 20, batch 3450, loss[loss=0.1697, ctc_loss=0.11, cr_loss=0.2988, over 17064.00 frames. ], tot_loss[loss=0.2154, ctc_loss=0.1435, cr_loss=0.3596, over 3363933.28 frames. ], batch size: 39, lr: 5.95e-03, grad_scale: 16.0
2024-09-23 22:39:09,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=361545.3333333333, ans=0.125
2024-09-23 22:39:25,483 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.67 vs. limit=22.5
2024-09-23 22:39:27,517 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.64 vs. limit=15.0
2024-09-23 22:39:34,087 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.266e+02 1.362e+02 1.520e+02 1.983e+02, threshold=2.723e+02, percent-clipped=0.0
2024-09-23 22:39:45,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=361638.6666666667, ans=0.0
2024-09-23 22:40:00,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=361685.3333333333, ans=0.2
2024-09-23 22:40:11,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=361732.0, ans=0.0
2024-09-23 22:40:16,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer_ff2.min_abs, batch_count=361732.0, ans=0.1
2024-09-23 22:40:21,866 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.13 vs. limit=6.0
2024-09-23 22:40:23,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=361778.6666666667, ans=0.125
2024-09-23 22:40:24,484 INFO [train.py:1198] (1/4) Epoch 20, batch 3500, loss[loss=0.2082, ctc_loss=0.1372, cr_loss=0.3547, over 17257.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1437, cr_loss=0.36, over 3349040.42 frames. ], batch size: 42, lr: 5.94e-03, grad_scale: 16.0
2024-09-23 22:40:24,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=361778.6666666667, ans=0.125
2024-09-23 22:40:24,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=361778.6666666667, ans=0.125
2024-09-23 22:40:45,862 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.42 vs. limit=6.0
2024-09-23 22:40:51,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=361825.3333333333, ans=0.125
2024-09-23 22:41:01,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=361872.0, ans=0.125
2024-09-23 22:41:13,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=361918.6666666667, ans=0.2
2024-09-23 22:41:39,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=361965.3333333333, ans=0.025
2024-09-23 22:41:44,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=361965.3333333333, ans=0.035
2024-09-23 22:41:47,245 INFO [train.py:1198] (1/4) Epoch 20, batch 3550, loss[loss=0.1939, ctc_loss=0.1244, cr_loss=0.3476, over 17222.00 frames. ], tot_loss[loss=0.2164, ctc_loss=0.1443, cr_loss=0.361, over 3358503.79 frames. ], batch size: 47, lr: 5.94e-03, grad_scale: 16.0
2024-09-23 22:42:15,393 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.278e+02 1.363e+02 1.489e+02 4.233e+02, threshold=2.726e+02, percent-clipped=1.0
2024-09-23 22:42:28,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=362105.3333333333, ans=0.07
2024-09-23 22:43:05,644 INFO [train.py:1198] (1/4) Epoch 20, batch 3600, loss[loss=0.192, ctc_loss=0.1257, cr_loss=0.3314, over 17282.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1436, cr_loss=0.3601, over 3362682.05 frames. ], batch size: 42, lr: 5.94e-03, grad_scale: 32.0
2024-09-23 22:43:18,541 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.96 vs. limit=12.0
2024-09-23 22:43:20,227 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=7.64 vs. limit=15.0
2024-09-23 22:43:23,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=362292.0, ans=0.035
2024-09-23 22:43:38,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=362338.6666666667, ans=0.04949747468305833
2024-09-23 22:43:50,027 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.23 vs. limit=15.0
2024-09-23 22:43:59,372 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.48 vs. limit=12.0
2024-09-23 22:44:05,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=362385.3333333333, ans=0.0
2024-09-23 22:44:23,798 INFO [train.py:1198] (1/4) Epoch 20, batch 3650, loss[loss=0.184, ctc_loss=0.121, cr_loss=0.3146, over 17104.00 frames. ], tot_loss[loss=0.2156, ctc_loss=0.1437, cr_loss=0.3599, over 3366029.82 frames. ], batch size: 43, lr: 5.94e-03, grad_scale: 32.0
2024-09-23 22:44:24,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=362478.6666666667, ans=0.1
2024-09-23 22:44:35,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=362478.6666666667, ans=0.125
2024-09-23 22:44:42,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=362525.3333333333, ans=0.125
2024-09-23 22:44:51,968 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.248e+02 1.353e+02 1.463e+02 2.228e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-23 22:45:23,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=362618.6666666667, ans=0.125
2024-09-23 22:45:30,545 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 22:45:41,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=362712.0, ans=0.125
2024-09-23 22:45:42,630 INFO [train.py:1198] (1/4) Epoch 20, batch 3700, loss[loss=0.2069, ctc_loss=0.1372, cr_loss=0.3484, over 17043.00 frames. ], tot_loss[loss=0.2162, ctc_loss=0.1442, cr_loss=0.3602, over 3350951.58 frames. ], batch size: 39, lr: 5.94e-03, grad_scale: 32.0
2024-09-23 22:45:47,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=362712.0, ans=0.025
2024-09-23 22:46:11,687 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=5.48 vs. limit=12.0
2024-09-23 22:46:32,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=362852.0, ans=0.2
2024-09-23 22:47:01,721 INFO [train.py:1198] (1/4) Epoch 20, batch 3750, loss[loss=0.2225, ctc_loss=0.1462, cr_loss=0.3814, over 17008.00 frames. ], tot_loss[loss=0.2151, ctc_loss=0.1435, cr_loss=0.3584, over 3351505.85 frames. ], batch size: 51, lr: 5.93e-03, grad_scale: 32.0
2024-09-23 22:47:03,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=362945.3333333333, ans=0.0
2024-09-23 22:47:17,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=362992.0, ans=0.1
2024-09-23 22:47:24,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=362992.0, ans=0.125
2024-09-23 22:47:27,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=362992.0, ans=0.125
2024-09-23 22:47:30,494 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.286e+02 1.371e+02 1.493e+02 3.473e+02, threshold=2.742e+02, percent-clipped=1.0
2024-09-23 22:47:46,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=363038.6666666667, ans=0.125
2024-09-23 22:47:56,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=363085.3333333333, ans=0.0
2024-09-23 22:47:57,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=363085.3333333333, ans=0.2
2024-09-23 22:47:58,249 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.01 vs. limit=15.0
2024-09-23 22:48:04,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=363132.0, ans=0.0
2024-09-23 22:48:10,926 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.76 vs. limit=15.0
2024-09-23 22:48:13,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=363132.0, ans=0.1
2024-09-23 22:48:22,097 INFO [train.py:1198] (1/4) Epoch 20, batch 3800, loss[loss=0.1865, ctc_loss=0.1217, cr_loss=0.3244, over 17182.00 frames. ], tot_loss[loss=0.2144, ctc_loss=0.1428, cr_loss=0.3581, over 3331432.37 frames. ], batch size: 41, lr: 5.93e-03, grad_scale: 32.0
2024-09-23 22:48:23,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=363178.6666666667, ans=0.2
2024-09-23 22:49:05,535 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=7.69 vs. limit=15.0
2024-09-23 22:49:06,718 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.18 vs. limit=22.5
2024-09-23 22:49:09,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=363318.6666666667, ans=0.1
2024-09-23 22:49:40,057 INFO [train.py:1198] (1/4) Epoch 20, batch 3850, loss[loss=0.2753, ctc_loss=0.1959, cr_loss=0.3969, over 11962.00 frames. ], tot_loss[loss=0.2154, ctc_loss=0.1438, cr_loss=0.358, over 3273538.76 frames. ], batch size: 123, lr: 5.93e-03, grad_scale: 32.0
2024-09-23 22:50:04,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=363458.6666666667, ans=0.0
2024-09-23 22:50:08,337 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.324e+02 1.457e+02 1.626e+02 2.908e+02, threshold=2.914e+02, percent-clipped=1.0
2024-09-23 22:50:08,679 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 22:50:20,860 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 22:50:31,774 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.93 vs. limit=6.0
2024-09-23 22:51:41,131 INFO [train.py:1198] (1/4) Epoch 21, batch 0, loss[loss=0.189, ctc_loss=0.1257, cr_loss=0.3169, over 17090.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1257, cr_loss=0.3169, over 17090.00 frames. ], batch size: 43, lr: 5.78e-03, grad_scale: 32.0
2024-09-23 22:51:41,131 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-23 22:51:57,015 INFO [train.py:1230] (1/4) Epoch 21, validation: loss=0.03907, ctc_loss=0.03907, cr_loss=7.91e-15, over 944034.00 frames. 
2024-09-23 22:51:57,016 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-23 22:52:38,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.whiten.whitening_limit, batch_count=363720.0, ans=12.0
2024-09-23 22:52:54,159 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.45 vs. limit=10.0
2024-09-23 22:52:57,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=363766.6666666667, ans=0.125
2024-09-23 22:53:14,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=363813.3333333333, ans=0.0
2024-09-23 22:53:18,865 INFO [train.py:1198] (1/4) Epoch 21, batch 50, loss[loss=0.2418, ctc_loss=0.1603, cr_loss=0.4074, over 17304.00 frames. ], tot_loss[loss=0.2187, ctc_loss=0.1452, cr_loss=0.3677, over 760610.11 frames. ], batch size: 49, lr: 5.78e-03, grad_scale: 32.0
2024-09-23 22:53:26,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=363860.0, ans=0.1
2024-09-23 22:53:41,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=363906.6666666667, ans=0.125
2024-09-23 22:53:56,544 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.319e+02 1.470e+02 1.661e+02 2.685e+02, threshold=2.941e+02, percent-clipped=0.0
2024-09-23 22:54:41,602 INFO [train.py:1198] (1/4) Epoch 21, batch 100, loss[loss=0.2432, ctc_loss=0.1633, cr_loss=0.3994, over 17052.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1432, cr_loss=0.3624, over 1331221.91 frames. ], batch size: 52, lr: 5.78e-03, grad_scale: 32.0
2024-09-23 22:54:56,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=364140.0, ans=0.125
2024-09-23 22:55:00,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=364140.0, ans=0.07
2024-09-23 22:55:03,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=364140.0, ans=0.1
2024-09-23 22:55:17,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=364186.6666666667, ans=0.05
2024-09-23 22:55:29,487 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.05 vs. limit=6.0
2024-09-23 22:55:32,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.57 vs. limit=15.0
2024-09-23 22:55:48,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=364280.0, ans=0.0
2024-09-23 22:55:50,407 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.10 vs. limit=15.0
2024-09-23 22:56:03,805 INFO [train.py:1198] (1/4) Epoch 21, batch 150, loss[loss=0.2285, ctc_loss=0.1548, cr_loss=0.3687, over 17218.00 frames. ], tot_loss[loss=0.2156, ctc_loss=0.1433, cr_loss=0.3617, over 1781149.96 frames. ], batch size: 47, lr: 5.78e-03, grad_scale: 32.0
2024-09-23 22:56:07,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=364326.6666666667, ans=0.125
2024-09-23 22:56:11,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=364326.6666666667, ans=0.0
2024-09-23 22:56:12,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=364326.6666666667, ans=0.125
2024-09-23 22:56:18,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=364373.3333333333, ans=0.125
2024-09-23 22:56:26,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=364373.3333333333, ans=0.125
2024-09-23 22:56:38,963 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.242e+02 1.342e+02 1.442e+02 2.042e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-23 22:57:04,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=364466.6666666667, ans=0.09899494936611666
2024-09-23 22:57:29,599 INFO [train.py:1198] (1/4) Epoch 21, batch 200, loss[loss=0.2413, ctc_loss=0.1661, cr_loss=0.3761, over 11685.00 frames. ], tot_loss[loss=0.2161, ctc_loss=0.1439, cr_loss=0.3613, over 2115211.45 frames. ], batch size: 124, lr: 5.78e-03, grad_scale: 32.0
2024-09-23 22:57:31,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=364560.0, ans=0.125
2024-09-23 22:57:31,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=364560.0, ans=0.125
2024-09-23 22:57:33,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=364560.0, ans=0.125
2024-09-23 22:57:44,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=364606.6666666667, ans=0.1
2024-09-23 22:57:46,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=364606.6666666667, ans=0.025
2024-09-23 22:58:07,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.34 vs. limit=10.0
2024-09-23 22:58:08,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=364653.3333333333, ans=0.0
2024-09-23 22:58:27,764 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.85 vs. limit=22.5
2024-09-23 22:58:35,331 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-23 22:58:52,293 INFO [train.py:1198] (1/4) Epoch 21, batch 250, loss[loss=0.2053, ctc_loss=0.1323, cr_loss=0.365, over 17038.00 frames. ], tot_loss[loss=0.2142, ctc_loss=0.1424, cr_loss=0.3593, over 2392996.49 frames. ], batch size: 39, lr: 5.77e-03, grad_scale: 32.0
2024-09-23 22:59:13,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=364840.0, ans=0.125
2024-09-23 22:59:18,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=364840.0, ans=0.1
2024-09-23 22:59:19,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=364840.0, ans=0.1
2024-09-23 22:59:23,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=364886.6666666667, ans=0.0
2024-09-23 22:59:27,565 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.024e+02 1.297e+02 1.364e+02 1.577e+02 2.161e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-23 22:59:45,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=364933.3333333333, ans=0.1
2024-09-23 22:59:48,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=364933.3333333333, ans=0.125
2024-09-23 23:00:15,863 INFO [train.py:1198] (1/4) Epoch 21, batch 300, loss[loss=0.2292, ctc_loss=0.1532, cr_loss=0.3802, over 16881.00 frames. ], tot_loss[loss=0.2134, ctc_loss=0.1419, cr_loss=0.3578, over 2599346.44 frames. ], batch size: 58, lr: 5.77e-03, grad_scale: 32.0
2024-09-23 23:00:28,163 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.77 vs. limit=6.0
2024-09-23 23:00:43,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=365073.3333333333, ans=0.0
2024-09-23 23:00:44,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=365073.3333333333, ans=0.125
2024-09-23 23:00:53,664 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.70 vs. limit=10.0
2024-09-23 23:00:56,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=365120.0, ans=0.2
2024-09-23 23:01:11,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=365166.6666666667, ans=0.0
2024-09-23 23:01:27,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=365213.3333333333, ans=0.0
2024-09-23 23:01:35,406 INFO [train.py:1198] (1/4) Epoch 21, batch 350, loss[loss=0.2059, ctc_loss=0.1359, cr_loss=0.35, over 17127.00 frames. ], tot_loss[loss=0.212, ctc_loss=0.1409, cr_loss=0.3557, over 2771890.17 frames. ], batch size: 48, lr: 5.77e-03, grad_scale: 32.0
2024-09-23 23:01:51,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=365260.0, ans=10.0
2024-09-23 23:02:04,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=365306.6666666667, ans=0.1
2024-09-23 23:02:16,762 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.263e+02 1.348e+02 1.461e+02 2.184e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-23 23:02:33,480 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.51 vs. limit=15.0
2024-09-23 23:03:01,498 INFO [train.py:1198] (1/4) Epoch 21, batch 400, loss[loss=0.1773, ctc_loss=0.1171, cr_loss=0.3011, over 17291.00 frames. ], tot_loss[loss=0.213, ctc_loss=0.1416, cr_loss=0.3569, over 2896229.29 frames. ], batch size: 42, lr: 5.77e-03, grad_scale: 32.0
2024-09-23 23:03:06,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=365493.3333333333, ans=0.09899494936611666
2024-09-23 23:03:08,059 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-23 23:03:37,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=365586.6666666667, ans=0.0
2024-09-23 23:04:15,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=365680.0, ans=0.125
2024-09-23 23:04:23,585 INFO [train.py:1198] (1/4) Epoch 21, batch 450, loss[loss=0.2111, ctc_loss=0.1419, cr_loss=0.3459, over 16760.00 frames. ], tot_loss[loss=0.2138, ctc_loss=0.142, cr_loss=0.3589, over 3004066.20 frames. ], batch size: 61, lr: 5.77e-03, grad_scale: 32.0
2024-09-23 23:04:46,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=365773.3333333333, ans=0.0
2024-09-23 23:04:48,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten.whitening_limit, batch_count=365773.3333333333, ans=15.0
2024-09-23 23:04:59,071 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.052e+02 1.246e+02 1.320e+02 1.437e+02 2.140e+02, threshold=2.640e+02, percent-clipped=0.0
2024-09-23 23:04:59,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=365820.0, ans=0.125
2024-09-23 23:05:38,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=365913.3333333333, ans=0.125
2024-09-23 23:05:46,207 INFO [train.py:1198] (1/4) Epoch 21, batch 500, loss[loss=0.2322, ctc_loss=0.1553, cr_loss=0.3845, over 17002.00 frames. ], tot_loss[loss=0.2145, ctc_loss=0.1426, cr_loss=0.3597, over 3085994.41 frames. ], batch size: 53, lr: 5.76e-03, grad_scale: 32.0
2024-09-23 23:06:48,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=366100.0, ans=0.125
2024-09-23 23:06:57,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=366146.6666666667, ans=0.2
2024-09-23 23:06:57,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=366146.6666666667, ans=0.0
2024-09-23 23:07:05,124 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.43 vs. limit=15.0
2024-09-23 23:07:05,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=366146.6666666667, ans=0.0
2024-09-23 23:07:11,259 INFO [train.py:1198] (1/4) Epoch 21, batch 550, loss[loss=0.2006, ctc_loss=0.1324, cr_loss=0.3408, over 17108.00 frames. ], tot_loss[loss=0.215, ctc_loss=0.1429, cr_loss=0.3604, over 3149645.50 frames. ], batch size: 49, lr: 5.76e-03, grad_scale: 32.0
2024-09-23 23:07:37,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=366240.0, ans=0.125
2024-09-23 23:07:46,483 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.227e+02 1.339e+02 1.480e+02 1.923e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-23 23:08:12,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=366333.3333333333, ans=0.125
2024-09-23 23:08:16,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=366380.0, ans=0.0
2024-09-23 23:08:17,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=366380.0, ans=0.0
2024-09-23 23:08:25,053 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.19 vs. limit=12.0
2024-09-23 23:08:32,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=366426.6666666667, ans=0.125
2024-09-23 23:08:33,668 INFO [train.py:1198] (1/4) Epoch 21, batch 600, loss[loss=0.2212, ctc_loss=0.1464, cr_loss=0.374, over 17259.00 frames. ], tot_loss[loss=0.2155, ctc_loss=0.1433, cr_loss=0.3607, over 3186335.87 frames. ], batch size: 44, lr: 5.76e-03, grad_scale: 32.0
2024-09-23 23:09:14,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=366520.0, ans=0.125
2024-09-23 23:09:53,532 INFO [train.py:1198] (1/4) Epoch 21, batch 650, loss[loss=0.2235, ctc_loss=0.152, cr_loss=0.3577, over 16740.00 frames. ], tot_loss[loss=0.215, ctc_loss=0.1431, cr_loss=0.3597, over 3227180.92 frames. ], batch size: 61, lr: 5.76e-03, grad_scale: 16.0
2024-09-23 23:09:54,047 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.16 vs. limit=12.0
2024-09-23 23:10:04,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=366660.0, ans=0.1
2024-09-23 23:10:05,198 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.35 vs. limit=15.0
2024-09-23 23:10:32,740 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.251e+02 1.353e+02 1.422e+02 2.083e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-23 23:11:00,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=366846.6666666667, ans=0.1
2024-09-23 23:11:02,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=366846.6666666667, ans=0.125
2024-09-23 23:11:13,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=366846.6666666667, ans=0.0
2024-09-23 23:11:15,972 INFO [train.py:1198] (1/4) Epoch 21, batch 700, loss[loss=0.2216, ctc_loss=0.1481, cr_loss=0.3673, over 16568.00 frames. ], tot_loss[loss=0.2158, ctc_loss=0.1435, cr_loss=0.3614, over 3260382.33 frames. ], batch size: 66, lr: 5.76e-03, grad_scale: 16.0
2024-09-23 23:11:19,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=366893.3333333333, ans=0.125
2024-09-23 23:11:23,431 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.25 vs. limit=15.0
2024-09-23 23:11:24,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=366893.3333333333, ans=0.125
2024-09-23 23:11:41,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=366940.0, ans=0.1
2024-09-23 23:12:10,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=367033.3333333333, ans=0.1
2024-09-23 23:12:27,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=367080.0, ans=0.125
2024-09-23 23:12:36,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=367080.0, ans=0.125
2024-09-23 23:12:41,459 INFO [train.py:1198] (1/4) Epoch 21, batch 750, loss[loss=0.2239, ctc_loss=0.1499, cr_loss=0.3701, over 17015.00 frames. ], tot_loss[loss=0.2162, ctc_loss=0.1438, cr_loss=0.3621, over 3264194.69 frames. ], batch size: 52, lr: 5.76e-03, grad_scale: 16.0
2024-09-23 23:12:54,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.max_abs, batch_count=367126.6666666667, ans=10.0
2024-09-23 23:12:56,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=367173.3333333333, ans=0.1
2024-09-23 23:13:12,309 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.98 vs. limit=22.5
2024-09-23 23:13:14,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=367220.0, ans=0.1
2024-09-23 23:13:21,100 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.308e+02 1.415e+02 1.587e+02 2.063e+02, threshold=2.831e+02, percent-clipped=0.0
2024-09-23 23:13:52,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=367313.3333333333, ans=0.125
2024-09-23 23:14:04,986 INFO [train.py:1198] (1/4) Epoch 21, batch 800, loss[loss=0.2719, ctc_loss=0.1913, cr_loss=0.4029, over 12409.00 frames. ], tot_loss[loss=0.2161, ctc_loss=0.1438, cr_loss=0.3612, over 3286467.20 frames. ], batch size: 123, lr: 5.75e-03, grad_scale: 32.0
2024-09-23 23:14:07,754 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.52 vs. limit=10.0
2024-09-23 23:14:12,367 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.62 vs. limit=22.5
2024-09-23 23:14:16,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=367360.0, ans=0.125
2024-09-23 23:14:24,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=367406.6666666667, ans=0.0
2024-09-23 23:14:41,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=367453.3333333333, ans=0.0
2024-09-23 23:14:48,809 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.86 vs. limit=6.0
2024-09-23 23:14:59,471 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.16 vs. limit=22.5
2024-09-23 23:15:00,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=367500.0, ans=0.0
2024-09-23 23:15:03,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=367500.0, ans=0.125
2024-09-23 23:15:13,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=367546.6666666667, ans=0.0
2024-09-23 23:15:14,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=367546.6666666667, ans=0.0
2024-09-23 23:15:14,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=367546.6666666667, ans=0.0
2024-09-23 23:15:27,331 INFO [train.py:1198] (1/4) Epoch 21, batch 850, loss[loss=0.231, ctc_loss=0.1533, cr_loss=0.3882, over 17313.00 frames. ], tot_loss[loss=0.2152, ctc_loss=0.1432, cr_loss=0.36, over 3309681.52 frames. ], batch size: 51, lr: 5.75e-03, grad_scale: 32.0
2024-09-23 23:16:05,379 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.235e+02 1.359e+02 1.551e+02 2.196e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-23 23:16:24,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=367733.3333333333, ans=0.025
2024-09-23 23:16:31,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=367733.3333333333, ans=0.125
2024-09-23 23:16:53,128 INFO [train.py:1198] (1/4) Epoch 21, batch 900, loss[loss=0.1989, ctc_loss=0.1326, cr_loss=0.3314, over 17067.00 frames. ], tot_loss[loss=0.215, ctc_loss=0.1431, cr_loss=0.3598, over 3315848.44 frames. ], batch size: 43, lr: 5.75e-03, grad_scale: 16.0
2024-09-23 23:17:04,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=367826.6666666667, ans=0.1
2024-09-23 23:17:15,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=367873.3333333333, ans=0.125
2024-09-23 23:18:06,329 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.28 vs. limit=22.5
2024-09-23 23:18:07,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=368013.3333333333, ans=0.125
2024-09-23 23:18:12,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=368013.3333333333, ans=0.125
2024-09-23 23:18:15,415 INFO [train.py:1198] (1/4) Epoch 21, batch 950, loss[loss=0.2159, ctc_loss=0.1438, cr_loss=0.3606, over 17024.00 frames. ], tot_loss[loss=0.2146, ctc_loss=0.1427, cr_loss=0.3594, over 3327244.50 frames. ], batch size: 51, lr: 5.75e-03, grad_scale: 16.0
2024-09-23 23:18:15,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=368060.0, ans=0.125
2024-09-23 23:18:22,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=368060.0, ans=0.125
2024-09-23 23:18:22,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=368060.0, ans=0.125
2024-09-23 23:18:23,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=368060.0, ans=0.125
2024-09-23 23:18:29,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=368106.6666666667, ans=0.0
2024-09-23 23:18:36,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=368106.6666666667, ans=0.2
2024-09-23 23:18:44,411 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=10.78 vs. limit=15.0
2024-09-23 23:18:52,646 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.89 vs. limit=15.0
2024-09-23 23:18:53,167 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.253e+02 1.356e+02 1.481e+02 2.989e+02, threshold=2.713e+02, percent-clipped=1.0
2024-09-23 23:19:34,878 INFO [train.py:1198] (1/4) Epoch 21, batch 1000, loss[loss=0.2369, ctc_loss=0.1613, cr_loss=0.3779, over 17096.00 frames. ], tot_loss[loss=0.2147, ctc_loss=0.1428, cr_loss=0.3597, over 3339932.80 frames. ], batch size: 49, lr: 5.75e-03, grad_scale: 16.0
2024-09-23 23:20:00,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=368340.0, ans=0.035
2024-09-23 23:20:40,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=368480.0, ans=0.0
2024-09-23 23:20:45,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=368480.0, ans=0.0
2024-09-23 23:20:55,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=368480.0, ans=0.125
2024-09-23 23:20:58,192 INFO [train.py:1198] (1/4) Epoch 21, batch 1050, loss[loss=0.1879, ctc_loss=0.1208, cr_loss=0.3356, over 16935.00 frames. ], tot_loss[loss=0.2149, ctc_loss=0.143, cr_loss=0.3599, over 3340616.55 frames. ], batch size: 42, lr: 5.74e-03, grad_scale: 16.0
2024-09-23 23:20:58,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=368526.6666666667, ans=0.125
2024-09-23 23:21:07,318 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.76 vs. limit=15.0
2024-09-23 23:21:08,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=368526.6666666667, ans=0.0
2024-09-23 23:21:09,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=368526.6666666667, ans=0.0
2024-09-23 23:21:20,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=368573.3333333333, ans=0.0
2024-09-23 23:21:39,785 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.259e+02 1.337e+02 1.472e+02 2.048e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-23 23:21:40,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=368620.0, ans=0.1
2024-09-23 23:22:24,101 INFO [train.py:1198] (1/4) Epoch 21, batch 1100, loss[loss=0.2138, ctc_loss=0.1394, cr_loss=0.3721, over 17012.00 frames. ], tot_loss[loss=0.2147, ctc_loss=0.1428, cr_loss=0.3596, over 3346030.10 frames. ], batch size: 44, lr: 5.74e-03, grad_scale: 16.0
2024-09-23 23:23:27,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=368900.0, ans=0.125
2024-09-23 23:23:43,797 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.70 vs. limit=15.0
2024-09-23 23:23:46,068 INFO [train.py:1198] (1/4) Epoch 21, batch 1150, loss[loss=0.1869, ctc_loss=0.1216, cr_loss=0.3269, over 16949.00 frames. ], tot_loss[loss=0.213, ctc_loss=0.1415, cr_loss=0.3573, over 3347332.01 frames. ], batch size: 42, lr: 5.74e-03, grad_scale: 16.0
2024-09-23 23:23:47,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=368993.3333333333, ans=0.2
2024-09-23 23:23:52,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=368993.3333333333, ans=0.015
2024-09-23 23:23:59,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=368993.3333333333, ans=0.125
2024-09-23 23:24:15,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=369040.0, ans=0.125
2024-09-23 23:24:21,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=369086.6666666667, ans=0.125
2024-09-23 23:24:24,272 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.242e+02 1.332e+02 1.452e+02 2.606e+02, threshold=2.664e+02, percent-clipped=0.0
2024-09-23 23:24:45,585 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=5.72 vs. limit=15.0
2024-09-23 23:25:08,569 INFO [train.py:1198] (1/4) Epoch 21, batch 1200, loss[loss=0.1761, ctc_loss=0.114, cr_loss=0.3108, over 16968.00 frames. ], tot_loss[loss=0.2133, ctc_loss=0.1418, cr_loss=0.3574, over 3354108.63 frames. ], batch size: 42, lr: 5.74e-03, grad_scale: 32.0
2024-09-23 23:25:21,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=369226.6666666667, ans=0.125
2024-09-23 23:25:23,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=369273.3333333333, ans=0.2
2024-09-23 23:26:05,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=369366.6666666667, ans=0.125
2024-09-23 23:26:21,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.07 vs. limit=15.0
2024-09-23 23:26:27,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=369413.3333333333, ans=0.1
2024-09-23 23:26:30,657 INFO [train.py:1198] (1/4) Epoch 21, batch 1250, loss[loss=0.2044, ctc_loss=0.1374, cr_loss=0.3348, over 17241.00 frames. ], tot_loss[loss=0.2138, ctc_loss=0.1422, cr_loss=0.358, over 3355299.87 frames. ], batch size: 50, lr: 5.74e-03, grad_scale: 16.0
2024-09-23 23:26:37,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.34 vs. limit=15.0
2024-09-23 23:26:59,973 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.97 vs. limit=6.0
2024-09-23 23:27:01,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=369506.6666666667, ans=0.125
2024-09-23 23:27:13,542 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.996e+01 1.239e+02 1.348e+02 1.443e+02 2.209e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-23 23:27:15,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=369553.3333333333, ans=0.0
2024-09-23 23:27:18,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=369553.3333333333, ans=0.125
2024-09-23 23:27:19,297 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.40 vs. limit=15.0
2024-09-23 23:27:31,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=369600.0, ans=0.125
2024-09-23 23:27:31,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=369600.0, ans=22.5
2024-09-23 23:27:56,568 INFO [train.py:1198] (1/4) Epoch 21, batch 1300, loss[loss=0.2191, ctc_loss=0.1447, cr_loss=0.3723, over 16996.00 frames. ], tot_loss[loss=0.2141, ctc_loss=0.1424, cr_loss=0.3585, over 3352322.13 frames. ], batch size: 56, lr: 5.74e-03, grad_scale: 16.0
2024-09-23 23:28:17,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=369740.0, ans=0.2
2024-09-23 23:28:17,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=369740.0, ans=0.125
2024-09-23 23:28:21,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=369740.0, ans=0.125
2024-09-23 23:29:16,749 INFO [train.py:1198] (1/4) Epoch 21, batch 1350, loss[loss=0.233, ctc_loss=0.1577, cr_loss=0.3767, over 17004.00 frames. ], tot_loss[loss=0.2139, ctc_loss=0.1423, cr_loss=0.3581, over 3353416.82 frames. ], batch size: 53, lr: 5.73e-03, grad_scale: 16.0
2024-09-23 23:29:23,699 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.55 vs. limit=15.0
2024-09-23 23:29:58,705 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.263e+02 1.337e+02 1.513e+02 2.330e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-23 23:30:38,707 INFO [train.py:1198] (1/4) Epoch 21, batch 1400, loss[loss=0.2347, ctc_loss=0.1579, cr_loss=0.3844, over 17068.00 frames. ], tot_loss[loss=0.2131, ctc_loss=0.1417, cr_loss=0.357, over 3349387.97 frames. ], batch size: 46, lr: 5.73e-03, grad_scale: 16.0
2024-09-23 23:30:47,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=370160.0, ans=0.125
2024-09-23 23:31:16,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=370253.3333333333, ans=0.0
2024-09-23 23:31:31,036 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.55 vs. limit=15.0
2024-09-23 23:32:03,657 INFO [train.py:1198] (1/4) Epoch 21, batch 1450, loss[loss=0.2423, ctc_loss=0.1704, cr_loss=0.3592, over 11917.00 frames. ], tot_loss[loss=0.2127, ctc_loss=0.1414, cr_loss=0.3563, over 3352171.06 frames. ], batch size: 123, lr: 5.73e-03, grad_scale: 16.0
2024-09-23 23:32:06,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.97 vs. limit=15.0
2024-09-23 23:32:17,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=370393.3333333333, ans=0.0
2024-09-23 23:32:27,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=370440.0, ans=0.125
2024-09-23 23:32:46,581 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.255e+02 1.316e+02 1.424e+02 2.117e+02, threshold=2.631e+02, percent-clipped=0.0
2024-09-23 23:32:48,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=370486.6666666667, ans=0.07
2024-09-23 23:32:51,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=370486.6666666667, ans=0.0
2024-09-23 23:33:10,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=370580.0, ans=0.125
2024-09-23 23:33:26,655 INFO [train.py:1198] (1/4) Epoch 21, batch 1500, loss[loss=0.198, ctc_loss=0.1292, cr_loss=0.3439, over 17274.00 frames. ], tot_loss[loss=0.2133, ctc_loss=0.1418, cr_loss=0.3573, over 3345856.39 frames. ], batch size: 42, lr: 5.73e-03, grad_scale: 16.0
2024-09-23 23:33:36,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=370626.6666666667, ans=0.1
2024-09-23 23:34:08,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=370720.0, ans=0.2
2024-09-23 23:34:20,530 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.75 vs. limit=10.0
2024-09-23 23:34:29,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=370813.3333333333, ans=0.09899494936611666
2024-09-23 23:34:29,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=370813.3333333333, ans=10.0
2024-09-23 23:34:29,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=370813.3333333333, ans=0.125
2024-09-23 23:34:41,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=370813.3333333333, ans=0.0
2024-09-23 23:34:49,370 INFO [train.py:1198] (1/4) Epoch 21, batch 1550, loss[loss=0.1981, ctc_loss=0.1294, cr_loss=0.3434, over 17268.00 frames. ], tot_loss[loss=0.2134, ctc_loss=0.1419, cr_loss=0.3574, over 3345238.78 frames. ], batch size: 44, lr: 5.73e-03, grad_scale: 16.0
2024-09-23 23:35:06,338 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.32 vs. limit=15.0
2024-09-23 23:35:15,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=370906.6666666667, ans=0.0
2024-09-23 23:35:21,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=370953.3333333333, ans=0.0
2024-09-23 23:35:29,477 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.240e+02 1.330e+02 1.451e+02 2.029e+02, threshold=2.659e+02, percent-clipped=0.0
2024-09-23 23:35:34,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=370953.3333333333, ans=0.125
2024-09-23 23:35:57,557 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.16 vs. limit=22.5
2024-09-23 23:36:01,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=371046.6666666667, ans=0.0
2024-09-23 23:36:12,039 INFO [train.py:1198] (1/4) Epoch 21, batch 1600, loss[loss=0.2159, ctc_loss=0.1439, cr_loss=0.3601, over 17360.00 frames. ], tot_loss[loss=0.2135, ctc_loss=0.1421, cr_loss=0.3571, over 3346140.07 frames. ], batch size: 48, lr: 5.73e-03, grad_scale: 32.0
2024-09-23 23:36:53,987 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=6.98 vs. limit=15.0
2024-09-23 23:37:37,163 INFO [train.py:1198] (1/4) Epoch 21, batch 1650, loss[loss=0.2597, ctc_loss=0.1778, cr_loss=0.4096, over 11711.00 frames. ], tot_loss[loss=0.2142, ctc_loss=0.1425, cr_loss=0.3587, over 3344686.28 frames. ], batch size: 123, lr: 5.72e-03, grad_scale: 32.0
2024-09-23 23:37:44,480 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.93 vs. limit=10.0
2024-09-23 23:37:59,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=371373.3333333333, ans=0.0
2024-09-23 23:38:16,964 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.286e+02 1.382e+02 1.548e+02 2.796e+02, threshold=2.764e+02, percent-clipped=1.0
2024-09-23 23:38:33,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=371466.6666666667, ans=0.2
2024-09-23 23:38:56,658 INFO [train.py:1198] (1/4) Epoch 21, batch 1700, loss[loss=0.2222, ctc_loss=0.1472, cr_loss=0.375, over 17229.00 frames. ], tot_loss[loss=0.2148, ctc_loss=0.1428, cr_loss=0.3597, over 3351436.92 frames. ], batch size: 50, lr: 5.72e-03, grad_scale: 32.0
2024-09-23 23:38:58,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=371560.0, ans=0.0
2024-09-23 23:39:33,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=371653.3333333333, ans=22.5
2024-09-23 23:39:37,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.40 vs. limit=22.5
2024-09-23 23:39:58,530 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.86 vs. limit=10.0
2024-09-23 23:40:02,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=371746.6666666667, ans=0.2
2024-09-23 23:40:10,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=371746.6666666667, ans=0.2
2024-09-23 23:40:18,411 INFO [train.py:1198] (1/4) Epoch 21, batch 1750, loss[loss=0.1984, ctc_loss=0.1272, cr_loss=0.3564, over 17081.00 frames. ], tot_loss[loss=0.2145, ctc_loss=0.1426, cr_loss=0.3594, over 3356369.17 frames. ], batch size: 43, lr: 5.72e-03, grad_scale: 32.0
2024-09-23 23:40:31,970 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.72 vs. limit=15.0
2024-09-23 23:40:42,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=371840.0, ans=0.125
2024-09-23 23:40:58,126 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.218e+02 1.302e+02 1.379e+02 1.745e+02, threshold=2.603e+02, percent-clipped=0.0
2024-09-23 23:41:03,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=371886.6666666667, ans=0.2
2024-09-23 23:41:28,859 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.75 vs. limit=15.0
2024-09-23 23:41:29,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=371980.0, ans=0.0
2024-09-23 23:41:37,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=371980.0, ans=0.125
2024-09-23 23:41:40,628 INFO [train.py:1198] (1/4) Epoch 21, batch 1800, loss[loss=0.2205, ctc_loss=0.1462, cr_loss=0.3713, over 17270.00 frames. ], tot_loss[loss=0.2146, ctc_loss=0.1428, cr_loss=0.3591, over 3361185.35 frames. ], batch size: 44, lr: 5.72e-03, grad_scale: 32.0
2024-09-23 23:42:03,147 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.54 vs. limit=15.0
2024-09-23 23:42:13,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=372120.0, ans=10.0
2024-09-23 23:42:44,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=372166.6666666667, ans=0.1
2024-09-23 23:42:51,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=372213.3333333333, ans=0.05
2024-09-23 23:43:05,767 INFO [train.py:1198] (1/4) Epoch 21, batch 1850, loss[loss=0.2484, ctc_loss=0.1676, cr_loss=0.4042, over 17009.00 frames. ], tot_loss[loss=0.2142, ctc_loss=0.1425, cr_loss=0.3588, over 3364862.34 frames. ], batch size: 56, lr: 5.72e-03, grad_scale: 32.0
2024-09-23 23:43:34,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=372306.6666666667, ans=0.0
2024-09-23 23:43:42,474 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=372353.3333333333, ans=0.0
2024-09-23 23:43:45,461 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.299e+02 1.413e+02 1.617e+02 3.494e+02, threshold=2.826e+02, percent-clipped=2.0
2024-09-23 23:43:49,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=372353.3333333333, ans=0.07
2024-09-23 23:43:54,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=372400.0, ans=0.2
2024-09-23 23:44:09,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=372446.6666666667, ans=0.1
2024-09-23 23:44:11,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=372446.6666666667, ans=0.025
2024-09-23 23:44:18,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=11.19 vs. limit=15.0
2024-09-23 23:44:28,064 INFO [train.py:1198] (1/4) Epoch 21, batch 1900, loss[loss=0.2138, ctc_loss=0.1426, cr_loss=0.356, over 16137.00 frames. ], tot_loss[loss=0.2153, ctc_loss=0.1434, cr_loss=0.3596, over 3345895.65 frames. ], batch size: 74, lr: 5.71e-03, grad_scale: 32.0
2024-09-23 23:44:39,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=372493.3333333333, ans=0.035
2024-09-23 23:44:53,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=372540.0, ans=0.2
2024-09-23 23:45:12,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=372586.6666666667, ans=0.125
2024-09-23 23:45:22,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=372633.3333333333, ans=0.125
2024-09-23 23:45:22,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=372633.3333333333, ans=0.125
2024-09-23 23:45:36,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=372680.0, ans=0.2
2024-09-23 23:45:38,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=372680.0, ans=0.0
2024-09-23 23:45:43,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=372680.0, ans=0.2
2024-09-23 23:45:47,537 INFO [train.py:1198] (1/4) Epoch 21, batch 1950, loss[loss=0.1905, ctc_loss=0.1246, cr_loss=0.3295, over 17181.00 frames. ], tot_loss[loss=0.215, ctc_loss=0.1431, cr_loss=0.3596, over 3348070.43 frames. ], batch size: 41, lr: 5.71e-03, grad_scale: 32.0
2024-09-23 23:46:12,686 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=372773.3333333333, ans=0.015
2024-09-23 23:46:12,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=372773.3333333333, ans=0.2
2024-09-23 23:46:17,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=372773.3333333333, ans=0.0
2024-09-23 23:46:28,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=372820.0, ans=0.2
2024-09-23 23:46:29,779 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.272e+02 1.378e+02 1.476e+02 2.541e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-23 23:46:30,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.whiten.whitening_limit, batch_count=372820.0, ans=12.0
2024-09-23 23:46:51,057 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.97 vs. limit=15.0
2024-09-23 23:47:04,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=372913.3333333333, ans=0.0
2024-09-23 23:47:12,283 INFO [train.py:1198] (1/4) Epoch 21, batch 2000, loss[loss=0.1694, ctc_loss=0.1124, cr_loss=0.285, over 17231.00 frames. ], tot_loss[loss=0.2153, ctc_loss=0.1434, cr_loss=0.3597, over 3347187.93 frames. ], batch size: 41, lr: 5.71e-03, grad_scale: 32.0
2024-09-23 23:47:31,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=373006.6666666667, ans=0.0
2024-09-23 23:47:34,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=373006.6666666667, ans=0.1
2024-09-23 23:47:36,492 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.09 vs. limit=15.0
2024-09-23 23:48:09,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=373100.0, ans=0.125
2024-09-23 23:48:09,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.52 vs. limit=15.0
2024-09-23 23:48:17,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=373146.6666666667, ans=0.125
2024-09-23 23:48:19,611 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.95 vs. limit=15.0
2024-09-23 23:48:34,484 INFO [train.py:1198] (1/4) Epoch 21, batch 2050, loss[loss=0.2121, ctc_loss=0.14, cr_loss=0.3603, over 16466.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1436, cr_loss=0.3608, over 3347309.36 frames. ], batch size: 66, lr: 5.71e-03, grad_scale: 32.0
2024-09-23 23:48:38,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=373193.3333333333, ans=0.0
2024-09-23 23:49:00,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=373240.0, ans=0.2
2024-09-23 23:49:14,304 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.256e+02 1.387e+02 1.531e+02 2.398e+02, threshold=2.775e+02, percent-clipped=0.0
2024-09-23 23:49:59,551 INFO [train.py:1198] (1/4) Epoch 21, batch 2100, loss[loss=0.2533, ctc_loss=0.1707, cr_loss=0.4133, over 14960.00 frames. ], tot_loss[loss=0.2154, ctc_loss=0.1433, cr_loss=0.3605, over 3353808.12 frames. ], batch size: 89, lr: 5.71e-03, grad_scale: 32.0
2024-09-23 23:50:04,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=373426.6666666667, ans=0.125
2024-09-23 23:50:25,884 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.82 vs. limit=15.0
2024-09-23 23:50:38,593 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.81 vs. limit=15.0
2024-09-23 23:51:22,279 INFO [train.py:1198] (1/4) Epoch 21, batch 2150, loss[loss=0.2109, ctc_loss=0.1392, cr_loss=0.3586, over 17303.00 frames. ], tot_loss[loss=0.2147, ctc_loss=0.1428, cr_loss=0.3595, over 3358383.23 frames. ], batch size: 46, lr: 5.71e-03, grad_scale: 32.0
2024-09-23 23:51:22,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=373660.0, ans=0.0
2024-09-23 23:51:40,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=373706.6666666667, ans=0.125
2024-09-23 23:51:57,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=373753.3333333333, ans=0.1
2024-09-23 23:52:04,668 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.257e+02 1.382e+02 1.555e+02 2.014e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-23 23:52:25,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=373800.0, ans=0.05
2024-09-23 23:52:33,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=373846.6666666667, ans=0.2
2024-09-23 23:52:47,116 INFO [train.py:1198] (1/4) Epoch 21, batch 2200, loss[loss=0.2237, ctc_loss=0.1507, cr_loss=0.3648, over 17022.00 frames. ], tot_loss[loss=0.214, ctc_loss=0.1422, cr_loss=0.3586, over 3358939.66 frames. ], batch size: 51, lr: 5.70e-03, grad_scale: 32.0
2024-09-23 23:52:53,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=373893.3333333333, ans=0.125
2024-09-23 23:53:11,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=373940.0, ans=0.125
2024-09-23 23:53:32,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=373986.6666666667, ans=0.07
2024-09-23 23:53:36,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_positive, batch_count=374033.3333333333, ans=0.05
2024-09-23 23:53:43,590 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.59 vs. limit=12.0
2024-09-23 23:53:54,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=374080.0, ans=0.125
2024-09-23 23:54:06,785 INFO [train.py:1198] (1/4) Epoch 21, batch 2250, loss[loss=0.224, ctc_loss=0.1503, cr_loss=0.3683, over 17042.00 frames. ], tot_loss[loss=0.2129, ctc_loss=0.1414, cr_loss=0.3576, over 3362493.76 frames. ], batch size: 52, lr: 5.70e-03, grad_scale: 32.0
2024-09-23 23:54:30,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=374173.3333333333, ans=0.0
2024-09-23 23:54:36,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer_ff3.min_abs, batch_count=374173.3333333333, ans=0.2
2024-09-23 23:54:49,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.304e+02 1.416e+02 1.549e+02 2.189e+02, threshold=2.832e+02, percent-clipped=0.0
2024-09-23 23:54:56,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=6.20 vs. limit=15.0
2024-09-23 23:55:03,932 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-23 23:55:29,455 INFO [train.py:1198] (1/4) Epoch 21, batch 2300, loss[loss=0.1731, ctc_loss=0.1087, cr_loss=0.3217, over 17258.00 frames. ], tot_loss[loss=0.2134, ctc_loss=0.1416, cr_loss=0.3589, over 3357293.39 frames. ], batch size: 42, lr: 5.70e-03, grad_scale: 32.0
2024-09-23 23:55:32,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=374360.0, ans=0.025
2024-09-23 23:55:45,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=374406.6666666667, ans=0.2
2024-09-23 23:55:50,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=374406.6666666667, ans=0.1
2024-09-23 23:55:52,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=374406.6666666667, ans=0.0
2024-09-23 23:56:10,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=374453.3333333333, ans=0.125
2024-09-23 23:56:54,560 INFO [train.py:1198] (1/4) Epoch 21, batch 2350, loss[loss=0.2015, ctc_loss=0.1311, cr_loss=0.3523, over 17169.00 frames. ], tot_loss[loss=0.2145, ctc_loss=0.1425, cr_loss=0.3601, over 3357608.04 frames. ], batch size: 45, lr: 5.70e-03, grad_scale: 32.0
2024-09-23 23:57:38,981 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.011e+02 1.274e+02 1.389e+02 1.537e+02 2.355e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-23 23:57:47,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=374733.3333333333, ans=0.125
2024-09-23 23:58:17,352 INFO [train.py:1198] (1/4) Epoch 21, batch 2400, loss[loss=0.1759, ctc_loss=0.1157, cr_loss=0.3009, over 17182.00 frames. ], tot_loss[loss=0.2142, ctc_loss=0.1424, cr_loss=0.3594, over 3344505.78 frames. ], batch size: 41, lr: 5.70e-03, grad_scale: 32.0
2024-09-23 23:59:03,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=374966.6666666667, ans=0.0
2024-09-23 23:59:22,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=375013.3333333333, ans=0.0
2024-09-23 23:59:39,561 INFO [train.py:1198] (1/4) Epoch 21, batch 2450, loss[loss=0.2096, ctc_loss=0.1378, cr_loss=0.359, over 16900.00 frames. ], tot_loss[loss=0.2137, ctc_loss=0.142, cr_loss=0.3584, over 3352446.45 frames. ], batch size: 58, lr: 5.70e-03, grad_scale: 32.0
2024-09-24 00:00:01,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.28 vs. limit=6.0
2024-09-24 00:00:19,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=375153.3333333333, ans=0.125
2024-09-24 00:00:20,975 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.266e+02 1.364e+02 1.496e+02 2.065e+02, threshold=2.727e+02, percent-clipped=0.0
2024-09-24 00:00:22,094 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.85 vs. limit=10.0
2024-09-24 00:00:27,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=375200.0, ans=0.125
2024-09-24 00:01:02,063 INFO [train.py:1198] (1/4) Epoch 21, batch 2500, loss[loss=0.265, ctc_loss=0.1799, cr_loss=0.4255, over 15012.00 frames. ], tot_loss[loss=0.2146, ctc_loss=0.1427, cr_loss=0.3596, over 3349996.07 frames. ], batch size: 89, lr: 5.69e-03, grad_scale: 32.0
2024-09-24 00:01:16,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=375340.0, ans=0.2
2024-09-24 00:01:21,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=375340.0, ans=0.125
2024-09-24 00:01:52,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=375433.3333333333, ans=0.1
2024-09-24 00:01:58,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=375433.3333333333, ans=0.125
2024-09-24 00:02:26,981 INFO [train.py:1198] (1/4) Epoch 21, batch 2550, loss[loss=0.2089, ctc_loss=0.1379, cr_loss=0.3551, over 17250.00 frames. ], tot_loss[loss=0.2152, ctc_loss=0.1431, cr_loss=0.3606, over 3349649.84 frames. ], batch size: 42, lr: 5.69e-03, grad_scale: 32.0
2024-09-24 00:02:51,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=375573.3333333333, ans=0.125
2024-09-24 00:03:08,316 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.019e+02 1.246e+02 1.350e+02 1.498e+02 2.836e+02, threshold=2.701e+02, percent-clipped=1.0
2024-09-24 00:03:46,558 INFO [train.py:1198] (1/4) Epoch 21, batch 2600, loss[loss=0.2051, ctc_loss=0.1352, cr_loss=0.3498, over 17108.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1434, cr_loss=0.3619, over 3354333.91 frames. ], batch size: 49, lr: 5.69e-03, grad_scale: 32.0
2024-09-24 00:03:54,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=375760.0, ans=0.125
2024-09-24 00:04:49,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=375900.0, ans=0.1
2024-09-24 00:05:02,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=375946.6666666667, ans=0.125
2024-09-24 00:05:08,530 INFO [train.py:1198] (1/4) Epoch 21, batch 2650, loss[loss=0.2399, ctc_loss=0.1621, cr_loss=0.3887, over 16505.00 frames. ], tot_loss[loss=0.2157, ctc_loss=0.1433, cr_loss=0.3616, over 3356787.53 frames. ], batch size: 66, lr: 5.69e-03, grad_scale: 32.0
2024-09-24 00:05:15,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=375993.3333333333, ans=0.2
2024-09-24 00:05:15,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=375993.3333333333, ans=0.025
2024-09-24 00:05:26,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=376040.0, ans=0.125
2024-09-24 00:05:51,664 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.12 vs. limit=6.0
2024-09-24 00:05:52,572 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.243e+02 1.351e+02 1.464e+02 2.576e+02, threshold=2.701e+02, percent-clipped=0.0
2024-09-24 00:05:52,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=376086.6666666667, ans=0.125
2024-09-24 00:06:04,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=376133.3333333333, ans=0.0
2024-09-24 00:06:28,136 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 00:06:32,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=376226.6666666667, ans=0.125
2024-09-24 00:06:33,544 INFO [train.py:1198] (1/4) Epoch 21, batch 2700, loss[loss=0.2338, ctc_loss=0.1565, cr_loss=0.3864, over 16572.00 frames. ], tot_loss[loss=0.2156, ctc_loss=0.1433, cr_loss=0.361, over 3357832.85 frames. ], batch size: 66, lr: 5.69e-03, grad_scale: 32.0
2024-09-24 00:06:48,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=376273.3333333333, ans=0.125
2024-09-24 00:07:22,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=376366.6666666667, ans=0.125
2024-09-24 00:07:55,723 INFO [train.py:1198] (1/4) Epoch 21, batch 2750, loss[loss=0.2133, ctc_loss=0.1407, cr_loss=0.3631, over 17201.00 frames. ], tot_loss[loss=0.215, ctc_loss=0.143, cr_loss=0.3604, over 3352068.71 frames. ], batch size: 47, lr: 5.68e-03, grad_scale: 32.0
2024-09-24 00:08:22,627 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=8.97 vs. limit=22.5
2024-09-24 00:08:29,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=376553.3333333333, ans=0.0
2024-09-24 00:08:37,708 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.226e+02 1.342e+02 1.468e+02 2.196e+02, threshold=2.683e+02, percent-clipped=0.0
2024-09-24 00:08:51,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.51 vs. limit=15.0
2024-09-24 00:09:02,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=376646.6666666667, ans=0.0
2024-09-24 00:09:10,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=376646.6666666667, ans=0.125
2024-09-24 00:09:18,226 INFO [train.py:1198] (1/4) Epoch 21, batch 2800, loss[loss=0.2021, ctc_loss=0.1351, cr_loss=0.3347, over 17307.00 frames. ], tot_loss[loss=0.2154, ctc_loss=0.1433, cr_loss=0.3605, over 3352794.32 frames. ], batch size: 49, lr: 5.68e-03, grad_scale: 32.0
2024-09-24 00:09:34,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=376740.0, ans=0.125
2024-09-24 00:09:42,686 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.63 vs. limit=15.0
2024-09-24 00:10:25,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=376880.0, ans=0.025
2024-09-24 00:10:37,015 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 00:10:38,125 INFO [train.py:1198] (1/4) Epoch 21, batch 2850, loss[loss=0.1909, ctc_loss=0.1231, cr_loss=0.3388, over 17259.00 frames. ], tot_loss[loss=0.2147, ctc_loss=0.1429, cr_loss=0.3592, over 3350597.91 frames. ], batch size: 44, lr: 5.68e-03, grad_scale: 32.0
2024-09-24 00:10:47,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=376926.6666666667, ans=0.025
2024-09-24 00:11:00,828 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.72 vs. limit=15.0
2024-09-24 00:11:21,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=377020.0, ans=0.025
2024-09-24 00:11:22,310 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.291e+02 1.370e+02 1.483e+02 1.856e+02, threshold=2.740e+02, percent-clipped=0.0
2024-09-24 00:11:47,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=377113.3333333333, ans=0.125
2024-09-24 00:11:58,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=377113.3333333333, ans=0.1
2024-09-24 00:12:03,410 INFO [train.py:1198] (1/4) Epoch 21, batch 2900, loss[loss=0.2122, ctc_loss=0.1421, cr_loss=0.3506, over 17341.00 frames. ], tot_loss[loss=0.2132, ctc_loss=0.1417, cr_loss=0.3573, over 3354819.24 frames. ], batch size: 48, lr: 5.68e-03, grad_scale: 32.0
2024-09-24 00:12:29,009 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 00:12:38,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=377253.3333333333, ans=0.1
2024-09-24 00:13:26,186 INFO [train.py:1198] (1/4) Epoch 21, batch 2950, loss[loss=0.2154, ctc_loss=0.1424, cr_loss=0.3649, over 16783.00 frames. ], tot_loss[loss=0.2126, ctc_loss=0.1413, cr_loss=0.3565, over 3360673.71 frames. ], batch size: 61, lr: 5.68e-03, grad_scale: 16.0
2024-09-24 00:13:29,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=377393.3333333333, ans=0.5
2024-09-24 00:14:11,336 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.234e+02 1.337e+02 1.460e+02 2.034e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-24 00:14:24,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=377533.3333333333, ans=0.09899494936611666
2024-09-24 00:14:27,690 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.29 vs. limit=12.0
2024-09-24 00:14:38,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=377580.0, ans=0.125
2024-09-24 00:14:47,420 INFO [train.py:1198] (1/4) Epoch 21, batch 3000, loss[loss=0.1944, ctc_loss=0.1252, cr_loss=0.3464, over 17261.00 frames. ], tot_loss[loss=0.2129, ctc_loss=0.1414, cr_loss=0.3571, over 3357540.18 frames. ], batch size: 42, lr: 5.68e-03, grad_scale: 16.0
2024-09-24 00:14:47,421 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 00:15:02,885 INFO [train.py:1230] (1/4) Epoch 21, validation: loss=0.03893, ctc_loss=0.03893, cr_loss=7.803e-15, over 944034.00 frames. 
2024-09-24 00:15:02,886 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 00:15:33,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=377720.0, ans=0.125
2024-09-24 00:15:35,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=377720.0, ans=0.025
2024-09-24 00:15:42,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=377720.0, ans=0.125
2024-09-24 00:16:11,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=377813.3333333333, ans=0.09899494936611666
2024-09-24 00:16:21,772 INFO [train.py:1198] (1/4) Epoch 21, batch 3050, loss[loss=0.1908, ctc_loss=0.1225, cr_loss=0.3412, over 17061.00 frames. ], tot_loss[loss=0.2132, ctc_loss=0.1418, cr_loss=0.3573, over 3354526.90 frames. ], batch size: 46, lr: 5.67e-03, grad_scale: 16.0
2024-09-24 00:16:26,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=377860.0, ans=0.0
2024-09-24 00:17:00,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=377953.3333333333, ans=0.125
2024-09-24 00:17:04,472 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.246e+02 1.316e+02 1.406e+02 3.085e+02, threshold=2.632e+02, percent-clipped=1.0
2024-09-24 00:17:40,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=378046.6666666667, ans=0.125
2024-09-24 00:17:43,200 INFO [train.py:1198] (1/4) Epoch 21, batch 3100, loss[loss=0.2051, ctc_loss=0.1359, cr_loss=0.3459, over 17221.00 frames. ], tot_loss[loss=0.2136, ctc_loss=0.142, cr_loss=0.358, over 3362730.62 frames. ], batch size: 50, lr: 5.67e-03, grad_scale: 16.0
2024-09-24 00:17:52,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=378093.3333333333, ans=0.0
2024-09-24 00:18:29,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=378233.3333333333, ans=0.125
2024-09-24 00:19:04,079 INFO [train.py:1198] (1/4) Epoch 21, batch 3150, loss[loss=0.1967, ctc_loss=0.1301, cr_loss=0.3333, over 17280.00 frames. ], tot_loss[loss=0.2139, ctc_loss=0.1422, cr_loss=0.3585, over 3360146.83 frames. ], batch size: 42, lr: 5.67e-03, grad_scale: 16.0
2024-09-24 00:19:13,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=378326.6666666667, ans=0.025
2024-09-24 00:19:29,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=378373.3333333333, ans=15.0
2024-09-24 00:19:35,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=378420.0, ans=0.2
2024-09-24 00:19:46,189 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.987e+01 1.255e+02 1.337e+02 1.461e+02 2.046e+02, threshold=2.675e+02, percent-clipped=0.0
2024-09-24 00:19:55,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=378466.6666666667, ans=0.2
2024-09-24 00:19:58,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=378466.6666666667, ans=0.125
2024-09-24 00:20:12,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=378513.3333333333, ans=0.0
2024-09-24 00:20:16,625 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.17 vs. limit=22.5
2024-09-24 00:20:21,949 INFO [train.py:1198] (1/4) Epoch 21, batch 3200, loss[loss=0.2039, ctc_loss=0.1378, cr_loss=0.3305, over 17226.00 frames. ], tot_loss[loss=0.214, ctc_loss=0.1422, cr_loss=0.3586, over 3360919.99 frames. ], batch size: 50, lr: 5.67e-03, grad_scale: 32.0
2024-09-24 00:20:42,502 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.12 vs. limit=10.0
2024-09-24 00:20:44,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=378606.6666666667, ans=0.09899494936611666
2024-09-24 00:21:11,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=378700.0, ans=0.2
2024-09-24 00:21:14,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=378700.0, ans=10.0
2024-09-24 00:21:25,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=378746.6666666667, ans=0.0
2024-09-24 00:21:42,247 INFO [train.py:1198] (1/4) Epoch 21, batch 3250, loss[loss=0.1857, ctc_loss=0.1191, cr_loss=0.3332, over 17179.00 frames. ], tot_loss[loss=0.2132, ctc_loss=0.1416, cr_loss=0.3579, over 3366340.72 frames. ], batch size: 41, lr: 5.67e-03, grad_scale: 32.0
2024-09-24 00:21:52,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.whiten.whitening_limit, batch_count=378793.3333333333, ans=12.0
2024-09-24 00:22:10,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=378840.0, ans=0.1
2024-09-24 00:22:15,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=378886.6666666667, ans=0.2
2024-09-24 00:22:24,295 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.280e+02 1.348e+02 1.474e+02 1.911e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-24 00:22:50,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=378980.0, ans=0.1
2024-09-24 00:22:51,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=378980.0, ans=0.125
2024-09-24 00:22:52,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=378980.0, ans=0.0
2024-09-24 00:23:00,167 INFO [train.py:1198] (1/4) Epoch 21, batch 3300, loss[loss=0.1923, ctc_loss=0.1234, cr_loss=0.3445, over 16362.00 frames. ], tot_loss[loss=0.2126, ctc_loss=0.1411, cr_loss=0.3573, over 3363257.83 frames. ], batch size: 36, lr: 5.67e-03, grad_scale: 32.0
2024-09-24 00:23:00,692 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.91 vs. limit=15.0
2024-09-24 00:23:14,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=379073.3333333333, ans=0.125
2024-09-24 00:23:48,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=379166.6666666667, ans=0.125
2024-09-24 00:23:53,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=379166.6666666667, ans=0.125
2024-09-24 00:23:58,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=379166.6666666667, ans=0.1
2024-09-24 00:24:18,313 INFO [train.py:1198] (1/4) Epoch 21, batch 3350, loss[loss=0.2309, ctc_loss=0.1497, cr_loss=0.4059, over 16446.00 frames. ], tot_loss[loss=0.2136, ctc_loss=0.142, cr_loss=0.3581, over 3357850.01 frames. ], batch size: 66, lr: 5.66e-03, grad_scale: 32.0
2024-09-24 00:24:26,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=379260.0, ans=0.125
2024-09-24 00:24:40,251 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.63 vs. limit=15.0
2024-09-24 00:24:58,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=379353.3333333333, ans=0.125
2024-09-24 00:25:01,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=379353.3333333333, ans=0.125
2024-09-24 00:25:02,489 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.292e+02 1.373e+02 1.509e+02 2.435e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-24 00:25:08,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=379400.0, ans=0.2
2024-09-24 00:25:09,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=379400.0, ans=0.1
2024-09-24 00:25:21,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=379446.6666666667, ans=0.125
2024-09-24 00:25:38,757 INFO [train.py:1198] (1/4) Epoch 21, batch 3400, loss[loss=0.2433, ctc_loss=0.164, cr_loss=0.3961, over 15979.00 frames. ], tot_loss[loss=0.2139, ctc_loss=0.1423, cr_loss=0.3581, over 3343707.61 frames. ], batch size: 74, lr: 5.66e-03, grad_scale: 32.0
2024-09-24 00:25:52,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=379540.0, ans=0.05
2024-09-24 00:26:13,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten.whitening_limit, batch_count=379586.6666666667, ans=15.0
2024-09-24 00:26:18,304 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.95 vs. limit=22.5
2024-09-24 00:26:28,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=379633.3333333333, ans=0.125
2024-09-24 00:26:28,848 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 00:26:36,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=379633.3333333333, ans=0.2
2024-09-24 00:26:43,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=379680.0, ans=0.125
2024-09-24 00:26:56,713 INFO [train.py:1198] (1/4) Epoch 21, batch 3450, loss[loss=0.1898, ctc_loss=0.1241, cr_loss=0.3289, over 17360.00 frames. ], tot_loss[loss=0.2145, ctc_loss=0.1427, cr_loss=0.3594, over 3355509.20 frames. ], batch size: 48, lr: 5.66e-03, grad_scale: 32.0
2024-09-24 00:27:20,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=379773.3333333333, ans=0.125
2024-09-24 00:27:38,510 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.292e+02 1.385e+02 1.500e+02 2.362e+02, threshold=2.770e+02, percent-clipped=0.0
2024-09-24 00:28:00,290 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.34 vs. limit=15.0
2024-09-24 00:28:16,468 INFO [train.py:1198] (1/4) Epoch 21, batch 3500, loss[loss=0.2261, ctc_loss=0.1506, cr_loss=0.3778, over 17323.00 frames. ], tot_loss[loss=0.2143, ctc_loss=0.1425, cr_loss=0.3593, over 3357823.24 frames. ], batch size: 51, lr: 5.66e-03, grad_scale: 32.0
2024-09-24 00:29:02,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=380053.3333333333, ans=0.125
2024-09-24 00:29:02,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=380053.3333333333, ans=0.0
2024-09-24 00:29:03,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.90 vs. limit=15.0
2024-09-24 00:29:14,200 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.42 vs. limit=15.0
2024-09-24 00:29:33,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=380146.6666666667, ans=0.125
2024-09-24 00:29:35,938 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.48 vs. limit=12.0
2024-09-24 00:29:36,654 INFO [train.py:1198] (1/4) Epoch 21, batch 3550, loss[loss=0.1806, ctc_loss=0.1163, cr_loss=0.3216, over 17276.00 frames. ], tot_loss[loss=0.2138, ctc_loss=0.1422, cr_loss=0.3583, over 3361866.12 frames. ], batch size: 42, lr: 5.66e-03, grad_scale: 32.0
2024-09-24 00:29:57,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=380240.0, ans=0.125
2024-09-24 00:29:57,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=380240.0, ans=0.2
2024-09-24 00:30:08,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=380286.6666666667, ans=0.125
2024-09-24 00:30:20,797 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.243e+02 1.324e+02 1.443e+02 2.322e+02, threshold=2.648e+02, percent-clipped=0.0
2024-09-24 00:30:39,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=380380.0, ans=0.125
2024-09-24 00:30:44,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=380380.0, ans=0.0
2024-09-24 00:30:44,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=380380.0, ans=0.2
2024-09-24 00:30:47,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=380380.0, ans=0.0
2024-09-24 00:30:55,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=380426.6666666667, ans=0.1
2024-09-24 00:30:56,886 INFO [train.py:1198] (1/4) Epoch 21, batch 3600, loss[loss=0.2841, ctc_loss=0.1995, cr_loss=0.4234, over 11985.00 frames. ], tot_loss[loss=0.2132, ctc_loss=0.1417, cr_loss=0.3574, over 3363933.78 frames. ], batch size: 123, lr: 5.66e-03, grad_scale: 32.0
2024-09-24 00:31:18,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=380473.3333333333, ans=0.0
2024-09-24 00:31:28,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=380520.0, ans=0.035
2024-09-24 00:31:35,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=380520.0, ans=0.125
2024-09-24 00:31:47,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn2.whiten.whitening_limit, batch_count=380566.6666666667, ans=22.5
2024-09-24 00:31:50,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=380566.6666666667, ans=0.0
2024-09-24 00:31:53,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=380566.6666666667, ans=0.125
2024-09-24 00:31:56,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=380566.6666666667, ans=0.2
2024-09-24 00:32:14,912 INFO [train.py:1198] (1/4) Epoch 21, batch 3650, loss[loss=0.226, ctc_loss=0.1493, cr_loss=0.3835, over 17182.00 frames. ], tot_loss[loss=0.2136, ctc_loss=0.1419, cr_loss=0.3584, over 3363837.57 frames. ], batch size: 45, lr: 5.65e-03, grad_scale: 32.0
2024-09-24 00:32:21,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=380660.0, ans=0.125
2024-09-24 00:32:54,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=380753.3333333333, ans=0.125
2024-09-24 00:32:56,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=380753.3333333333, ans=0.125
2024-09-24 00:32:57,524 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.280e+02 1.369e+02 1.514e+02 2.640e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-24 00:33:04,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=380800.0, ans=0.1
2024-09-24 00:33:09,827 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.49 vs. limit=10.0
2024-09-24 00:33:23,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=380846.6666666667, ans=0.0
2024-09-24 00:33:23,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=380846.6666666667, ans=0.0
2024-09-24 00:33:27,217 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=10.27 vs. limit=22.5
2024-09-24 00:33:27,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=380846.6666666667, ans=0.1
2024-09-24 00:33:35,126 INFO [train.py:1198] (1/4) Epoch 21, batch 3700, loss[loss=0.2038, ctc_loss=0.1388, cr_loss=0.3251, over 17315.00 frames. ], tot_loss[loss=0.2119, ctc_loss=0.1406, cr_loss=0.3564, over 3356705.84 frames. ], batch size: 51, lr: 5.65e-03, grad_scale: 32.0
2024-09-24 00:34:13,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=380986.6666666667, ans=0.2
2024-09-24 00:34:33,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=381033.3333333333, ans=0.125
2024-09-24 00:34:53,989 INFO [train.py:1198] (1/4) Epoch 21, batch 3750, loss[loss=0.1733, ctc_loss=0.1158, cr_loss=0.2872, over 17036.00 frames. ], tot_loss[loss=0.2137, ctc_loss=0.142, cr_loss=0.3585, over 3353009.58 frames. ], batch size: 39, lr: 5.65e-03, grad_scale: 32.0
2024-09-24 00:35:03,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=381126.6666666667, ans=0.125
2024-09-24 00:35:11,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=381173.3333333333, ans=0.1
2024-09-24 00:35:16,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=381173.3333333333, ans=0.0
2024-09-24 00:35:27,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=381220.0, ans=15.0
2024-09-24 00:35:36,265 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.271e+02 1.353e+02 1.516e+02 2.351e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-24 00:35:44,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=381266.6666666667, ans=0.1
2024-09-24 00:35:53,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=381266.6666666667, ans=0.125
2024-09-24 00:36:03,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=381313.3333333333, ans=0.125
2024-09-24 00:36:11,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=381360.0, ans=0.125
2024-09-24 00:36:12,540 INFO [train.py:1198] (1/4) Epoch 21, batch 3800, loss[loss=0.1952, ctc_loss=0.1286, cr_loss=0.3332, over 16664.00 frames. ], tot_loss[loss=0.2153, ctc_loss=0.1432, cr_loss=0.3604, over 3337762.52 frames. ], batch size: 37, lr: 5.65e-03, grad_scale: 32.0
2024-09-24 00:36:19,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=381360.0, ans=0.0
2024-09-24 00:36:50,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=381453.3333333333, ans=0.0
2024-09-24 00:36:53,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=381453.3333333333, ans=0.0
2024-09-24 00:36:53,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=381453.3333333333, ans=0.125
2024-09-24 00:36:57,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=381453.3333333333, ans=0.0
2024-09-24 00:37:01,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=381500.0, ans=0.1
2024-09-24 00:37:06,589 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 00:37:29,109 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.18 vs. limit=22.5
2024-09-24 00:37:31,011 INFO [train.py:1198] (1/4) Epoch 21, batch 3850, loss[loss=0.2148, ctc_loss=0.1412, cr_loss=0.368, over 17145.00 frames. ], tot_loss[loss=0.2156, ctc_loss=0.1435, cr_loss=0.3603, over 3314614.94 frames. ], batch size: 48, lr: 5.65e-03, grad_scale: 32.0
2024-09-24 00:37:31,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=381593.3333333333, ans=0.125
2024-09-24 00:37:50,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=381640.0, ans=0.125
2024-09-24 00:38:11,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=381686.6666666667, ans=0.025
2024-09-24 00:38:14,455 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.386e+02 1.498e+02 1.641e+02 2.451e+02, threshold=2.996e+02, percent-clipped=0.0
2024-09-24 00:38:24,824 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=20.74 vs. limit=22.5
2024-09-24 00:38:41,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=381780.0, ans=0.125
2024-09-24 00:39:39,374 INFO [train.py:1198] (1/4) Epoch 22, batch 0, loss[loss=0.1957, ctc_loss=0.1312, cr_loss=0.3225, over 17095.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1312, cr_loss=0.3225, over 17095.00 frames. ], batch size: 49, lr: 5.51e-03, grad_scale: 32.0
2024-09-24 00:39:39,375 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 00:39:54,642 INFO [train.py:1230] (1/4) Epoch 22, validation: loss=0.03827, ctc_loss=0.03827, cr_loss=8.092e-15, over 944034.00 frames. 
2024-09-24 00:39:54,642 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 00:39:58,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=381808.0, ans=0.2
2024-09-24 00:40:07,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=381808.0, ans=0.0
2024-09-24 00:40:13,036 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.31 vs. limit=15.0
2024-09-24 00:40:26,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=381854.6666666667, ans=0.0
2024-09-24 00:40:26,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=381854.6666666667, ans=0.0
2024-09-24 00:40:56,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=381948.0, ans=0.1
2024-09-24 00:41:11,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=381994.6666666667, ans=0.125
2024-09-24 00:41:17,586 INFO [train.py:1198] (1/4) Epoch 22, batch 50, loss[loss=0.1869, ctc_loss=0.1258, cr_loss=0.3055, over 17179.00 frames. ], tot_loss[loss=0.2094, ctc_loss=0.1389, cr_loss=0.3522, over 755830.28 frames. ], batch size: 45, lr: 5.51e-03, grad_scale: 32.0
2024-09-24 00:41:47,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=382134.6666666667, ans=0.0
2024-09-24 00:42:06,803 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.269e+02 1.377e+02 1.581e+02 4.736e+02, threshold=2.753e+02, percent-clipped=1.0
2024-09-24 00:42:07,754 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=8.63 vs. limit=15.0
2024-09-24 00:42:10,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=382181.3333333333, ans=0.025
2024-09-24 00:42:18,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.63 vs. limit=15.0
2024-09-24 00:42:23,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=382228.0, ans=0.1
2024-09-24 00:42:26,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=382228.0, ans=0.1
2024-09-24 00:42:27,351 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.38 vs. limit=5.0
2024-09-24 00:42:38,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=382274.6666666667, ans=10.0
2024-09-24 00:42:40,205 INFO [train.py:1198] (1/4) Epoch 22, batch 100, loss[loss=0.188, ctc_loss=0.1259, cr_loss=0.3102, over 17096.00 frames. ], tot_loss[loss=0.2109, ctc_loss=0.14, cr_loss=0.3546, over 1330550.59 frames. ], batch size: 43, lr: 5.51e-03, grad_scale: 32.0
2024-09-24 00:42:48,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=382274.6666666667, ans=0.125
2024-09-24 00:43:59,805 INFO [train.py:1198] (1/4) Epoch 22, batch 150, loss[loss=0.2064, ctc_loss=0.1368, cr_loss=0.348, over 16786.00 frames. ], tot_loss[loss=0.2123, ctc_loss=0.1407, cr_loss=0.358, over 1786735.30 frames. ], batch size: 61, lr: 5.51e-03, grad_scale: 32.0
2024-09-24 00:44:12,868 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.13 vs. limit=15.0
2024-09-24 00:44:34,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=382554.6666666667, ans=0.2
2024-09-24 00:44:52,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=382648.0, ans=0.125
2024-09-24 00:44:55,447 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.262e+02 1.352e+02 1.515e+02 2.166e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-24 00:45:05,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=382648.0, ans=0.07
2024-09-24 00:45:29,258 INFO [train.py:1198] (1/4) Epoch 22, batch 200, loss[loss=0.2212, ctc_loss=0.1487, cr_loss=0.3624, over 17221.00 frames. ], tot_loss[loss=0.2125, ctc_loss=0.1408, cr_loss=0.3585, over 2132844.37 frames. ], batch size: 50, lr: 5.51e-03, grad_scale: 32.0
2024-09-24 00:45:29,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=382741.3333333333, ans=0.5
2024-09-24 00:46:05,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=382834.6666666667, ans=0.1
2024-09-24 00:46:31,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=382928.0, ans=0.125
2024-09-24 00:46:48,688 INFO [train.py:1198] (1/4) Epoch 22, batch 250, loss[loss=0.2118, ctc_loss=0.1384, cr_loss=0.367, over 17067.00 frames. ], tot_loss[loss=0.2141, ctc_loss=0.1421, cr_loss=0.3601, over 2395352.42 frames. ], batch size: 46, lr: 5.50e-03, grad_scale: 32.0
2024-09-24 00:47:41,218 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.260e+02 1.348e+02 1.458e+02 2.828e+02, threshold=2.696e+02, percent-clipped=1.0
2024-09-24 00:47:45,654 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.12 vs. limit=15.0
2024-09-24 00:47:55,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=383161.3333333333, ans=0.07
2024-09-24 00:47:57,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=383161.3333333333, ans=0.025
2024-09-24 00:48:08,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=383161.3333333333, ans=0.025
2024-09-24 00:48:11,507 INFO [train.py:1198] (1/4) Epoch 22, batch 300, loss[loss=0.2068, ctc_loss=0.1338, cr_loss=0.3647, over 17162.00 frames. ], tot_loss[loss=0.2127, ctc_loss=0.1411, cr_loss=0.3581, over 2606010.09 frames. ], batch size: 45, lr: 5.50e-03, grad_scale: 32.0
2024-09-24 00:48:31,912 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.54 vs. limit=8.0
2024-09-24 00:48:55,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=383301.3333333333, ans=0.125
2024-09-24 00:49:02,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=383348.0, ans=0.125
2024-09-24 00:49:22,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=383394.6666666667, ans=0.125
2024-09-24 00:49:30,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=383394.6666666667, ans=0.125
2024-09-24 00:49:32,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=383394.6666666667, ans=0.2
2024-09-24 00:49:37,075 INFO [train.py:1198] (1/4) Epoch 22, batch 350, loss[loss=0.2091, ctc_loss=0.1374, cr_loss=0.3586, over 17160.00 frames. ], tot_loss[loss=0.2116, ctc_loss=0.1403, cr_loss=0.3561, over 2770917.55 frames. ], batch size: 45, lr: 5.50e-03, grad_scale: 32.0
2024-09-24 00:49:42,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=383441.3333333333, ans=0.2
2024-09-24 00:50:12,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=383534.6666666667, ans=0.125
2024-09-24 00:50:28,764 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.255e+02 1.333e+02 1.486e+02 2.174e+02, threshold=2.666e+02, percent-clipped=0.0
2024-09-24 00:50:32,922 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.05 vs. limit=15.0
2024-09-24 00:50:43,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=383628.0, ans=0.125
2024-09-24 00:50:46,008 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.74 vs. limit=22.5
2024-09-24 00:50:59,528 INFO [train.py:1198] (1/4) Epoch 22, batch 400, loss[loss=0.2328, ctc_loss=0.153, cr_loss=0.3991, over 17046.00 frames. ], tot_loss[loss=0.2111, ctc_loss=0.1399, cr_loss=0.3558, over 2907889.93 frames. ], batch size: 46, lr: 5.50e-03, grad_scale: 32.0
2024-09-24 00:51:05,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=17.33 vs. limit=22.5
2024-09-24 00:51:34,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=383768.0, ans=0.025
2024-09-24 00:52:19,399 INFO [train.py:1198] (1/4) Epoch 22, batch 450, loss[loss=0.2501, ctc_loss=0.1684, cr_loss=0.4083, over 17202.00 frames. ], tot_loss[loss=0.2109, ctc_loss=0.1398, cr_loss=0.3554, over 3013373.05 frames. ], batch size: 55, lr: 5.50e-03, grad_scale: 32.0
2024-09-24 00:52:23,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=383908.0, ans=0.07
2024-09-24 00:52:48,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=383954.6666666667, ans=0.0
2024-09-24 00:52:51,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=383954.6666666667, ans=0.125
2024-09-24 00:52:59,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=384001.3333333333, ans=0.125
2024-09-24 00:53:05,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=384001.3333333333, ans=0.0
2024-09-24 00:53:11,769 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.289e+02 1.376e+02 1.526e+02 3.562e+02, threshold=2.753e+02, percent-clipped=1.0
2024-09-24 00:53:41,446 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.64 vs. limit=15.0
2024-09-24 00:53:41,925 INFO [train.py:1198] (1/4) Epoch 22, batch 500, loss[loss=0.2351, ctc_loss=0.1586, cr_loss=0.3826, over 14840.00 frames. ], tot_loss[loss=0.2119, ctc_loss=0.1406, cr_loss=0.3568, over 3085141.56 frames. ], batch size: 89, lr: 5.50e-03, grad_scale: 32.0
2024-09-24 00:53:42,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=384141.3333333333, ans=0.125
2024-09-24 00:54:46,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=384281.3333333333, ans=0.125
2024-09-24 00:54:53,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=384328.0, ans=0.125
2024-09-24 00:54:59,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=384328.0, ans=0.125
2024-09-24 00:55:00,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=384328.0, ans=0.125
2024-09-24 00:55:04,514 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=14.31 vs. limit=22.5
2024-09-24 00:55:07,115 INFO [train.py:1198] (1/4) Epoch 22, batch 550, loss[loss=0.2191, ctc_loss=0.1405, cr_loss=0.3932, over 17240.00 frames. ], tot_loss[loss=0.2114, ctc_loss=0.1401, cr_loss=0.3562, over 3150844.46 frames. ], batch size: 50, lr: 5.49e-03, grad_scale: 32.0
2024-09-24 00:55:07,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=384374.6666666667, ans=0.125
2024-09-24 00:55:34,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=384421.3333333333, ans=0.1
2024-09-24 00:55:38,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=384421.3333333333, ans=0.0
2024-09-24 00:55:59,357 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.258e+02 1.357e+02 1.512e+02 2.238e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-24 00:56:03,601 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=11.31 vs. limit=22.5
2024-09-24 00:56:28,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=384608.0, ans=0.1
2024-09-24 00:56:30,180 INFO [train.py:1198] (1/4) Epoch 22, batch 600, loss[loss=0.2382, ctc_loss=0.1587, cr_loss=0.3972, over 17190.00 frames. ], tot_loss[loss=0.2115, ctc_loss=0.1402, cr_loss=0.3567, over 3199853.86 frames. ], batch size: 55, lr: 5.49e-03, grad_scale: 32.0
2024-09-24 00:56:33,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=384608.0, ans=0.1
2024-09-24 00:57:05,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=384701.3333333333, ans=0.07
2024-09-24 00:57:15,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=384701.3333333333, ans=0.0
2024-09-24 00:57:36,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=384794.6666666667, ans=0.1
2024-09-24 00:57:37,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.72 vs. limit=15.0
2024-09-24 00:57:39,282 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=15.23 vs. limit=22.5
2024-09-24 00:57:52,482 INFO [train.py:1198] (1/4) Epoch 22, batch 650, loss[loss=0.2048, ctc_loss=0.1354, cr_loss=0.3471, over 17091.00 frames. ], tot_loss[loss=0.2122, ctc_loss=0.1409, cr_loss=0.3569, over 3226313.89 frames. ], batch size: 43, lr: 5.49e-03, grad_scale: 16.0
2024-09-24 00:58:08,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=384888.0, ans=0.0
2024-09-24 00:58:26,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=384934.6666666667, ans=0.025
2024-09-24 00:58:31,205 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 00:58:39,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=384981.3333333333, ans=0.125
2024-09-24 00:58:43,614 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.285e+02 1.401e+02 1.551e+02 2.497e+02, threshold=2.802e+02, percent-clipped=0.0
2024-09-24 00:58:50,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=384981.3333333333, ans=0.125
2024-09-24 00:58:51,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=16.93 vs. limit=22.5
2024-09-24 00:58:57,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=385028.0, ans=0.0
2024-09-24 00:58:58,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=385028.0, ans=0.015
2024-09-24 00:58:58,731 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 00:59:06,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=385028.0, ans=0.0
2024-09-24 00:59:15,116 INFO [train.py:1198] (1/4) Epoch 22, batch 700, loss[loss=0.2291, ctc_loss=0.1529, cr_loss=0.3813, over 16997.00 frames. ], tot_loss[loss=0.212, ctc_loss=0.1407, cr_loss=0.3566, over 3256174.73 frames. ], batch size: 56, lr: 5.49e-03, grad_scale: 16.0
2024-09-24 00:59:17,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=385074.6666666667, ans=22.5
2024-09-24 01:00:01,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=385168.0, ans=0.0
2024-09-24 01:00:20,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=385214.6666666667, ans=0.1
2024-09-24 01:00:40,659 INFO [train.py:1198] (1/4) Epoch 22, batch 750, loss[loss=0.1895, ctc_loss=0.1238, cr_loss=0.3286, over 17254.00 frames. ], tot_loss[loss=0.2126, ctc_loss=0.1411, cr_loss=0.3571, over 3284025.47 frames. ], batch size: 44, lr: 5.49e-03, grad_scale: 8.0
2024-09-24 01:01:03,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=385354.6666666667, ans=0.125
2024-09-24 01:01:13,442 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.79 vs. limit=22.5
2024-09-24 01:01:27,625 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=6.21 vs. limit=15.0
2024-09-24 01:01:30,241 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:01:33,050 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.232e+02 1.335e+02 1.516e+02 2.427e+02, threshold=2.671e+02, percent-clipped=0.0
2024-09-24 01:01:54,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=385494.6666666667, ans=0.1
2024-09-24 01:02:00,313 INFO [train.py:1198] (1/4) Epoch 22, batch 800, loss[loss=0.1904, ctc_loss=0.1233, cr_loss=0.3354, over 17094.00 frames. ], tot_loss[loss=0.213, ctc_loss=0.1415, cr_loss=0.3579, over 3290746.95 frames. ], batch size: 40, lr: 5.49e-03, grad_scale: 16.0
2024-09-24 01:02:11,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=385541.3333333333, ans=0.125
2024-09-24 01:02:44,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=385634.6666666667, ans=0.1
2024-09-24 01:02:51,794 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.98 vs. limit=15.0
2024-09-24 01:03:02,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=385681.3333333333, ans=0.0
2024-09-24 01:03:17,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=385728.0, ans=0.0
2024-09-24 01:03:23,165 INFO [train.py:1198] (1/4) Epoch 22, batch 850, loss[loss=0.2058, ctc_loss=0.1347, cr_loss=0.3553, over 17285.00 frames. ], tot_loss[loss=0.2125, ctc_loss=0.1412, cr_loss=0.3567, over 3305935.68 frames. ], batch size: 46, lr: 5.48e-03, grad_scale: 16.0
2024-09-24 01:03:23,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=385774.6666666667, ans=0.1
2024-09-24 01:03:44,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=385821.3333333333, ans=0.125
2024-09-24 01:03:45,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=385821.3333333333, ans=0.125
2024-09-24 01:03:53,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.82 vs. limit=15.0
2024-09-24 01:04:03,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=385868.0, ans=0.125
2024-09-24 01:04:09,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=385868.0, ans=0.1
2024-09-24 01:04:18,868 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.270e+02 1.386e+02 1.514e+02 2.172e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-24 01:04:48,969 INFO [train.py:1198] (1/4) Epoch 22, batch 900, loss[loss=0.2192, ctc_loss=0.1465, cr_loss=0.3632, over 17302.00 frames. ], tot_loss[loss=0.2121, ctc_loss=0.1408, cr_loss=0.3565, over 3317767.12 frames. ], batch size: 51, lr: 5.48e-03, grad_scale: 16.0
2024-09-24 01:04:49,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=386008.0, ans=0.2
2024-09-24 01:05:07,545 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.88 vs. limit=22.5
2024-09-24 01:05:11,771 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:05:35,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=386101.3333333333, ans=0.0
2024-09-24 01:06:06,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=12.19 vs. limit=15.0
2024-09-24 01:06:11,837 INFO [train.py:1198] (1/4) Epoch 22, batch 950, loss[loss=0.1953, ctc_loss=0.1287, cr_loss=0.3331, over 17246.00 frames. ], tot_loss[loss=0.2119, ctc_loss=0.1406, cr_loss=0.3568, over 3337803.54 frames. ], batch size: 42, lr: 5.48e-03, grad_scale: 16.0
2024-09-24 01:06:42,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=386334.6666666667, ans=0.0
2024-09-24 01:06:45,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=386334.6666666667, ans=0.125
2024-09-24 01:07:02,605 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:07:03,757 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.224e+02 1.313e+02 1.395e+02 1.890e+02, threshold=2.625e+02, percent-clipped=0.0
2024-09-24 01:07:31,055 INFO [train.py:1198] (1/4) Epoch 22, batch 1000, loss[loss=0.2367, ctc_loss=0.1592, cr_loss=0.3878, over 16724.00 frames. ], tot_loss[loss=0.2106, ctc_loss=0.1397, cr_loss=0.3549, over 3350987.89 frames. ], batch size: 61, lr: 5.48e-03, grad_scale: 16.0
2024-09-24 01:07:31,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=386474.6666666667, ans=0.1
2024-09-24 01:07:51,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=386521.3333333333, ans=0.125
2024-09-24 01:08:53,198 INFO [train.py:1198] (1/4) Epoch 22, batch 1050, loss[loss=0.1949, ctc_loss=0.1289, cr_loss=0.3298, over 17297.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1389, cr_loss=0.3533, over 3357429.03 frames. ], batch size: 51, lr: 5.48e-03, grad_scale: 16.0
2024-09-24 01:09:13,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=386754.6666666667, ans=0.0
2024-09-24 01:09:21,663 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=6.12 vs. limit=12.0
2024-09-24 01:09:22,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=386754.6666666667, ans=0.125
2024-09-24 01:09:50,909 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.253e+02 1.357e+02 1.506e+02 3.378e+02, threshold=2.715e+02, percent-clipped=1.0
2024-09-24 01:09:57,888 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.59 vs. limit=12.0
2024-09-24 01:10:20,569 INFO [train.py:1198] (1/4) Epoch 22, batch 1100, loss[loss=0.1816, ctc_loss=0.1176, cr_loss=0.32, over 17070.00 frames. ], tot_loss[loss=0.2094, ctc_loss=0.1388, cr_loss=0.3528, over 3366370.86 frames. ], batch size: 46, lr: 5.48e-03, grad_scale: 16.0
2024-09-24 01:10:28,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=386941.3333333333, ans=0.0
2024-09-24 01:10:37,862 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.90 vs. limit=22.5
2024-09-24 01:11:07,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=387081.3333333333, ans=0.2
2024-09-24 01:11:07,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=387081.3333333333, ans=0.07
2024-09-24 01:11:21,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=387081.3333333333, ans=0.2
2024-09-24 01:11:23,912 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.40 vs. limit=6.0
2024-09-24 01:11:40,357 INFO [train.py:1198] (1/4) Epoch 22, batch 1150, loss[loss=0.2078, ctc_loss=0.1376, cr_loss=0.3508, over 17148.00 frames. ], tot_loss[loss=0.209, ctc_loss=0.1385, cr_loss=0.3527, over 3371481.54 frames. ], batch size: 48, lr: 5.47e-03, grad_scale: 16.0
2024-09-24 01:11:53,634 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:12:04,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=387221.3333333333, ans=0.125
2024-09-24 01:12:21,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=387268.0, ans=0.0
2024-09-24 01:12:36,141 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.223e+02 1.330e+02 1.443e+02 2.592e+02, threshold=2.661e+02, percent-clipped=0.0
2024-09-24 01:12:52,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=387361.3333333333, ans=0.125
2024-09-24 01:13:03,659 INFO [train.py:1198] (1/4) Epoch 22, batch 1200, loss[loss=0.1947, ctc_loss=0.1277, cr_loss=0.3353, over 17150.00 frames. ], tot_loss[loss=0.2091, ctc_loss=0.1383, cr_loss=0.3537, over 3372604.26 frames. ], batch size: 48, lr: 5.47e-03, grad_scale: 32.0
2024-09-24 01:13:10,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=387408.0, ans=0.125
2024-09-24 01:13:45,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=387501.3333333333, ans=0.07
2024-09-24 01:14:05,180 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.49 vs. limit=15.0
2024-09-24 01:14:25,965 INFO [train.py:1198] (1/4) Epoch 22, batch 1250, loss[loss=0.2206, ctc_loss=0.1486, cr_loss=0.36, over 16522.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1388, cr_loss=0.3539, over 3361318.68 frames. ], batch size: 66, lr: 5.47e-03, grad_scale: 32.0
2024-09-24 01:14:40,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=387641.3333333333, ans=0.0
2024-09-24 01:14:44,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=387688.0, ans=0.0
2024-09-24 01:14:54,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=387688.0, ans=0.1
2024-09-24 01:14:59,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=387734.6666666667, ans=0.125
2024-09-24 01:15:23,793 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.292e+02 1.416e+02 1.548e+02 3.016e+02, threshold=2.831e+02, percent-clipped=1.0
2024-09-24 01:15:24,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=387781.3333333333, ans=0.2
2024-09-24 01:15:30,746 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=15.79 vs. limit=22.5
2024-09-24 01:15:46,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=387828.0, ans=0.125
2024-09-24 01:15:48,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=387828.0, ans=0.0
2024-09-24 01:15:48,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=387828.0, ans=0.1
2024-09-24 01:15:50,985 INFO [train.py:1198] (1/4) Epoch 22, batch 1300, loss[loss=0.1987, ctc_loss=0.1307, cr_loss=0.3402, over 17357.00 frames. ], tot_loss[loss=0.2105, ctc_loss=0.1394, cr_loss=0.3553, over 3366021.77 frames. ], batch size: 48, lr: 5.47e-03, grad_scale: 32.0
2024-09-24 01:16:34,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=387968.0, ans=0.125
2024-09-24 01:16:56,249 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:16:57,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=388061.3333333333, ans=0.0
2024-09-24 01:17:07,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=388061.3333333333, ans=0.2
2024-09-24 01:17:10,176 INFO [train.py:1198] (1/4) Epoch 22, batch 1350, loss[loss=0.2211, ctc_loss=0.1484, cr_loss=0.3632, over 16456.00 frames. ], tot_loss[loss=0.2124, ctc_loss=0.1409, cr_loss=0.3575, over 3348675.96 frames. ], batch size: 66, lr: 5.47e-03, grad_scale: 32.0
2024-09-24 01:17:29,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=388154.6666666667, ans=0.125
2024-09-24 01:17:29,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=388154.6666666667, ans=0.0
2024-09-24 01:17:32,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=388154.6666666667, ans=0.125
2024-09-24 01:17:51,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=388201.3333333333, ans=0.0
2024-09-24 01:18:05,574 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.287e+02 1.390e+02 1.521e+02 2.749e+02, threshold=2.781e+02, percent-clipped=0.0
2024-09-24 01:18:30,712 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.79 vs. limit=15.0
2024-09-24 01:18:32,814 INFO [train.py:1198] (1/4) Epoch 22, batch 1400, loss[loss=0.1845, ctc_loss=0.1202, cr_loss=0.3216, over 17015.00 frames. ], tot_loss[loss=0.212, ctc_loss=0.1406, cr_loss=0.3569, over 3347854.61 frames. ], batch size: 44, lr: 5.47e-03, grad_scale: 32.0
2024-09-24 01:18:47,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=388388.0, ans=0.0
2024-09-24 01:18:49,257 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:18:53,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=388388.0, ans=0.1
2024-09-24 01:19:00,588 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.05 vs. limit=15.0
2024-09-24 01:19:07,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=388434.6666666667, ans=0.1
2024-09-24 01:19:30,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=388481.3333333333, ans=0.1
2024-09-24 01:19:45,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=388528.0, ans=0.125
2024-09-24 01:19:46,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=388528.0, ans=0.0
2024-09-24 01:19:49,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=388528.0, ans=0.2
2024-09-24 01:19:57,777 INFO [train.py:1198] (1/4) Epoch 22, batch 1450, loss[loss=0.2306, ctc_loss=0.1546, cr_loss=0.3802, over 17091.00 frames. ], tot_loss[loss=0.2111, ctc_loss=0.14, cr_loss=0.3557, over 3350152.78 frames. ], batch size: 49, lr: 5.46e-03, grad_scale: 32.0
2024-09-24 01:20:00,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=388574.6666666667, ans=0.2
2024-09-24 01:20:07,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=388574.6666666667, ans=0.0
2024-09-24 01:20:24,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=388621.3333333333, ans=0.07
2024-09-24 01:20:48,425 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:20:52,837 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.236e+02 1.340e+02 1.484e+02 2.143e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-24 01:21:10,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=388761.3333333333, ans=0.125
2024-09-24 01:21:15,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=388761.3333333333, ans=0.05
2024-09-24 01:21:17,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=388761.3333333333, ans=0.125
2024-09-24 01:21:18,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=388808.0, ans=0.125
2024-09-24 01:21:20,282 INFO [train.py:1198] (1/4) Epoch 22, batch 1500, loss[loss=0.2071, ctc_loss=0.1359, cr_loss=0.3559, over 17310.00 frames. ], tot_loss[loss=0.2109, ctc_loss=0.1397, cr_loss=0.3557, over 3359359.33 frames. ], batch size: 51, lr: 5.46e-03, grad_scale: 32.0
2024-09-24 01:21:25,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=388808.0, ans=0.0
2024-09-24 01:21:41,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=388854.6666666667, ans=0.125
2024-09-24 01:22:00,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.90 vs. limit=22.5
2024-09-24 01:22:33,623 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.90 vs. limit=15.0
2024-09-24 01:22:42,364 INFO [train.py:1198] (1/4) Epoch 22, batch 1550, loss[loss=0.1933, ctc_loss=0.1275, cr_loss=0.3287, over 17123.00 frames. ], tot_loss[loss=0.2106, ctc_loss=0.1394, cr_loss=0.3558, over 3365285.86 frames. ], batch size: 40, lr: 5.46e-03, grad_scale: 16.0
2024-09-24 01:23:00,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=389088.0, ans=0.125
2024-09-24 01:23:14,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=389134.6666666667, ans=0.125
2024-09-24 01:23:35,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=389181.3333333333, ans=0.0
2024-09-24 01:23:37,030 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.256e+02 1.352e+02 1.468e+02 3.649e+02, threshold=2.704e+02, percent-clipped=1.0
2024-09-24 01:24:02,774 INFO [train.py:1198] (1/4) Epoch 22, batch 1600, loss[loss=0.2132, ctc_loss=0.1449, cr_loss=0.3418, over 16950.00 frames. ], tot_loss[loss=0.2109, ctc_loss=0.1397, cr_loss=0.3562, over 3365875.95 frames. ], batch size: 58, lr: 5.46e-03, grad_scale: 32.0
2024-09-24 01:24:08,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=389274.6666666667, ans=0.125
2024-09-24 01:24:23,403 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.64 vs. limit=10.0
2024-09-24 01:24:26,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=389321.3333333333, ans=0.025
2024-09-24 01:24:38,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=389368.0, ans=0.125
2024-09-24 01:24:43,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=389368.0, ans=0.1
2024-09-24 01:24:57,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.43 vs. limit=12.0
2024-09-24 01:25:10,039 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.16 vs. limit=10.0
2024-09-24 01:25:10,128 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.72 vs. limit=6.0
2024-09-24 01:25:13,499 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.55 vs. limit=10.0
2024-09-24 01:25:15,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=389461.3333333333, ans=0.125
2024-09-24 01:25:23,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=389461.3333333333, ans=0.125
2024-09-24 01:25:29,441 INFO [train.py:1198] (1/4) Epoch 22, batch 1650, loss[loss=0.2196, ctc_loss=0.1436, cr_loss=0.3801, over 17076.00 frames. ], tot_loss[loss=0.2115, ctc_loss=0.1401, cr_loss=0.3567, over 3349228.24 frames. ], batch size: 46, lr: 5.46e-03, grad_scale: 32.0
2024-09-24 01:25:31,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=389508.0, ans=0.0
2024-09-24 01:25:37,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=389508.0, ans=10.0
2024-09-24 01:25:53,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=389554.6666666667, ans=0.125
2024-09-24 01:26:16,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=389648.0, ans=0.0
2024-09-24 01:26:23,522 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.218e+02 1.295e+02 1.408e+02 1.987e+02, threshold=2.589e+02, percent-clipped=0.0
2024-09-24 01:26:41,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=389694.6666666667, ans=0.0
2024-09-24 01:26:43,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=389694.6666666667, ans=0.125
2024-09-24 01:26:44,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.max_positive, batch_count=389694.6666666667, ans=0.95
2024-09-24 01:26:49,450 INFO [train.py:1198] (1/4) Epoch 22, batch 1700, loss[loss=0.2355, ctc_loss=0.1572, cr_loss=0.3915, over 17005.00 frames. ], tot_loss[loss=0.2118, ctc_loss=0.1404, cr_loss=0.3568, over 3351631.11 frames. ], batch size: 53, lr: 5.46e-03, grad_scale: 32.0
2024-09-24 01:27:24,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=389834.6666666667, ans=0.2
2024-09-24 01:27:31,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=389834.6666666667, ans=0.125
2024-09-24 01:27:40,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=389881.3333333333, ans=0.125
2024-09-24 01:27:43,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=389881.3333333333, ans=0.0
2024-09-24 01:27:45,679 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.19 vs. limit=15.0
2024-09-24 01:27:55,364 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=11.39 vs. limit=22.5
2024-09-24 01:28:04,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=389928.0, ans=0.0
2024-09-24 01:28:09,291 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=389928.0, ans=0.0
2024-09-24 01:28:12,340 INFO [train.py:1198] (1/4) Epoch 22, batch 1750, loss[loss=0.1904, ctc_loss=0.1242, cr_loss=0.3311, over 17274.00 frames. ], tot_loss[loss=0.2122, ctc_loss=0.1407, cr_loss=0.3574, over 3361160.12 frames. ], batch size: 44, lr: 5.46e-03, grad_scale: 32.0
2024-09-24 01:28:12,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=389974.6666666667, ans=0.125
2024-09-24 01:28:18,061 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.91 vs. limit=22.5
2024-09-24 01:28:30,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=390021.3333333333, ans=0.125
2024-09-24 01:28:55,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=390068.0, ans=0.1
2024-09-24 01:29:09,553 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.027e+02 1.289e+02 1.384e+02 1.529e+02 2.458e+02, threshold=2.768e+02, percent-clipped=0.0
2024-09-24 01:29:37,592 INFO [train.py:1198] (1/4) Epoch 22, batch 1800, loss[loss=0.1859, ctc_loss=0.1203, cr_loss=0.3278, over 17193.00 frames. ], tot_loss[loss=0.2137, ctc_loss=0.1418, cr_loss=0.3591, over 3351467.96 frames. ], batch size: 41, lr: 5.45e-03, grad_scale: 32.0
2024-09-24 01:29:41,861 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=5.60 vs. limit=15.0
2024-09-24 01:30:09,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=390254.6666666667, ans=0.1
2024-09-24 01:30:20,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=390301.3333333333, ans=0.125
2024-09-24 01:30:22,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=390301.3333333333, ans=0.07
2024-09-24 01:30:38,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=390348.0, ans=0.125
2024-09-24 01:30:44,633 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:31:00,215 INFO [train.py:1198] (1/4) Epoch 22, batch 1850, loss[loss=0.2195, ctc_loss=0.1474, cr_loss=0.3605, over 17262.00 frames. ], tot_loss[loss=0.2131, ctc_loss=0.1415, cr_loss=0.3581, over 3356412.13 frames. ], batch size: 44, lr: 5.45e-03, grad_scale: 32.0
2024-09-24 01:31:13,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=390441.3333333333, ans=0.125
2024-09-24 01:31:14,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=390488.0, ans=0.025
2024-09-24 01:31:21,184 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:31:29,279 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.39 vs. limit=15.0
2024-09-24 01:31:43,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=390534.6666666667, ans=0.0
2024-09-24 01:31:50,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=390581.3333333333, ans=0.125
2024-09-24 01:31:54,070 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.256e+02 1.338e+02 1.460e+02 2.391e+02, threshold=2.675e+02, percent-clipped=0.0
2024-09-24 01:31:56,691 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.39 vs. limit=15.0
2024-09-24 01:32:07,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=390628.0, ans=0.1
2024-09-24 01:32:11,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=390628.0, ans=0.125
2024-09-24 01:32:11,308 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:32:21,846 INFO [train.py:1198] (1/4) Epoch 22, batch 1900, loss[loss=0.2148, ctc_loss=0.1426, cr_loss=0.3609, over 17006.00 frames. ], tot_loss[loss=0.2137, ctc_loss=0.142, cr_loss=0.3588, over 3351140.00 frames. ], batch size: 44, lr: 5.45e-03, grad_scale: 32.0
2024-09-24 01:32:35,385 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.53 vs. limit=15.0
2024-09-24 01:33:19,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=390814.6666666667, ans=0.125
2024-09-24 01:33:25,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.60 vs. limit=15.0
2024-09-24 01:33:35,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=390861.3333333333, ans=0.1
2024-09-24 01:33:41,708 INFO [train.py:1198] (1/4) Epoch 22, batch 1950, loss[loss=0.2275, ctc_loss=0.1497, cr_loss=0.3889, over 16869.00 frames. ], tot_loss[loss=0.2142, ctc_loss=0.1423, cr_loss=0.3597, over 3356857.39 frames. ], batch size: 58, lr: 5.45e-03, grad_scale: 32.0
2024-09-24 01:33:52,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=390908.0, ans=0.125
2024-09-24 01:34:14,715 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.63 vs. limit=15.0
2024-09-24 01:34:22,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=391001.3333333333, ans=0.05
2024-09-24 01:34:41,243 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.293e+02 1.356e+02 1.523e+02 3.316e+02, threshold=2.712e+02, percent-clipped=1.0
2024-09-24 01:35:09,057 INFO [train.py:1198] (1/4) Epoch 22, batch 2000, loss[loss=0.2433, ctc_loss=0.1664, cr_loss=0.3844, over 15871.00 frames. ], tot_loss[loss=0.2149, ctc_loss=0.1429, cr_loss=0.3601, over 3342327.04 frames. ], batch size: 74, lr: 5.45e-03, grad_scale: 32.0
2024-09-24 01:35:15,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=391141.3333333333, ans=0.125
2024-09-24 01:35:20,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=391141.3333333333, ans=0.2
2024-09-24 01:35:31,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=391188.0, ans=0.2
2024-09-24 01:35:32,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=391188.0, ans=0.1
2024-09-24 01:35:37,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=391188.0, ans=0.0
2024-09-24 01:36:13,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=391328.0, ans=0.125
2024-09-24 01:36:28,617 INFO [train.py:1198] (1/4) Epoch 22, batch 2050, loss[loss=0.2104, ctc_loss=0.1397, cr_loss=0.3537, over 17105.00 frames. ], tot_loss[loss=0.215, ctc_loss=0.1429, cr_loss=0.3605, over 3345574.95 frames. ], batch size: 49, lr: 5.45e-03, grad_scale: 16.0
2024-09-24 01:36:30,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=391374.6666666667, ans=0.125
2024-09-24 01:36:34,642 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=10.01 vs. limit=15.0
2024-09-24 01:36:35,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=391374.6666666667, ans=0.125
2024-09-24 01:37:07,055 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.12 vs. limit=10.0
2024-09-24 01:37:10,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=391468.0, ans=0.1
2024-09-24 01:37:24,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=391514.6666666667, ans=0.125
2024-09-24 01:37:27,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.255e+02 1.364e+02 1.443e+02 3.272e+02, threshold=2.728e+02, percent-clipped=1.0
2024-09-24 01:37:51,062 INFO [train.py:1198] (1/4) Epoch 22, batch 2100, loss[loss=0.2559, ctc_loss=0.1749, cr_loss=0.4052, over 16586.00 frames. ], tot_loss[loss=0.2151, ctc_loss=0.1429, cr_loss=0.361, over 3346956.56 frames. ], batch size: 66, lr: 5.44e-03, grad_scale: 16.0
2024-09-24 01:38:10,648 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.55 vs. limit=22.5
2024-09-24 01:38:18,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=391654.6666666667, ans=0.0
2024-09-24 01:38:26,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=391701.3333333333, ans=0.125
2024-09-24 01:39:16,044 INFO [train.py:1198] (1/4) Epoch 22, batch 2150, loss[loss=0.2214, ctc_loss=0.1462, cr_loss=0.376, over 17005.00 frames. ], tot_loss[loss=0.2141, ctc_loss=0.142, cr_loss=0.3604, over 3351342.10 frames. ], batch size: 51, lr: 5.44e-03, grad_scale: 16.0
2024-09-24 01:39:26,157 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.80 vs. limit=15.0
2024-09-24 01:39:32,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=391888.0, ans=0.07
2024-09-24 01:39:58,955 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.61 vs. limit=15.0
2024-09-24 01:40:07,795 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.87 vs. limit=5.0
2024-09-24 01:40:16,908 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.257e+02 1.348e+02 1.506e+02 2.274e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-24 01:40:32,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=392028.0, ans=0.025
2024-09-24 01:40:40,616 INFO [train.py:1198] (1/4) Epoch 22, batch 2200, loss[loss=0.1998, ctc_loss=0.1328, cr_loss=0.3354, over 17037.00 frames. ], tot_loss[loss=0.214, ctc_loss=0.142, cr_loss=0.3601, over 3348677.56 frames. ], batch size: 44, lr: 5.44e-03, grad_scale: 16.0
2024-09-24 01:40:40,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=392074.6666666667, ans=0.0
2024-09-24 01:40:47,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=392074.6666666667, ans=0.2
2024-09-24 01:40:56,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=392121.3333333333, ans=0.125
2024-09-24 01:41:13,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.58 vs. limit=15.0
2024-09-24 01:41:15,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=392168.0, ans=0.1
2024-09-24 01:42:03,260 INFO [train.py:1198] (1/4) Epoch 22, batch 2250, loss[loss=0.2225, ctc_loss=0.1464, cr_loss=0.3804, over 17256.00 frames. ], tot_loss[loss=0.2129, ctc_loss=0.1412, cr_loss=0.3585, over 3352298.62 frames. ], batch size: 44, lr: 5.44e-03, grad_scale: 16.0
2024-09-24 01:42:06,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=392308.0, ans=0.2
2024-09-24 01:42:08,750 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.66 vs. limit=15.0
2024-09-24 01:42:20,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=392354.6666666667, ans=0.0
2024-09-24 01:42:24,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=392354.6666666667, ans=0.125
2024-09-24 01:42:59,349 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.230e+02 1.321e+02 1.423e+02 2.235e+02, threshold=2.642e+02, percent-clipped=0.0
2024-09-24 01:43:23,763 INFO [train.py:1198] (1/4) Epoch 22, batch 2300, loss[loss=0.1714, ctc_loss=0.111, cr_loss=0.3023, over 17034.00 frames. ], tot_loss[loss=0.2124, ctc_loss=0.1409, cr_loss=0.3574, over 3353145.94 frames. ], batch size: 39, lr: 5.44e-03, grad_scale: 16.0
2024-09-24 01:43:42,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=392588.0, ans=0.125
2024-09-24 01:43:53,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.min_positive, batch_count=392588.0, ans=0.05
2024-09-24 01:44:17,805 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=6.85 vs. limit=15.0
2024-09-24 01:44:51,567 INFO [train.py:1198] (1/4) Epoch 22, batch 2350, loss[loss=0.2071, ctc_loss=0.1363, cr_loss=0.3541, over 17219.00 frames. ], tot_loss[loss=0.212, ctc_loss=0.1405, cr_loss=0.3573, over 3358776.55 frames. ], batch size: 47, lr: 5.44e-03, grad_scale: 16.0
2024-09-24 01:45:01,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=392774.6666666667, ans=0.0
2024-09-24 01:45:01,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=392774.6666666667, ans=0.0
2024-09-24 01:45:17,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=392821.3333333333, ans=0.125
2024-09-24 01:45:21,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=392868.0, ans=0.125
2024-09-24 01:45:33,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=392868.0, ans=0.125
2024-09-24 01:45:41,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=392914.6666666667, ans=0.1
2024-09-24 01:45:43,623 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.33 vs. limit=10.0
2024-09-24 01:45:47,291 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.237e+02 1.319e+02 1.405e+02 2.078e+02, threshold=2.638e+02, percent-clipped=0.0
2024-09-24 01:46:02,450 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.18 vs. limit=22.5
2024-09-24 01:46:11,755 INFO [train.py:1198] (1/4) Epoch 22, batch 2400, loss[loss=0.2442, ctc_loss=0.1673, cr_loss=0.3841, over 16608.00 frames. ], tot_loss[loss=0.2128, ctc_loss=0.1412, cr_loss=0.3578, over 3349089.77 frames. ], batch size: 66, lr: 5.43e-03, grad_scale: 32.0
2024-09-24 01:46:15,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=393008.0, ans=0.0
2024-09-24 01:46:23,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=393008.0, ans=0.125
2024-09-24 01:46:29,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=393054.6666666667, ans=0.125
2024-09-24 01:47:19,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.25 vs. limit=15.0
2024-09-24 01:47:22,990 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=7.20 vs. limit=15.0
2024-09-24 01:47:34,554 INFO [train.py:1198] (1/4) Epoch 22, batch 2450, loss[loss=0.1957, ctc_loss=0.1254, cr_loss=0.3515, over 17172.00 frames. ], tot_loss[loss=0.2135, ctc_loss=0.1417, cr_loss=0.359, over 3351286.63 frames. ], batch size: 45, lr: 5.43e-03, grad_scale: 32.0
2024-09-24 01:47:40,420 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.64 vs. limit=15.0
2024-09-24 01:48:03,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=393288.0, ans=0.07
2024-09-24 01:48:03,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=393288.0, ans=0.95
2024-09-24 01:48:30,720 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.273e+02 1.360e+02 1.558e+02 2.301e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-24 01:48:36,120 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.21 vs. limit=15.0
2024-09-24 01:48:52,911 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:48:56,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=393474.6666666667, ans=0.125
2024-09-24 01:48:57,400 INFO [train.py:1198] (1/4) Epoch 22, batch 2500, loss[loss=0.1992, ctc_loss=0.1317, cr_loss=0.3375, over 17198.00 frames. ], tot_loss[loss=0.2126, ctc_loss=0.141, cr_loss=0.358, over 3359795.42 frames. ], batch size: 55, lr: 5.43e-03, grad_scale: 32.0
2024-09-24 01:49:02,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=393474.6666666667, ans=0.0
2024-09-24 01:49:38,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=393568.0, ans=0.0
2024-09-24 01:50:00,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.81 vs. limit=6.0
2024-09-24 01:50:03,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.83 vs. limit=15.0
2024-09-24 01:50:11,046 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:50:22,182 INFO [train.py:1198] (1/4) Epoch 22, batch 2550, loss[loss=0.1997, ctc_loss=0.1314, cr_loss=0.3411, over 17167.00 frames. ], tot_loss[loss=0.2128, ctc_loss=0.1412, cr_loss=0.3581, over 3354725.79 frames. ], batch size: 45, lr: 5.43e-03, grad_scale: 32.0
2024-09-24 01:50:24,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=393708.0, ans=0.125
2024-09-24 01:51:04,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=393801.3333333333, ans=0.125
2024-09-24 01:51:12,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=393848.0, ans=0.0
2024-09-24 01:51:18,854 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.258e+02 1.341e+02 1.475e+02 2.313e+02, threshold=2.682e+02, percent-clipped=0.0
2024-09-24 01:51:23,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=393848.0, ans=0.1
2024-09-24 01:51:43,206 INFO [train.py:1198] (1/4) Epoch 22, batch 2600, loss[loss=0.1889, ctc_loss=0.1245, cr_loss=0.3219, over 16966.00 frames. ], tot_loss[loss=0.2127, ctc_loss=0.1412, cr_loss=0.3572, over 3348778.73 frames. ], batch size: 42, lr: 5.43e-03, grad_scale: 32.0
2024-09-24 01:51:43,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=393941.3333333333, ans=0.125
2024-09-24 01:51:54,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=393941.3333333333, ans=0.2
2024-09-24 01:52:33,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=394081.3333333333, ans=0.05
2024-09-24 01:53:06,119 INFO [train.py:1198] (1/4) Epoch 22, batch 2650, loss[loss=0.211, ctc_loss=0.142, cr_loss=0.3453, over 16915.00 frames. ], tot_loss[loss=0.2131, ctc_loss=0.1415, cr_loss=0.3577, over 3338870.83 frames. ], batch size: 58, lr: 5.43e-03, grad_scale: 32.0
2024-09-24 01:53:06,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=394174.6666666667, ans=0.0
2024-09-24 01:53:09,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=394174.6666666667, ans=0.1
2024-09-24 01:53:17,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=394174.6666666667, ans=0.2
2024-09-24 01:53:19,832 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.45 vs. limit=10.0
2024-09-24 01:53:29,790 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.96 vs. limit=15.0
2024-09-24 01:53:49,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=394268.0, ans=0.125
2024-09-24 01:54:07,322 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.255e+02 1.320e+02 1.434e+02 1.908e+02, threshold=2.640e+02, percent-clipped=0.0
2024-09-24 01:54:31,190 INFO [train.py:1198] (1/4) Epoch 22, batch 2700, loss[loss=0.1827, ctc_loss=0.1176, cr_loss=0.3252, over 17077.00 frames. ], tot_loss[loss=0.2129, ctc_loss=0.1414, cr_loss=0.3578, over 3343453.00 frames. ], batch size: 43, lr: 5.42e-03, grad_scale: 32.0
2024-09-24 01:54:51,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=394454.6666666667, ans=0.0
2024-09-24 01:55:36,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=394594.6666666667, ans=0.125
2024-09-24 01:55:53,387 INFO [train.py:1198] (1/4) Epoch 22, batch 2750, loss[loss=0.2004, ctc_loss=0.1325, cr_loss=0.3399, over 17316.00 frames. ], tot_loss[loss=0.211, ctc_loss=0.14, cr_loss=0.3551, over 3344894.44 frames. ], batch size: 51, lr: 5.42e-03, grad_scale: 32.0
2024-09-24 01:55:55,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=394641.3333333333, ans=0.0
2024-09-24 01:55:59,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=394641.3333333333, ans=0.125
2024-09-24 01:56:25,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=394734.6666666667, ans=0.125
2024-09-24 01:56:31,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=394734.6666666667, ans=0.1
2024-09-24 01:56:43,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=394781.3333333333, ans=0.125
2024-09-24 01:56:52,095 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.297e+02 1.428e+02 1.577e+02 2.482e+02, threshold=2.855e+02, percent-clipped=0.0
2024-09-24 01:56:57,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=394781.3333333333, ans=0.0
2024-09-24 01:56:58,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=394828.0, ans=0.0
2024-09-24 01:57:03,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.52 vs. limit=15.0
2024-09-24 01:57:03,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=394828.0, ans=0.1
2024-09-24 01:57:05,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=394828.0, ans=0.2
2024-09-24 01:57:16,458 INFO [train.py:1198] (1/4) Epoch 22, batch 2800, loss[loss=0.1967, ctc_loss=0.1316, cr_loss=0.3253, over 17210.00 frames. ], tot_loss[loss=0.2104, ctc_loss=0.1394, cr_loss=0.3552, over 3359026.41 frames. ], batch size: 47, lr: 5.42e-03, grad_scale: 32.0
2024-09-24 01:57:30,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=394921.3333333333, ans=0.0
2024-09-24 01:58:10,295 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 01:58:12,522 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=5.05 vs. limit=10.0
2024-09-24 01:58:32,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=395061.3333333333, ans=0.2
2024-09-24 01:58:38,324 INFO [train.py:1198] (1/4) Epoch 22, batch 2850, loss[loss=0.2176, ctc_loss=0.1467, cr_loss=0.3545, over 16711.00 frames. ], tot_loss[loss=0.2095, ctc_loss=0.1387, cr_loss=0.3544, over 3360280.32 frames. ], batch size: 61, lr: 5.42e-03, grad_scale: 32.0
2024-09-24 01:58:40,806 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=8.83 vs. limit=15.0
2024-09-24 01:58:49,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=395108.0, ans=0.05
2024-09-24 01:59:00,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=395154.6666666667, ans=0.0
2024-09-24 01:59:36,607 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.264e+02 1.351e+02 1.437e+02 2.135e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-24 01:59:55,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=395294.6666666667, ans=0.125
2024-09-24 02:00:03,215 INFO [train.py:1198] (1/4) Epoch 22, batch 2900, loss[loss=0.1939, ctc_loss=0.1297, cr_loss=0.3207, over 17301.00 frames. ], tot_loss[loss=0.2093, ctc_loss=0.1385, cr_loss=0.354, over 3363398.51 frames. ], batch size: 49, lr: 5.42e-03, grad_scale: 32.0
2024-09-24 02:00:23,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=395388.0, ans=0.0
2024-09-24 02:00:33,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=395434.6666666667, ans=0.125
2024-09-24 02:01:06,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.20 vs. limit=15.0
2024-09-24 02:01:07,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=395528.0, ans=0.0
2024-09-24 02:01:23,078 INFO [train.py:1198] (1/4) Epoch 22, batch 2950, loss[loss=0.2005, ctc_loss=0.1339, cr_loss=0.3329, over 17230.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1388, cr_loss=0.3549, over 3365652.15 frames. ], batch size: 50, lr: 5.42e-03, grad_scale: 32.0
2024-09-24 02:01:26,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=395574.6666666667, ans=10.0
2024-09-24 02:02:02,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=395668.0, ans=0.025
2024-09-24 02:02:21,884 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.266e+02 1.346e+02 1.444e+02 1.756e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-24 02:02:25,611 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.85 vs. limit=22.5
2024-09-24 02:02:37,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=395761.3333333333, ans=0.0
2024-09-24 02:02:45,309 INFO [train.py:1198] (1/4) Epoch 22, batch 3000, loss[loss=0.2198, ctc_loss=0.1453, cr_loss=0.3726, over 16919.00 frames. ], tot_loss[loss=0.2113, ctc_loss=0.1399, cr_loss=0.3568, over 3361674.62 frames. ], batch size: 58, lr: 5.42e-03, grad_scale: 32.0
2024-09-24 02:02:45,309 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 02:03:00,740 INFO [train.py:1230] (1/4) Epoch 22, validation: loss=0.03869, ctc_loss=0.03869, cr_loss=8.188e-15, over 944034.00 frames. 
2024-09-24 02:03:00,740 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 02:03:25,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=395854.6666666667, ans=0.025
2024-09-24 02:03:38,620 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.79 vs. limit=15.0
2024-09-24 02:03:44,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=395901.3333333333, ans=0.125
2024-09-24 02:03:55,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=395948.0, ans=0.0
2024-09-24 02:04:21,130 INFO [train.py:1198] (1/4) Epoch 22, batch 3050, loss[loss=0.2668, ctc_loss=0.1823, cr_loss=0.4224, over 14970.00 frames. ], tot_loss[loss=0.212, ctc_loss=0.1406, cr_loss=0.3574, over 3354760.83 frames. ], batch size: 89, lr: 5.41e-03, grad_scale: 32.0
2024-09-24 02:04:35,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=396088.0, ans=0.125
2024-09-24 02:04:43,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=396088.0, ans=10.0
2024-09-24 02:04:50,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=396134.6666666667, ans=0.125
2024-09-24 02:04:54,624 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.85 vs. limit=6.0
2024-09-24 02:04:58,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=396134.6666666667, ans=0.025
2024-09-24 02:05:00,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=396134.6666666667, ans=0.125
2024-09-24 02:05:04,154 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.36 vs. limit=15.0
2024-09-24 02:05:09,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=396181.3333333333, ans=0.125
2024-09-24 02:05:15,411 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.242e+02 1.330e+02 1.474e+02 2.506e+02, threshold=2.661e+02, percent-clipped=0.0
2024-09-24 02:05:39,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=396274.6666666667, ans=0.1
2024-09-24 02:05:41,006 INFO [train.py:1198] (1/4) Epoch 22, batch 3100, loss[loss=0.1822, ctc_loss=0.119, cr_loss=0.3161, over 17058.00 frames. ], tot_loss[loss=0.2123, ctc_loss=0.1408, cr_loss=0.3575, over 3347920.80 frames. ], batch size: 39, lr: 5.41e-03, grad_scale: 32.0
2024-09-24 02:05:42,198 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.94 vs. limit=6.0
2024-09-24 02:06:07,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=396321.3333333333, ans=0.125
2024-09-24 02:06:24,232 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=12.56 vs. limit=15.0
2024-09-24 02:06:26,739 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.62 vs. limit=15.0
2024-09-24 02:06:31,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=396414.6666666667, ans=0.0
2024-09-24 02:06:50,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.67 vs. limit=12.0
2024-09-24 02:06:58,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=396461.3333333333, ans=0.05
2024-09-24 02:07:01,231 INFO [train.py:1198] (1/4) Epoch 22, batch 3150, loss[loss=0.2339, ctc_loss=0.1564, cr_loss=0.3873, over 17205.00 frames. ], tot_loss[loss=0.2117, ctc_loss=0.1403, cr_loss=0.3569, over 3358206.93 frames. ], batch size: 47, lr: 5.41e-03, grad_scale: 16.0
2024-09-24 02:07:04,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=396508.0, ans=0.125
2024-09-24 02:07:16,060 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.32 vs. limit=12.0
2024-09-24 02:07:48,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=396648.0, ans=0.125
2024-09-24 02:07:50,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=396648.0, ans=0.125
2024-09-24 02:07:57,788 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.290e+02 1.399e+02 1.555e+02 2.844e+02, threshold=2.797e+02, percent-clipped=1.0
2024-09-24 02:08:13,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=396694.6666666667, ans=0.125
2024-09-24 02:08:18,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=396741.3333333333, ans=0.125
2024-09-24 02:08:19,586 INFO [train.py:1198] (1/4) Epoch 22, batch 3200, loss[loss=0.2117, ctc_loss=0.1418, cr_loss=0.3496, over 17364.00 frames. ], tot_loss[loss=0.2123, ctc_loss=0.1408, cr_loss=0.3577, over 3349195.54 frames. ], batch size: 48, lr: 5.41e-03, grad_scale: 32.0
2024-09-24 02:08:30,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=396741.3333333333, ans=0.025
2024-09-24 02:08:46,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=396788.0, ans=0.0
2024-09-24 02:09:13,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=396881.3333333333, ans=0.1
2024-09-24 02:09:19,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=396881.3333333333, ans=0.125
2024-09-24 02:09:36,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=396974.6666666667, ans=0.125
2024-09-24 02:09:37,797 INFO [train.py:1198] (1/4) Epoch 22, batch 3250, loss[loss=0.2442, ctc_loss=0.1625, cr_loss=0.4087, over 17051.00 frames. ], tot_loss[loss=0.2123, ctc_loss=0.1407, cr_loss=0.3583, over 3357442.75 frames. ], batch size: 52, lr: 5.41e-03, grad_scale: 32.0
2024-09-24 02:09:41,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=396974.6666666667, ans=0.025
2024-09-24 02:09:42,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=396974.6666666667, ans=0.0
2024-09-24 02:09:48,069 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.18 vs. limit=15.0
2024-09-24 02:10:20,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=397068.0, ans=0.0
2024-09-24 02:10:33,680 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.276e+02 1.353e+02 1.572e+02 3.957e+02, threshold=2.706e+02, percent-clipped=1.0
2024-09-24 02:10:44,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=397161.3333333333, ans=0.0
2024-09-24 02:10:49,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=397161.3333333333, ans=0.125
2024-09-24 02:10:49,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=397161.3333333333, ans=0.07
2024-09-24 02:10:55,243 INFO [train.py:1198] (1/4) Epoch 22, batch 3300, loss[loss=0.1749, ctc_loss=0.1112, cr_loss=0.3188, over 16951.00 frames. ], tot_loss[loss=0.2107, ctc_loss=0.1395, cr_loss=0.3562, over 3367245.98 frames. ], batch size: 42, lr: 5.41e-03, grad_scale: 32.0
2024-09-24 02:11:22,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=397254.6666666667, ans=0.125
2024-09-24 02:11:47,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=397348.0, ans=0.0
2024-09-24 02:12:07,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=397394.6666666667, ans=0.1
2024-09-24 02:12:10,157 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=6.37 vs. limit=15.0
2024-09-24 02:12:15,301 INFO [train.py:1198] (1/4) Epoch 22, batch 3350, loss[loss=0.1958, ctc_loss=0.1299, cr_loss=0.3293, over 17012.00 frames. ], tot_loss[loss=0.2087, ctc_loss=0.138, cr_loss=0.3531, over 3363522.93 frames. ], batch size: 44, lr: 5.40e-03, grad_scale: 32.0
2024-09-24 02:12:18,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.27 vs. limit=10.0
2024-09-24 02:12:20,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=397441.3333333333, ans=0.1
2024-09-24 02:12:22,632 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.15 vs. limit=15.0
2024-09-24 02:12:22,925 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=7.72 vs. limit=15.0
2024-09-24 02:12:42,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=397488.0, ans=0.125
2024-09-24 02:12:54,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=397534.6666666667, ans=0.125
2024-09-24 02:13:00,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=397581.3333333333, ans=0.2
2024-09-24 02:13:05,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=397581.3333333333, ans=0.0
2024-09-24 02:13:11,463 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.283e+02 1.436e+02 1.658e+02 2.229e+02, threshold=2.872e+02, percent-clipped=0.0
2024-09-24 02:13:18,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=397628.0, ans=6.0
2024-09-24 02:13:33,123 INFO [train.py:1198] (1/4) Epoch 22, batch 3400, loss[loss=0.2032, ctc_loss=0.1319, cr_loss=0.3564, over 17265.00 frames. ], tot_loss[loss=0.2095, ctc_loss=0.1386, cr_loss=0.3543, over 3364914.12 frames. ], batch size: 44, lr: 5.40e-03, grad_scale: 32.0
2024-09-24 02:13:47,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=397721.3333333333, ans=0.07
2024-09-24 02:13:58,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=397721.3333333333, ans=0.125
2024-09-24 02:14:01,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=397721.3333333333, ans=0.125
2024-09-24 02:14:03,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=397768.0, ans=0.125
2024-09-24 02:14:03,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=397768.0, ans=0.07
2024-09-24 02:14:15,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=397768.0, ans=0.125
2024-09-24 02:14:23,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.43 vs. limit=15.0
2024-09-24 02:14:39,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=397861.3333333333, ans=0.1
2024-09-24 02:14:42,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=397861.3333333333, ans=0.125
2024-09-24 02:14:53,441 INFO [train.py:1198] (1/4) Epoch 22, batch 3450, loss[loss=0.2008, ctc_loss=0.131, cr_loss=0.349, over 17265.00 frames. ], tot_loss[loss=0.2107, ctc_loss=0.1396, cr_loss=0.3555, over 3352920.73 frames. ], batch size: 42, lr: 5.40e-03, grad_scale: 32.0
2024-09-24 02:14:54,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.30 vs. limit=15.0
2024-09-24 02:14:55,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=397908.0, ans=0.125
2024-09-24 02:15:21,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=397954.6666666667, ans=0.0
2024-09-24 02:15:25,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=398001.3333333333, ans=0.125
2024-09-24 02:15:32,102 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.37 vs. limit=12.0
2024-09-24 02:15:51,969 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.275e+02 1.378e+02 1.514e+02 2.011e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-24 02:16:07,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=398094.6666666667, ans=0.125
2024-09-24 02:16:13,998 INFO [train.py:1198] (1/4) Epoch 22, batch 3500, loss[loss=0.2208, ctc_loss=0.1433, cr_loss=0.3877, over 17153.00 frames. ], tot_loss[loss=0.2104, ctc_loss=0.1392, cr_loss=0.3557, over 3358815.77 frames. ], batch size: 45, lr: 5.40e-03, grad_scale: 32.0
2024-09-24 02:16:16,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=398141.3333333333, ans=0.125
2024-09-24 02:16:21,356 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.93 vs. limit=22.5
2024-09-24 02:16:23,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=398141.3333333333, ans=0.125
2024-09-24 02:16:27,013 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 02:16:38,559 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.04 vs. limit=10.0
2024-09-24 02:16:43,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=398234.6666666667, ans=0.0
2024-09-24 02:16:54,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=398234.6666666667, ans=0.125
2024-09-24 02:16:54,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=398234.6666666667, ans=0.0
2024-09-24 02:17:14,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=398281.3333333333, ans=0.2
2024-09-24 02:17:25,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=398328.0, ans=0.125
2024-09-24 02:17:32,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=398374.6666666667, ans=0.07
2024-09-24 02:17:34,216 INFO [train.py:1198] (1/4) Epoch 22, batch 3550, loss[loss=0.2673, ctc_loss=0.1809, cr_loss=0.432, over 16494.00 frames. ], tot_loss[loss=0.211, ctc_loss=0.1398, cr_loss=0.3563, over 3350603.84 frames. ], batch size: 66, lr: 5.40e-03, grad_scale: 32.0
2024-09-24 02:17:34,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=398374.6666666667, ans=0.0
2024-09-24 02:17:37,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=398374.6666666667, ans=0.1
2024-09-24 02:17:56,212 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 02:17:56,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=398421.3333333333, ans=0.07
2024-09-24 02:18:07,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=398468.0, ans=0.1
2024-09-24 02:18:24,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=398514.6666666667, ans=0.0
2024-09-24 02:18:32,117 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.257e+02 1.363e+02 1.494e+02 1.950e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-24 02:18:44,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=398561.3333333333, ans=0.125
2024-09-24 02:18:46,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=398561.3333333333, ans=0.125
2024-09-24 02:18:52,335 INFO [train.py:1198] (1/4) Epoch 22, batch 3600, loss[loss=0.2182, ctc_loss=0.1447, cr_loss=0.367, over 17109.00 frames. ], tot_loss[loss=0.2118, ctc_loss=0.1404, cr_loss=0.3571, over 3337189.18 frames. ], batch size: 49, lr: 5.40e-03, grad_scale: 32.0
2024-09-24 02:19:37,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=398748.0, ans=0.2
2024-09-24 02:19:39,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=398748.0, ans=0.125
2024-09-24 02:19:41,543 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.11 vs. limit=12.0
2024-09-24 02:20:01,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=398794.6666666667, ans=0.05
2024-09-24 02:20:10,105 INFO [train.py:1198] (1/4) Epoch 22, batch 3650, loss[loss=0.1979, ctc_loss=0.1302, cr_loss=0.3384, over 17216.00 frames. ], tot_loss[loss=0.2123, ctc_loss=0.1407, cr_loss=0.3577, over 3340115.15 frames. ], batch size: 50, lr: 5.39e-03, grad_scale: 32.0
2024-09-24 02:20:13,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=398841.3333333333, ans=0.09899494936611666
2024-09-24 02:20:16,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=398841.3333333333, ans=0.1
2024-09-24 02:20:21,701 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=10.95 vs. limit=12.0
2024-09-24 02:20:27,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=398888.0, ans=0.1
2024-09-24 02:20:38,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=398888.0, ans=0.125
2024-09-24 02:20:40,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=398888.0, ans=0.125
2024-09-24 02:21:06,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=398981.3333333333, ans=0.0
2024-09-24 02:21:09,556 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.259e+02 1.359e+02 1.456e+02 2.035e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-24 02:21:24,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=399028.0, ans=0.0
2024-09-24 02:21:30,491 INFO [train.py:1198] (1/4) Epoch 22, batch 3700, loss[loss=0.2431, ctc_loss=0.1696, cr_loss=0.3678, over 12609.00 frames. ], tot_loss[loss=0.2126, ctc_loss=0.141, cr_loss=0.358, over 3345162.44 frames. ], batch size: 123, lr: 5.39e-03, grad_scale: 32.0
2024-09-24 02:22:15,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=399214.6666666667, ans=0.1
2024-09-24 02:22:34,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=399261.3333333333, ans=0.0
2024-09-24 02:22:37,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=399261.3333333333, ans=0.025
2024-09-24 02:22:48,075 INFO [train.py:1198] (1/4) Epoch 22, batch 3750, loss[loss=0.2272, ctc_loss=0.1557, cr_loss=0.3573, over 15962.00 frames. ], tot_loss[loss=0.2131, ctc_loss=0.1414, cr_loss=0.3581, over 3342703.02 frames. ], batch size: 74, lr: 5.39e-03, grad_scale: 32.0
2024-09-24 02:22:59,462 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.70 vs. limit=15.0
2024-09-24 02:23:40,485 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.02 vs. limit=22.5
2024-09-24 02:23:46,154 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.265e+02 1.370e+02 1.488e+02 1.870e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-24 02:23:48,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=399448.0, ans=0.1
2024-09-24 02:24:07,239 INFO [train.py:1198] (1/4) Epoch 22, batch 3800, loss[loss=0.2592, ctc_loss=0.1767, cr_loss=0.4123, over 14989.00 frames. ], tot_loss[loss=0.2134, ctc_loss=0.1418, cr_loss=0.3584, over 3327651.95 frames. ], batch size: 89, lr: 5.39e-03, grad_scale: 32.0
2024-09-24 02:24:19,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=399541.3333333333, ans=0.125
2024-09-24 02:24:38,298 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.65 vs. limit=15.0
2024-09-24 02:25:03,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.31 vs. limit=15.0
2024-09-24 02:25:06,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=399728.0, ans=0.07
2024-09-24 02:25:15,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=399728.0, ans=0.125
2024-09-24 02:25:18,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=399728.0, ans=0.05
2024-09-24 02:25:22,548 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 02:25:23,718 INFO [train.py:1198] (1/4) Epoch 22, batch 3850, loss[loss=0.2366, ctc_loss=0.1564, cr_loss=0.4011, over 16852.00 frames. ], tot_loss[loss=0.2148, ctc_loss=0.143, cr_loss=0.3591, over 3298500.59 frames. ], batch size: 58, lr: 5.39e-03, grad_scale: 16.0
2024-09-24 02:25:50,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=399821.3333333333, ans=0.125
2024-09-24 02:26:10,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=399914.6666666667, ans=0.125
2024-09-24 02:26:22,073 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.394e+02 1.521e+02 1.653e+02 2.855e+02, threshold=3.042e+02, percent-clipped=1.0
2024-09-24 02:26:29,158 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.61 vs. limit=15.0
2024-09-24 02:27:26,201 INFO [train.py:1198] (1/4) Epoch 23, batch 0, loss[loss=0.2382, ctc_loss=0.1577, cr_loss=0.4024, over 16653.00 frames. ], tot_loss[loss=0.2382, ctc_loss=0.1577, cr_loss=0.4024, over 16653.00 frames. ], batch size: 61, lr: 5.27e-03, grad_scale: 32.0
2024-09-24 02:27:26,201 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 02:27:39,649 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.5.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([2.2985, 3.4295, 2.6915, 3.2865], device='cuda:1')
2024-09-24 02:27:41,789 INFO [train.py:1230] (1/4) Epoch 23, validation: loss=0.03754, ctc_loss=0.03754, cr_loss=8.311e-15, over 944034.00 frames. 
2024-09-24 02:27:41,789 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 02:28:11,779 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.05 vs. limit=12.0
2024-09-24 02:28:33,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.43 vs. limit=15.0
2024-09-24 02:28:33,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.84 vs. limit=6.0
2024-09-24 02:28:55,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=400180.6666666667, ans=0.125
2024-09-24 02:28:55,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=400180.6666666667, ans=0.1
2024-09-24 02:29:04,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=400227.3333333333, ans=0.125
2024-09-24 02:29:06,169 INFO [train.py:1198] (1/4) Epoch 23, batch 50, loss[loss=0.2187, ctc_loss=0.1469, cr_loss=0.3588, over 16053.00 frames. ], tot_loss[loss=0.2122, ctc_loss=0.1406, cr_loss=0.3578, over 752273.47 frames. ], batch size: 74, lr: 5.26e-03, grad_scale: 32.0
2024-09-24 02:29:46,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=400320.6666666667, ans=0.0
2024-09-24 02:29:59,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=400367.3333333333, ans=0.125
2024-09-24 02:30:00,829 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 02:30:01,082 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.41 vs. limit=10.0
2024-09-24 02:30:08,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=400414.0, ans=0.1
2024-09-24 02:30:08,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=400414.0, ans=0.125
2024-09-24 02:30:11,720 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.262e+02 1.335e+02 1.476e+02 2.366e+02, threshold=2.670e+02, percent-clipped=0.0
2024-09-24 02:30:26,017 INFO [train.py:1198] (1/4) Epoch 23, batch 100, loss[loss=0.2106, ctc_loss=0.1389, cr_loss=0.3588, over 17265.00 frames. ], tot_loss[loss=0.2077, ctc_loss=0.1372, cr_loss=0.3525, over 1331459.13 frames. ], batch size: 42, lr: 5.26e-03, grad_scale: 32.0
2024-09-24 02:31:18,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=400600.6666666667, ans=0.125
2024-09-24 02:31:33,722 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.12 vs. limit=15.0
2024-09-24 02:31:50,459 INFO [train.py:1198] (1/4) Epoch 23, batch 150, loss[loss=0.215, ctc_loss=0.1436, cr_loss=0.3569, over 17359.00 frames. ], tot_loss[loss=0.2081, ctc_loss=0.1375, cr_loss=0.3531, over 1784287.70 frames. ], batch size: 48, lr: 5.26e-03, grad_scale: 32.0
2024-09-24 02:31:51,278 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.64 vs. limit=15.0
2024-09-24 02:32:14,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=400740.6666666667, ans=0.0
2024-09-24 02:32:22,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=400787.3333333333, ans=0.125
2024-09-24 02:32:25,179 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.32 vs. limit=15.0
2024-09-24 02:32:34,156 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.73 vs. limit=10.0
2024-09-24 02:32:41,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=400834.0, ans=0.0
2024-09-24 02:32:53,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=400880.6666666667, ans=0.0
2024-09-24 02:32:53,215 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 02:32:55,955 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.238e+02 1.330e+02 1.442e+02 1.852e+02, threshold=2.660e+02, percent-clipped=0.0
2024-09-24 02:32:57,028 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.20 vs. limit=15.0
2024-09-24 02:33:13,124 INFO [train.py:1198] (1/4) Epoch 23, batch 200, loss[loss=0.1699, ctc_loss=0.1102, cr_loss=0.2988, over 16732.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.1374, cr_loss=0.3528, over 2135701.99 frames. ], batch size: 37, lr: 5.26e-03, grad_scale: 32.0
2024-09-24 02:33:19,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=400927.3333333333, ans=0.0
2024-09-24 02:33:22,985 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 02:33:30,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=400974.0, ans=0.125
2024-09-24 02:34:01,707 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.87 vs. limit=15.0
2024-09-24 02:34:05,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=401067.3333333333, ans=0.025
2024-09-24 02:34:07,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=401067.3333333333, ans=0.125
2024-09-24 02:34:18,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=401114.0, ans=0.125
2024-09-24 02:34:35,729 INFO [train.py:1198] (1/4) Epoch 23, batch 250, loss[loss=0.2186, ctc_loss=0.1473, cr_loss=0.3564, over 17021.00 frames. ], tot_loss[loss=0.2092, ctc_loss=0.1385, cr_loss=0.3538, over 2406391.99 frames. ], batch size: 53, lr: 5.26e-03, grad_scale: 32.0
2024-09-24 02:34:48,900 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 02:35:15,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=401254.0, ans=10.0
2024-09-24 02:35:30,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=401300.6666666667, ans=0.125
2024-09-24 02:35:41,288 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.253e+02 1.336e+02 1.468e+02 3.243e+02, threshold=2.673e+02, percent-clipped=1.0
2024-09-24 02:35:41,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=401347.3333333333, ans=0.125
2024-09-24 02:35:44,196 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=9.63 vs. limit=15.0
2024-09-24 02:35:46,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=401347.3333333333, ans=0.125
2024-09-24 02:35:55,852 INFO [train.py:1198] (1/4) Epoch 23, batch 300, loss[loss=0.2394, ctc_loss=0.1634, cr_loss=0.3803, over 17068.00 frames. ], tot_loss[loss=0.2112, ctc_loss=0.1399, cr_loss=0.3565, over 2619348.29 frames. ], batch size: 46, lr: 5.26e-03, grad_scale: 32.0
2024-09-24 02:35:57,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=401394.0, ans=0.0
2024-09-24 02:36:05,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=401394.0, ans=0.1
2024-09-24 02:36:16,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=401440.6666666667, ans=0.0
2024-09-24 02:36:31,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=401487.3333333333, ans=0.2
2024-09-24 02:36:34,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=401487.3333333333, ans=0.05
2024-09-24 02:36:56,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=401534.0, ans=0.125
2024-09-24 02:37:19,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=401627.3333333333, ans=0.05
2024-09-24 02:37:20,386 INFO [train.py:1198] (1/4) Epoch 23, batch 350, loss[loss=0.1885, ctc_loss=0.1236, cr_loss=0.3243, over 17163.00 frames. ], tot_loss[loss=0.2107, ctc_loss=0.1394, cr_loss=0.3562, over 2790532.66 frames. ], batch size: 45, lr: 5.26e-03, grad_scale: 32.0
2024-09-24 02:37:45,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=401674.0, ans=0.0
2024-09-24 02:38:25,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=401814.0, ans=0.2
2024-09-24 02:38:28,350 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.260e+02 1.355e+02 1.507e+02 3.262e+02, threshold=2.709e+02, percent-clipped=1.0
2024-09-24 02:38:42,870 INFO [train.py:1198] (1/4) Epoch 23, batch 400, loss[loss=0.188, ctc_loss=0.1238, cr_loss=0.3209, over 17264.00 frames. ], tot_loss[loss=0.2103, ctc_loss=0.1391, cr_loss=0.356, over 2913411.47 frames. ], batch size: 42, lr: 5.25e-03, grad_scale: 32.0
2024-09-24 02:38:55,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=401860.6666666667, ans=0.0
2024-09-24 02:39:03,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=401907.3333333333, ans=0.0
2024-09-24 02:39:57,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=402047.3333333333, ans=0.0
2024-09-24 02:39:59,639 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.87 vs. limit=10.0
2024-09-24 02:40:04,967 INFO [train.py:1198] (1/4) Epoch 23, batch 450, loss[loss=0.2082, ctc_loss=0.1352, cr_loss=0.3649, over 17222.00 frames. ], tot_loss[loss=0.2111, ctc_loss=0.1398, cr_loss=0.3566, over 2999484.92 frames. ], batch size: 47, lr: 5.25e-03, grad_scale: 32.0
2024-09-24 02:40:08,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=402094.0, ans=0.0
2024-09-24 02:40:46,686 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=402187.3333333333, ans=0.125
2024-09-24 02:41:07,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=402234.0, ans=0.125
2024-09-24 02:41:08,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=402234.0, ans=0.125
2024-09-24 02:41:13,370 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.257e+02 1.349e+02 1.505e+02 2.195e+02, threshold=2.697e+02, percent-clipped=0.0
2024-09-24 02:41:17,723 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.27 vs. limit=15.0
2024-09-24 02:41:21,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=402280.6666666667, ans=0.0
2024-09-24 02:41:26,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=402327.3333333333, ans=0.07
2024-09-24 02:41:27,648 INFO [train.py:1198] (1/4) Epoch 23, batch 500, loss[loss=0.2072, ctc_loss=0.1389, cr_loss=0.3417, over 17195.00 frames. ], tot_loss[loss=0.2108, ctc_loss=0.1396, cr_loss=0.3558, over 3086594.63 frames. ], batch size: 47, lr: 5.25e-03, grad_scale: 32.0
2024-09-24 02:41:40,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=402327.3333333333, ans=0.125
2024-09-24 02:41:53,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=402374.0, ans=0.2
2024-09-24 02:42:24,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=402467.3333333333, ans=0.2
2024-09-24 02:42:36,819 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.47 vs. limit=15.0
2024-09-24 02:42:41,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=402514.0, ans=0.125
2024-09-24 02:42:50,030 INFO [train.py:1198] (1/4) Epoch 23, batch 550, loss[loss=0.1974, ctc_loss=0.1266, cr_loss=0.3544, over 16954.00 frames. ], tot_loss[loss=0.2112, ctc_loss=0.14, cr_loss=0.3562, over 3140862.68 frames. ], batch size: 42, lr: 5.25e-03, grad_scale: 32.0
2024-09-24 02:42:58,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=402560.6666666667, ans=0.125
2024-09-24 02:43:23,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=402654.0, ans=0.0
2024-09-24 02:43:23,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=402654.0, ans=0.125
2024-09-24 02:43:37,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=402654.0, ans=0.125
2024-09-24 02:43:39,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=402700.6666666667, ans=0.1
2024-09-24 02:43:57,895 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.240e+02 1.347e+02 1.438e+02 1.839e+02, threshold=2.693e+02, percent-clipped=0.0
2024-09-24 02:44:12,301 INFO [train.py:1198] (1/4) Epoch 23, batch 600, loss[loss=0.185, ctc_loss=0.1193, cr_loss=0.3284, over 17092.00 frames. ], tot_loss[loss=0.2112, ctc_loss=0.1398, cr_loss=0.3567, over 3187465.42 frames. ], batch size: 43, lr: 5.25e-03, grad_scale: 32.0
2024-09-24 02:44:26,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=402840.6666666667, ans=0.125
2024-09-24 02:45:20,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=402980.6666666667, ans=0.0
2024-09-24 02:45:32,502 INFO [train.py:1198] (1/4) Epoch 23, batch 650, loss[loss=0.2459, ctc_loss=0.1636, cr_loss=0.4113, over 17096.00 frames. ], tot_loss[loss=0.2103, ctc_loss=0.1393, cr_loss=0.3555, over 3225443.86 frames. ], batch size: 49, lr: 5.25e-03, grad_scale: 32.0
2024-09-24 02:45:50,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=403074.0, ans=0.125
2024-09-24 02:45:57,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.81 vs. limit=10.0
2024-09-24 02:46:15,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=403120.6666666667, ans=0.0
2024-09-24 02:46:16,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=403120.6666666667, ans=0.125
2024-09-24 02:46:43,831 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.212e+02 1.314e+02 1.389e+02 1.753e+02, threshold=2.627e+02, percent-clipped=0.0
2024-09-24 02:46:57,957 INFO [train.py:1198] (1/4) Epoch 23, batch 700, loss[loss=0.1981, ctc_loss=0.1268, cr_loss=0.3566, over 17030.00 frames. ], tot_loss[loss=0.2101, ctc_loss=0.139, cr_loss=0.3553, over 3252431.44 frames. ], batch size: 44, lr: 5.24e-03, grad_scale: 32.0
2024-09-24 02:47:07,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=403260.6666666667, ans=0.0
2024-09-24 02:47:14,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=403307.3333333333, ans=0.125
2024-09-24 02:47:19,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=403307.3333333333, ans=0.125
2024-09-24 02:47:27,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=403307.3333333333, ans=0.0
2024-09-24 02:47:49,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.01 vs. limit=15.0
2024-09-24 02:48:00,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=403400.6666666667, ans=0.0
2024-09-24 02:48:01,187 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.29 vs. limit=6.0
2024-09-24 02:48:20,957 INFO [train.py:1198] (1/4) Epoch 23, batch 750, loss[loss=0.2278, ctc_loss=0.1503, cr_loss=0.3874, over 17039.00 frames. ], tot_loss[loss=0.2102, ctc_loss=0.1391, cr_loss=0.3555, over 3281723.83 frames. ], batch size: 52, lr: 5.24e-03, grad_scale: 32.0
2024-09-24 02:48:24,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.max_abs, batch_count=403494.0, ans=10.0
2024-09-24 02:48:29,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=403494.0, ans=0.0
2024-09-24 02:48:34,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=403494.0, ans=0.5
2024-09-24 02:48:47,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=403540.6666666667, ans=0.0
2024-09-24 02:48:49,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=403540.6666666667, ans=0.2
2024-09-24 02:49:08,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=403587.3333333333, ans=0.125
2024-09-24 02:49:14,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=403634.0, ans=0.125
2024-09-24 02:49:30,261 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.305e+02 1.385e+02 1.543e+02 2.308e+02, threshold=2.771e+02, percent-clipped=0.0
2024-09-24 02:49:43,266 INFO [train.py:1198] (1/4) Epoch 23, batch 800, loss[loss=0.1996, ctc_loss=0.1295, cr_loss=0.3508, over 17214.00 frames. ], tot_loss[loss=0.2108, ctc_loss=0.1396, cr_loss=0.3564, over 3294750.81 frames. ], batch size: 47, lr: 5.24e-03, grad_scale: 32.0
2024-09-24 02:50:24,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=403820.6666666667, ans=0.0
2024-09-24 02:50:27,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=403820.6666666667, ans=0.125
2024-09-24 02:50:49,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=403914.0, ans=0.125
2024-09-24 02:51:08,893 INFO [train.py:1198] (1/4) Epoch 23, batch 850, loss[loss=0.2168, ctc_loss=0.1454, cr_loss=0.3571, over 17025.00 frames. ], tot_loss[loss=0.2108, ctc_loss=0.1395, cr_loss=0.3565, over 3314365.05 frames. ], batch size: 56, lr: 5.24e-03, grad_scale: 16.0
2024-09-24 02:51:10,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=403960.6666666667, ans=0.125
2024-09-24 02:51:52,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=404054.0, ans=0.1
2024-09-24 02:52:00,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=404100.6666666667, ans=0.2
2024-09-24 02:52:17,547 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.256e+02 1.343e+02 1.446e+02 2.174e+02, threshold=2.686e+02, percent-clipped=0.0
2024-09-24 02:52:22,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=404147.3333333333, ans=0.2
2024-09-24 02:52:24,759 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.74 vs. limit=15.0
2024-09-24 02:52:24,768 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=7.30 vs. limit=15.0
2024-09-24 02:52:28,723 INFO [train.py:1198] (1/4) Epoch 23, batch 900, loss[loss=0.1756, ctc_loss=0.114, cr_loss=0.3078, over 16961.00 frames. ], tot_loss[loss=0.2112, ctc_loss=0.1398, cr_loss=0.3572, over 3328050.05 frames. ], batch size: 42, lr: 5.24e-03, grad_scale: 16.0
2024-09-24 02:52:38,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=404194.0, ans=0.025
2024-09-24 02:52:40,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.53 vs. limit=12.0
2024-09-24 02:53:05,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=404287.3333333333, ans=0.0
2024-09-24 02:53:24,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=404334.0, ans=0.125
2024-09-24 02:53:39,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=404380.6666666667, ans=0.125
2024-09-24 02:53:53,644 INFO [train.py:1198] (1/4) Epoch 23, batch 950, loss[loss=0.1888, ctc_loss=0.1244, cr_loss=0.3218, over 17081.00 frames. ], tot_loss[loss=0.2108, ctc_loss=0.1394, cr_loss=0.357, over 3341298.73 frames. ], batch size: 43, lr: 5.24e-03, grad_scale: 16.0
2024-09-24 02:54:29,772 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.38 vs. limit=15.0
2024-09-24 02:54:39,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=404520.6666666667, ans=0.125
2024-09-24 02:54:53,993 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.41 vs. limit=22.5
2024-09-24 02:55:02,711 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.281e+02 1.400e+02 1.572e+02 2.113e+02, threshold=2.800e+02, percent-clipped=0.0
2024-09-24 02:55:13,696 INFO [train.py:1198] (1/4) Epoch 23, batch 1000, loss[loss=0.1773, ctc_loss=0.1144, cr_loss=0.3143, over 17368.00 frames. ], tot_loss[loss=0.2106, ctc_loss=0.1392, cr_loss=0.3572, over 3346877.92 frames. ], batch size: 48, lr: 5.24e-03, grad_scale: 16.0
2024-09-24 02:55:13,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=404660.6666666667, ans=0.125
2024-09-24 02:55:33,938 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.45 vs. limit=15.0
2024-09-24 02:56:09,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=404800.6666666667, ans=0.09899494936611666
2024-09-24 02:56:36,219 INFO [train.py:1198] (1/4) Epoch 23, batch 1050, loss[loss=0.1989, ctc_loss=0.1298, cr_loss=0.3459, over 17005.00 frames. ], tot_loss[loss=0.2106, ctc_loss=0.1393, cr_loss=0.3569, over 3340687.05 frames. ], batch size: 51, lr: 5.23e-03, grad_scale: 16.0
2024-09-24 02:56:41,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=404894.0, ans=0.125
2024-09-24 02:57:01,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=404940.6666666667, ans=0.0
2024-09-24 02:57:09,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=404987.3333333333, ans=0.125
2024-09-24 02:57:19,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=404987.3333333333, ans=0.125
2024-09-24 02:57:22,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=405034.0, ans=0.0
2024-09-24 02:57:35,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.27 vs. limit=6.0
2024-09-24 02:57:46,811 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.300e+02 1.378e+02 1.529e+02 2.270e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-24 02:57:51,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=405080.6666666667, ans=0.0
2024-09-24 02:57:57,824 INFO [train.py:1198] (1/4) Epoch 23, batch 1100, loss[loss=0.2077, ctc_loss=0.1374, cr_loss=0.3515, over 17212.00 frames. ], tot_loss[loss=0.2101, ctc_loss=0.1388, cr_loss=0.3566, over 3352539.03 frames. ], batch size: 47, lr: 5.23e-03, grad_scale: 16.0
2024-09-24 02:58:32,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=405220.6666666667, ans=0.0
2024-09-24 02:58:38,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=405220.6666666667, ans=0.125
2024-09-24 02:58:45,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=405220.6666666667, ans=0.2
2024-09-24 02:58:56,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=405267.3333333333, ans=0.0
2024-09-24 02:59:06,647 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.41 vs. limit=15.0
2024-09-24 02:59:07,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=405314.0, ans=0.1
2024-09-24 02:59:20,147 INFO [train.py:1198] (1/4) Epoch 23, batch 1150, loss[loss=0.2233, ctc_loss=0.1494, cr_loss=0.3696, over 17304.00 frames. ], tot_loss[loss=0.2099, ctc_loss=0.1387, cr_loss=0.3565, over 3357405.88 frames. ], batch size: 51, lr: 5.23e-03, grad_scale: 16.0
2024-09-24 02:59:25,551 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.01 vs. limit=6.0
2024-09-24 02:59:37,308 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.72 vs. limit=12.0
2024-09-24 02:59:38,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=405407.3333333333, ans=0.125
2024-09-24 02:59:50,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=405454.0, ans=0.125
2024-09-24 02:59:54,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=405454.0, ans=0.0
2024-09-24 02:59:55,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=405454.0, ans=0.125
2024-09-24 03:00:05,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=405454.0, ans=0.125
2024-09-24 03:00:29,098 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.259e+02 1.339e+02 1.439e+02 1.652e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-24 03:00:40,374 INFO [train.py:1198] (1/4) Epoch 23, batch 1200, loss[loss=0.1852, ctc_loss=0.1203, cr_loss=0.3245, over 17359.00 frames. ], tot_loss[loss=0.2108, ctc_loss=0.1394, cr_loss=0.3571, over 3347433.16 frames. ], batch size: 48, lr: 5.23e-03, grad_scale: 32.0
2024-09-24 03:01:05,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=405640.6666666667, ans=0.125
2024-09-24 03:01:19,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=405687.3333333333, ans=0.0
2024-09-24 03:01:20,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=405687.3333333333, ans=0.125
2024-09-24 03:01:32,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=405734.0, ans=0.0
2024-09-24 03:01:54,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=405780.6666666667, ans=0.0
2024-09-24 03:02:05,557 INFO [train.py:1198] (1/4) Epoch 23, batch 1250, loss[loss=0.2314, ctc_loss=0.1539, cr_loss=0.3877, over 17044.00 frames. ], tot_loss[loss=0.2104, ctc_loss=0.1392, cr_loss=0.3564, over 3351801.65 frames. ], batch size: 56, lr: 5.23e-03, grad_scale: 32.0
2024-09-24 03:02:05,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=405827.3333333333, ans=0.125
2024-09-24 03:02:09,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=405827.3333333333, ans=0.5
2024-09-24 03:02:12,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=405827.3333333333, ans=0.125
2024-09-24 03:02:17,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=405827.3333333333, ans=0.125
2024-09-24 03:02:18,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=405827.3333333333, ans=0.1
2024-09-24 03:02:39,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=405920.6666666667, ans=0.125
2024-09-24 03:03:18,868 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.257e+02 1.354e+02 1.461e+02 2.818e+02, threshold=2.708e+02, percent-clipped=1.0
2024-09-24 03:03:29,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=406060.6666666667, ans=0.125
2024-09-24 03:03:30,974 INFO [train.py:1198] (1/4) Epoch 23, batch 1300, loss[loss=0.1741, ctc_loss=0.113, cr_loss=0.3055, over 17081.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1386, cr_loss=0.3548, over 3359057.04 frames. ], batch size: 46, lr: 5.23e-03, grad_scale: 16.0
2024-09-24 03:03:42,553 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.31 vs. limit=22.5
2024-09-24 03:03:43,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.75 vs. limit=22.5
2024-09-24 03:04:04,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=406154.0, ans=0.2
2024-09-24 03:04:06,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=406154.0, ans=0.0
2024-09-24 03:04:30,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=406200.6666666667, ans=0.2
2024-09-24 03:04:35,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=406247.3333333333, ans=0.125
2024-09-24 03:04:36,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=406247.3333333333, ans=0.09899494936611666
2024-09-24 03:04:44,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=406247.3333333333, ans=0.0
2024-09-24 03:04:50,622 INFO [train.py:1198] (1/4) Epoch 23, batch 1350, loss[loss=0.2116, ctc_loss=0.1409, cr_loss=0.3536, over 17308.00 frames. ], tot_loss[loss=0.2091, ctc_loss=0.1383, cr_loss=0.3541, over 3365700.92 frames. ], batch size: 49, lr: 5.23e-03, grad_scale: 8.0
2024-09-24 03:05:06,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=406340.6666666667, ans=0.0
2024-09-24 03:05:06,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=406340.6666666667, ans=0.125
2024-09-24 03:05:14,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=406340.6666666667, ans=0.125
2024-09-24 03:05:47,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=406434.0, ans=22.5
2024-09-24 03:06:07,397 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.244e+02 1.338e+02 1.449e+02 2.733e+02, threshold=2.676e+02, percent-clipped=1.0
2024-09-24 03:06:07,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=406480.6666666667, ans=0.0
2024-09-24 03:06:14,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=406527.3333333333, ans=0.125
2024-09-24 03:06:15,519 INFO [train.py:1198] (1/4) Epoch 23, batch 1400, loss[loss=0.2497, ctc_loss=0.1711, cr_loss=0.3932, over 14910.00 frames. ], tot_loss[loss=0.2093, ctc_loss=0.1384, cr_loss=0.3544, over 3372384.59 frames. ], batch size: 88, lr: 5.22e-03, grad_scale: 8.0
2024-09-24 03:06:15,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=406527.3333333333, ans=0.025
2024-09-24 03:06:49,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=406620.6666666667, ans=0.125
2024-09-24 03:06:50,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=406620.6666666667, ans=0.0
2024-09-24 03:07:18,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=406714.0, ans=0.0
2024-09-24 03:07:29,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=406714.0, ans=0.125
2024-09-24 03:07:35,430 INFO [train.py:1198] (1/4) Epoch 23, batch 1450, loss[loss=0.1966, ctc_loss=0.1302, cr_loss=0.3318, over 17065.00 frames. ], tot_loss[loss=0.2077, ctc_loss=0.1373, cr_loss=0.3522, over 3368803.34 frames. ], batch size: 46, lr: 5.22e-03, grad_scale: 8.0
2024-09-24 03:07:35,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=406760.6666666667, ans=0.2
2024-09-24 03:08:00,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=406807.3333333333, ans=0.1
2024-09-24 03:08:16,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=406854.0, ans=0.2
2024-09-24 03:08:38,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=406900.6666666667, ans=0.125
2024-09-24 03:08:51,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=406947.3333333333, ans=0.025
2024-09-24 03:08:52,575 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.227e+02 1.307e+02 1.393e+02 1.809e+02, threshold=2.613e+02, percent-clipped=0.0
2024-09-24 03:09:00,495 INFO [train.py:1198] (1/4) Epoch 23, batch 1500, loss[loss=0.2557, ctc_loss=0.1716, cr_loss=0.4202, over 17046.00 frames. ], tot_loss[loss=0.2089, ctc_loss=0.1381, cr_loss=0.3541, over 3364612.85 frames. ], batch size: 52, lr: 5.22e-03, grad_scale: 8.0
2024-09-24 03:09:13,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=406994.0, ans=0.0
2024-09-24 03:09:55,342 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.50 vs. limit=15.0
2024-09-24 03:10:09,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=407180.6666666667, ans=0.0
2024-09-24 03:10:20,386 INFO [train.py:1198] (1/4) Epoch 23, batch 1550, loss[loss=0.2271, ctc_loss=0.1484, cr_loss=0.3934, over 17246.00 frames. ], tot_loss[loss=0.21, ctc_loss=0.1388, cr_loss=0.3559, over 3364471.53 frames. ], batch size: 55, lr: 5.22e-03, grad_scale: 8.0
2024-09-24 03:10:26,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=407227.3333333333, ans=0.1
2024-09-24 03:10:31,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=407227.3333333333, ans=0.07
2024-09-24 03:10:50,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=407320.6666666667, ans=10.0
2024-09-24 03:11:07,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=407320.6666666667, ans=0.0
2024-09-24 03:11:37,267 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.292e+02 1.390e+02 1.535e+02 2.050e+02, threshold=2.781e+02, percent-clipped=0.0
2024-09-24 03:11:45,285 INFO [train.py:1198] (1/4) Epoch 23, batch 1600, loss[loss=0.1886, ctc_loss=0.122, cr_loss=0.3329, over 17183.00 frames. ], tot_loss[loss=0.2091, ctc_loss=0.1382, cr_loss=0.3546, over 3360232.45 frames. ], batch size: 41, lr: 5.22e-03, grad_scale: 16.0
2024-09-24 03:12:19,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=407554.0, ans=0.125
2024-09-24 03:12:23,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=407554.0, ans=0.125
2024-09-24 03:12:32,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=407600.6666666667, ans=0.125
2024-09-24 03:12:32,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=407600.6666666667, ans=0.125
2024-09-24 03:12:36,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=407600.6666666667, ans=0.09899494936611666
2024-09-24 03:13:08,010 INFO [train.py:1198] (1/4) Epoch 23, batch 1650, loss[loss=0.2127, ctc_loss=0.1383, cr_loss=0.3716, over 17207.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1386, cr_loss=0.3551, over 3355586.12 frames. ], batch size: 47, lr: 5.22e-03, grad_scale: 16.0
2024-09-24 03:13:11,608 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 03:13:32,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=407740.6666666667, ans=0.125
2024-09-24 03:13:45,917 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.14 vs. limit=15.0
2024-09-24 03:13:46,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=407787.3333333333, ans=0.2
2024-09-24 03:14:21,924 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.251e+02 1.320e+02 1.451e+02 2.604e+02, threshold=2.640e+02, percent-clipped=0.0
2024-09-24 03:14:24,702 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.11 vs. limit=15.0
2024-09-24 03:14:29,899 INFO [train.py:1198] (1/4) Epoch 23, batch 1700, loss[loss=0.2623, ctc_loss=0.1854, cr_loss=0.3846, over 11840.00 frames. ], tot_loss[loss=0.2112, ctc_loss=0.1397, cr_loss=0.3571, over 3339931.60 frames. ], batch size: 123, lr: 5.22e-03, grad_scale: 16.0
2024-09-24 03:14:44,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.22 vs. limit=15.0
2024-09-24 03:15:14,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=408020.6666666667, ans=0.125
2024-09-24 03:15:26,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=408067.3333333333, ans=0.2
2024-09-24 03:15:38,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=408114.0, ans=0.0
2024-09-24 03:15:52,352 INFO [train.py:1198] (1/4) Epoch 23, batch 1750, loss[loss=0.2161, ctc_loss=0.1431, cr_loss=0.3651, over 17302.00 frames. ], tot_loss[loss=0.2098, ctc_loss=0.1387, cr_loss=0.3556, over 3350459.55 frames. ], batch size: 46, lr: 5.21e-03, grad_scale: 16.0
2024-09-24 03:15:55,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=408160.6666666667, ans=0.0
2024-09-24 03:16:33,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=408254.0, ans=0.125
2024-09-24 03:16:39,985 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.78 vs. limit=22.5
2024-09-24 03:16:51,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.88 vs. limit=22.5
2024-09-24 03:16:57,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=408347.3333333333, ans=0.1
2024-09-24 03:17:06,658 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.263e+02 1.353e+02 1.472e+02 2.567e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-24 03:17:14,533 INFO [train.py:1198] (1/4) Epoch 23, batch 1800, loss[loss=0.1811, ctc_loss=0.1169, cr_loss=0.3207, over 16934.00 frames. ], tot_loss[loss=0.2102, ctc_loss=0.1391, cr_loss=0.3556, over 3341030.71 frames. ], batch size: 42, lr: 5.21e-03, grad_scale: 16.0
2024-09-24 03:17:21,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=408394.0, ans=0.2
2024-09-24 03:17:41,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=408440.6666666667, ans=0.07
2024-09-24 03:17:42,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=408440.6666666667, ans=0.125
2024-09-24 03:17:57,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=408487.3333333333, ans=0.0
2024-09-24 03:18:01,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=10.14 vs. limit=22.5
2024-09-24 03:18:39,947 INFO [train.py:1198] (1/4) Epoch 23, batch 1850, loss[loss=0.2331, ctc_loss=0.1536, cr_loss=0.3974, over 17025.00 frames. ], tot_loss[loss=0.2087, ctc_loss=0.1379, cr_loss=0.3538, over 3348415.72 frames. ], batch size: 53, lr: 5.21e-03, grad_scale: 16.0
2024-09-24 03:18:42,177 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.37 vs. limit=15.0
2024-09-24 03:19:29,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=408767.3333333333, ans=0.0
2024-09-24 03:19:50,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=408814.0, ans=0.1
2024-09-24 03:19:50,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=408814.0, ans=0.125
2024-09-24 03:19:51,998 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.253e+02 1.335e+02 1.430e+02 2.025e+02, threshold=2.670e+02, percent-clipped=0.0
2024-09-24 03:19:59,998 INFO [train.py:1198] (1/4) Epoch 23, batch 1900, loss[loss=0.1946, ctc_loss=0.1266, cr_loss=0.3399, over 17051.00 frames. ], tot_loss[loss=0.2075, ctc_loss=0.1371, cr_loss=0.3522, over 3354726.22 frames. ], batch size: 39, lr: 5.21e-03, grad_scale: 16.0
2024-09-24 03:20:00,993 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.68 vs. limit=6.0
2024-09-24 03:20:44,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=408954.0, ans=0.125
2024-09-24 03:20:55,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=409000.6666666667, ans=0.0
2024-09-24 03:20:59,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=409000.6666666667, ans=0.07
2024-09-24 03:21:06,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=409000.6666666667, ans=0.125
2024-09-24 03:21:25,580 INFO [train.py:1198] (1/4) Epoch 23, batch 1950, loss[loss=0.1871, ctc_loss=0.1186, cr_loss=0.3425, over 16973.00 frames. ], tot_loss[loss=0.2075, ctc_loss=0.1369, cr_loss=0.3527, over 3359598.99 frames. ], batch size: 42, lr: 5.21e-03, grad_scale: 16.0
2024-09-24 03:21:33,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=409094.0, ans=0.1
2024-09-24 03:21:39,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=409094.0, ans=15.0
2024-09-24 03:21:55,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=409140.6666666667, ans=0.2
2024-09-24 03:22:18,035 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.98 vs. limit=10.0
2024-09-24 03:22:20,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=409234.0, ans=0.0
2024-09-24 03:22:24,318 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.02 vs. limit=12.0
2024-09-24 03:22:40,198 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.272e+02 1.356e+02 1.502e+02 2.746e+02, threshold=2.712e+02, percent-clipped=1.0
2024-09-24 03:22:48,040 INFO [train.py:1198] (1/4) Epoch 23, batch 2000, loss[loss=0.2377, ctc_loss=0.1587, cr_loss=0.395, over 16980.00 frames. ], tot_loss[loss=0.2089, ctc_loss=0.138, cr_loss=0.3545, over 3354551.62 frames. ], batch size: 53, lr: 5.21e-03, grad_scale: 32.0
2024-09-24 03:23:04,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=409374.0, ans=0.125
2024-09-24 03:23:40,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=409467.3333333333, ans=0.07
2024-09-24 03:23:49,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=409467.3333333333, ans=0.0
2024-09-24 03:24:10,501 INFO [train.py:1198] (1/4) Epoch 23, batch 2050, loss[loss=0.1916, ctc_loss=0.1274, cr_loss=0.3213, over 17090.00 frames. ], tot_loss[loss=0.2095, ctc_loss=0.1385, cr_loss=0.3551, over 3346599.97 frames. ], batch size: 40, lr: 5.20e-03, grad_scale: 16.0
2024-09-24 03:24:10,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=409560.6666666667, ans=0.0
2024-09-24 03:24:14,165 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 03:24:19,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.63 vs. limit=15.0
2024-09-24 03:24:20,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=409560.6666666667, ans=0.5
2024-09-24 03:24:54,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=409654.0, ans=0.1
2024-09-24 03:25:07,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=409700.6666666667, ans=0.125
2024-09-24 03:25:07,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=409700.6666666667, ans=0.2
2024-09-24 03:25:24,757 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.284e+02 1.371e+02 1.476e+02 1.863e+02, threshold=2.742e+02, percent-clipped=0.0
2024-09-24 03:25:31,152 INFO [train.py:1198] (1/4) Epoch 23, batch 2100, loss[loss=0.1813, ctc_loss=0.1192, cr_loss=0.3102, over 16957.00 frames. ], tot_loss[loss=0.2104, ctc_loss=0.1392, cr_loss=0.356, over 3342432.89 frames. ], batch size: 42, lr: 5.20e-03, grad_scale: 16.0
2024-09-24 03:25:36,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=409794.0, ans=0.125
2024-09-24 03:25:38,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=409794.0, ans=0.1
2024-09-24 03:25:40,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=409794.0, ans=0.0
2024-09-24 03:25:45,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=409794.0, ans=0.1
2024-09-24 03:25:56,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=409840.6666666667, ans=0.125
2024-09-24 03:26:11,562 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 03:26:14,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=409887.3333333333, ans=0.025
2024-09-24 03:26:21,176 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 03:26:56,009 INFO [train.py:1198] (1/4) Epoch 23, batch 2150, loss[loss=0.1939, ctc_loss=0.1274, cr_loss=0.3323, over 17194.00 frames. ], tot_loss[loss=0.2113, ctc_loss=0.1398, cr_loss=0.3574, over 3343096.21 frames. ], batch size: 41, lr: 5.20e-03, grad_scale: 16.0
2024-09-24 03:27:07,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=410027.3333333333, ans=0.125
2024-09-24 03:27:45,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=410167.3333333333, ans=0.125
2024-09-24 03:27:50,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=410167.3333333333, ans=0.1
2024-09-24 03:27:50,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=410167.3333333333, ans=0.125
2024-09-24 03:27:50,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=410167.3333333333, ans=0.5
2024-09-24 03:28:15,213 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.254e+02 1.321e+02 1.428e+02 2.805e+02, threshold=2.642e+02, percent-clipped=1.0
2024-09-24 03:28:15,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=410214.0, ans=0.0
2024-09-24 03:28:18,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=410214.0, ans=0.125
2024-09-24 03:28:21,744 INFO [train.py:1198] (1/4) Epoch 23, batch 2200, loss[loss=0.229, ctc_loss=0.1514, cr_loss=0.388, over 17363.00 frames. ], tot_loss[loss=0.2117, ctc_loss=0.1402, cr_loss=0.3578, over 3340011.08 frames. ], batch size: 48, lr: 5.20e-03, grad_scale: 16.0
2024-09-24 03:28:34,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=410260.6666666667, ans=0.0
2024-09-24 03:29:26,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=410447.3333333333, ans=0.125
2024-09-24 03:29:38,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=410447.3333333333, ans=0.025
2024-09-24 03:29:40,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=410494.0, ans=0.1
2024-09-24 03:29:41,637 INFO [train.py:1198] (1/4) Epoch 23, batch 2250, loss[loss=0.2412, ctc_loss=0.1696, cr_loss=0.358, over 12274.00 frames. ], tot_loss[loss=0.2115, ctc_loss=0.14, cr_loss=0.3573, over 3335806.08 frames. ], batch size: 123, lr: 5.20e-03, grad_scale: 16.0
2024-09-24 03:29:43,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=410494.0, ans=0.125
2024-09-24 03:30:35,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=410634.0, ans=0.0
2024-09-24 03:30:37,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=410634.0, ans=0.0
2024-09-24 03:30:42,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=410634.0, ans=0.125
2024-09-24 03:30:49,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=410680.6666666667, ans=0.1
2024-09-24 03:31:02,825 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.239e+02 1.305e+02 1.382e+02 2.270e+02, threshold=2.609e+02, percent-clipped=0.0
2024-09-24 03:31:09,443 INFO [train.py:1198] (1/4) Epoch 23, batch 2300, loss[loss=0.2252, ctc_loss=0.1498, cr_loss=0.3771, over 17159.00 frames. ], tot_loss[loss=0.2099, ctc_loss=0.1389, cr_loss=0.3549, over 3346434.30 frames. ], batch size: 48, lr: 5.20e-03, grad_scale: 16.0
2024-09-24 03:31:26,626 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.05 vs. limit=10.0
2024-09-24 03:31:40,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=410820.6666666667, ans=0.125
2024-09-24 03:31:40,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=410820.6666666667, ans=0.125
2024-09-24 03:32:02,839 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=9.15 vs. limit=15.0
2024-09-24 03:32:12,528 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.53 vs. limit=6.0
2024-09-24 03:32:21,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=410914.0, ans=0.125
2024-09-24 03:32:31,674 INFO [train.py:1198] (1/4) Epoch 23, batch 2350, loss[loss=0.1802, ctc_loss=0.1175, cr_loss=0.3139, over 17011.00 frames. ], tot_loss[loss=0.2079, ctc_loss=0.1375, cr_loss=0.3523, over 3347588.70 frames. ], batch size: 51, lr: 5.20e-03, grad_scale: 16.0
2024-09-24 03:32:35,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=410960.6666666667, ans=0.025
2024-09-24 03:32:40,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=410960.6666666667, ans=0.2
2024-09-24 03:33:01,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=411007.3333333333, ans=0.125
2024-09-24 03:33:16,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=411054.0, ans=0.125
2024-09-24 03:33:34,332 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.90 vs. limit=15.0
2024-09-24 03:33:47,828 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.259e+02 1.351e+02 1.489e+02 2.005e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-24 03:33:53,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=411194.0, ans=0.125
2024-09-24 03:33:54,327 INFO [train.py:1198] (1/4) Epoch 23, batch 2400, loss[loss=0.1735, ctc_loss=0.1117, cr_loss=0.3089, over 16950.00 frames. ], tot_loss[loss=0.2092, ctc_loss=0.1384, cr_loss=0.354, over 3347368.09 frames. ], batch size: 42, lr: 5.19e-03, grad_scale: 32.0
2024-09-24 03:34:01,401 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.03 vs. limit=15.0
2024-09-24 03:34:46,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=411334.0, ans=0.025
2024-09-24 03:34:47,905 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.01 vs. limit=22.5
2024-09-24 03:34:53,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=411334.0, ans=0.1
2024-09-24 03:35:14,351 INFO [train.py:1198] (1/4) Epoch 23, batch 2450, loss[loss=0.2029, ctc_loss=0.1341, cr_loss=0.3438, over 17116.00 frames. ], tot_loss[loss=0.2093, ctc_loss=0.1384, cr_loss=0.3543, over 3351297.30 frames. ], batch size: 40, lr: 5.19e-03, grad_scale: 16.0
2024-09-24 03:35:18,048 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.97 vs. limit=15.0
2024-09-24 03:35:33,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=411474.0, ans=0.035
2024-09-24 03:35:40,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=411474.0, ans=0.125
2024-09-24 03:36:04,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=411520.6666666667, ans=0.125
2024-09-24 03:36:10,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=411567.3333333333, ans=0.0
2024-09-24 03:36:27,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=411614.0, ans=0.125
2024-09-24 03:36:30,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=411614.0, ans=0.1
2024-09-24 03:36:35,083 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.268e+02 1.337e+02 1.491e+02 2.179e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-24 03:36:39,835 INFO [train.py:1198] (1/4) Epoch 23, batch 2500, loss[loss=0.2169, ctc_loss=0.1457, cr_loss=0.3563, over 16984.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1387, cr_loss=0.3551, over 3351249.48 frames. ], batch size: 56, lr: 5.19e-03, grad_scale: 16.0
2024-09-24 03:36:40,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=411660.6666666667, ans=0.125
2024-09-24 03:36:54,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=411707.3333333333, ans=0.1
2024-09-24 03:37:02,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=411707.3333333333, ans=0.125
2024-09-24 03:37:24,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=411754.0, ans=0.125
2024-09-24 03:37:37,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=411800.6666666667, ans=0.125
2024-09-24 03:37:41,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=411800.6666666667, ans=0.2
2024-09-24 03:37:45,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=411847.3333333333, ans=0.2
2024-09-24 03:38:02,886 INFO [train.py:1198] (1/4) Epoch 23, batch 2550, loss[loss=0.187, ctc_loss=0.122, cr_loss=0.3251, over 16777.00 frames. ], tot_loss[loss=0.2092, ctc_loss=0.1383, cr_loss=0.3544, over 3355157.25 frames. ], batch size: 37, lr: 5.19e-03, grad_scale: 16.0
2024-09-24 03:38:16,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=411894.0, ans=0.0
2024-09-24 03:38:27,034 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.48 vs. limit=22.5
2024-09-24 03:38:30,358 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.64 vs. limit=15.0
2024-09-24 03:39:00,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=412034.0, ans=0.125
2024-09-24 03:39:20,140 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.256e+02 1.367e+02 1.488e+02 2.148e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-24 03:39:24,809 INFO [train.py:1198] (1/4) Epoch 23, batch 2600, loss[loss=0.1832, ctc_loss=0.117, cr_loss=0.3312, over 17097.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1386, cr_loss=0.3551, over 3357943.18 frames. ], batch size: 43, lr: 5.19e-03, grad_scale: 16.0
2024-09-24 03:39:31,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=412127.3333333333, ans=0.0
2024-09-24 03:39:33,225 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.47 vs. limit=15.0
2024-09-24 03:40:03,976 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=14.43 vs. limit=15.0
2024-09-24 03:40:05,458 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.76 vs. limit=15.0
2024-09-24 03:40:06,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=412220.6666666667, ans=0.0
2024-09-24 03:40:21,986 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.34 vs. limit=15.0
2024-09-24 03:40:37,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=412314.0, ans=0.0
2024-09-24 03:40:43,262 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.25 vs. limit=22.5
2024-09-24 03:40:49,869 INFO [train.py:1198] (1/4) Epoch 23, batch 2650, loss[loss=0.2342, ctc_loss=0.1578, cr_loss=0.3818, over 16904.00 frames. ], tot_loss[loss=0.2103, ctc_loss=0.1392, cr_loss=0.3558, over 3359019.54 frames. ], batch size: 58, lr: 5.19e-03, grad_scale: 16.0
2024-09-24 03:40:51,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=412360.6666666667, ans=0.2
2024-09-24 03:41:16,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=412407.3333333333, ans=15.0
2024-09-24 03:41:17,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=412407.3333333333, ans=0.07
2024-09-24 03:41:27,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=412454.0, ans=0.125
2024-09-24 03:41:35,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=412454.0, ans=0.0
2024-09-24 03:41:54,069 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.80 vs. limit=15.0
2024-09-24 03:41:55,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=412547.3333333333, ans=0.125
2024-09-24 03:41:58,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=412547.3333333333, ans=0.07
2024-09-24 03:41:59,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=412547.3333333333, ans=0.1
2024-09-24 03:42:02,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=412547.3333333333, ans=0.125
2024-09-24 03:42:05,849 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.276e+02 1.370e+02 1.490e+02 2.063e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-24 03:42:09,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=412594.0, ans=0.07
2024-09-24 03:42:10,851 INFO [train.py:1198] (1/4) Epoch 23, batch 2700, loss[loss=0.1964, ctc_loss=0.1312, cr_loss=0.3257, over 17171.00 frames. ], tot_loss[loss=0.2093, ctc_loss=0.1384, cr_loss=0.3545, over 3363968.45 frames. ], batch size: 45, lr: 5.19e-03, grad_scale: 16.0
2024-09-24 03:42:12,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=412594.0, ans=0.125
2024-09-24 03:42:24,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=412594.0, ans=0.1
2024-09-24 03:42:28,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=412640.6666666667, ans=0.125
2024-09-24 03:42:30,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.55 vs. limit=12.0
2024-09-24 03:42:36,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.99 vs. limit=22.5
2024-09-24 03:42:49,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer_ff2.min_abs, batch_count=412687.3333333333, ans=0.1
2024-09-24 03:42:54,996 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=7.23 vs. limit=15.0
2024-09-24 03:43:14,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=412734.0, ans=0.125
2024-09-24 03:43:35,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=412827.3333333333, ans=0.1
2024-09-24 03:43:36,486 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=4.29 vs. limit=15.0
2024-09-24 03:43:37,018 INFO [train.py:1198] (1/4) Epoch 23, batch 2750, loss[loss=0.2501, ctc_loss=0.1671, cr_loss=0.4151, over 15207.00 frames. ], tot_loss[loss=0.2094, ctc_loss=0.1385, cr_loss=0.3541, over 3355783.34 frames. ], batch size: 89, lr: 5.18e-03, grad_scale: 16.0
2024-09-24 03:44:22,564 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.48 vs. limit=6.0
2024-09-24 03:44:33,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.min_positive, batch_count=412967.3333333333, ans=0.025
2024-09-24 03:44:52,047 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.288e+02 1.407e+02 1.537e+02 4.593e+02, threshold=2.814e+02, percent-clipped=2.0
2024-09-24 03:44:56,674 INFO [train.py:1198] (1/4) Epoch 23, batch 2800, loss[loss=0.2122, ctc_loss=0.1395, cr_loss=0.3636, over 17306.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1387, cr_loss=0.3546, over 3349982.72 frames. ], batch size: 51, lr: 5.18e-03, grad_scale: 32.0
2024-09-24 03:45:13,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=413107.3333333333, ans=0.1
2024-09-24 03:45:26,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=413107.3333333333, ans=0.1
2024-09-24 03:45:29,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=413154.0, ans=0.0
2024-09-24 03:45:50,438 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.38 vs. limit=15.0
2024-09-24 03:45:51,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=413200.6666666667, ans=0.0
2024-09-24 03:45:53,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten.whitening_limit, batch_count=413200.6666666667, ans=15.0
2024-09-24 03:46:01,407 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 03:46:09,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=413247.3333333333, ans=0.125
2024-09-24 03:46:21,928 INFO [train.py:1198] (1/4) Epoch 23, batch 2850, loss[loss=0.211, ctc_loss=0.1338, cr_loss=0.3856, over 17023.00 frames. ], tot_loss[loss=0.2082, ctc_loss=0.1376, cr_loss=0.3531, over 3357691.43 frames. ], batch size: 51, lr: 5.18e-03, grad_scale: 32.0
2024-09-24 03:46:32,310 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.46 vs. limit=15.0
2024-09-24 03:46:33,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=413294.0, ans=0.0
2024-09-24 03:47:32,558 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.57 vs. limit=12.0
2024-09-24 03:47:39,673 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.282e+02 1.396e+02 1.534e+02 2.289e+02, threshold=2.792e+02, percent-clipped=0.0
2024-09-24 03:47:44,651 INFO [train.py:1198] (1/4) Epoch 23, batch 2900, loss[loss=0.1805, ctc_loss=0.1189, cr_loss=0.3082, over 17281.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.1374, cr_loss=0.3528, over 3355958.15 frames. ], batch size: 42, lr: 5.18e-03, grad_scale: 32.0
2024-09-24 03:47:57,853 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.18 vs. limit=6.0
2024-09-24 03:48:05,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=413574.0, ans=0.0
2024-09-24 03:48:08,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=413574.0, ans=0.125
2024-09-24 03:48:11,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=413574.0, ans=0.125
2024-09-24 03:48:13,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=413574.0, ans=0.0
2024-09-24 03:48:16,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=413574.0, ans=0.09899494936611666
2024-09-24 03:48:16,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=413574.0, ans=0.2
2024-09-24 03:48:24,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=413620.6666666667, ans=0.125
2024-09-24 03:48:42,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=413667.3333333333, ans=0.0
2024-09-24 03:48:59,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=413714.0, ans=0.125
2024-09-24 03:48:59,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=413714.0, ans=0.2
2024-09-24 03:49:07,438 INFO [train.py:1198] (1/4) Epoch 23, batch 2950, loss[loss=0.2114, ctc_loss=0.1388, cr_loss=0.3631, over 17314.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1386, cr_loss=0.3549, over 3354083.61 frames. ], batch size: 51, lr: 5.18e-03, grad_scale: 32.0
2024-09-24 03:49:16,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=413760.6666666667, ans=0.125
2024-09-24 03:49:44,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=413854.0, ans=0.125
2024-09-24 03:49:45,290 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.40 vs. limit=15.0
2024-09-24 03:50:08,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=413900.6666666667, ans=0.125
2024-09-24 03:50:22,355 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.031e+02 1.232e+02 1.296e+02 1.402e+02 3.065e+02, threshold=2.592e+02, percent-clipped=1.0
2024-09-24 03:50:24,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=413947.3333333333, ans=0.125
2024-09-24 03:50:27,184 INFO [train.py:1198] (1/4) Epoch 23, batch 3000, loss[loss=0.2171, ctc_loss=0.1425, cr_loss=0.3732, over 17362.00 frames. ], tot_loss[loss=0.2099, ctc_loss=0.1388, cr_loss=0.3555, over 3357524.21 frames. ], batch size: 48, lr: 5.18e-03, grad_scale: 32.0
2024-09-24 03:50:27,184 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 03:50:42,639 INFO [train.py:1230] (1/4) Epoch 23, validation: loss=0.03816, ctc_loss=0.03816, cr_loss=8.083e-15, over 944034.00 frames. 
2024-09-24 03:50:42,640 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 03:50:53,189 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.54 vs. limit=12.0
2024-09-24 03:50:57,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=414040.6666666667, ans=0.2
2024-09-24 03:51:32,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=414134.0, ans=0.0
2024-09-24 03:51:43,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=414134.0, ans=0.025
2024-09-24 03:51:51,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=414180.6666666667, ans=0.0
2024-09-24 03:52:03,780 INFO [train.py:1198] (1/4) Epoch 23, batch 3050, loss[loss=0.1733, ctc_loss=0.1122, cr_loss=0.3055, over 17025.00 frames. ], tot_loss[loss=0.209, ctc_loss=0.138, cr_loss=0.355, over 3370265.61 frames. ], batch size: 51, lr: 5.18e-03, grad_scale: 16.0
2024-09-24 03:52:04,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=414227.3333333333, ans=0.125
2024-09-24 03:52:05,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=414227.3333333333, ans=0.0
2024-09-24 03:52:10,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=414227.3333333333, ans=0.0
2024-09-24 03:52:15,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=414227.3333333333, ans=0.1
2024-09-24 03:53:05,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=414414.0, ans=0.0
2024-09-24 03:53:17,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=414414.0, ans=0.125
2024-09-24 03:53:19,031 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.284e+02 1.405e+02 1.521e+02 2.229e+02, threshold=2.810e+02, percent-clipped=0.0
2024-09-24 03:53:22,205 INFO [train.py:1198] (1/4) Epoch 23, batch 3100, loss[loss=0.1916, ctc_loss=0.1233, cr_loss=0.3411, over 17215.00 frames. ], tot_loss[loss=0.2087, ctc_loss=0.1378, cr_loss=0.3545, over 3366998.83 frames. ], batch size: 47, lr: 5.17e-03, grad_scale: 16.0
2024-09-24 03:53:22,984 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.19 vs. limit=15.0
2024-09-24 03:53:42,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=414507.3333333333, ans=0.125
2024-09-24 03:53:44,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=414507.3333333333, ans=0.125
2024-09-24 03:54:14,481 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.61 vs. limit=15.0
2024-09-24 03:54:21,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.whiten.whitening_limit, batch_count=414600.6666666667, ans=15.0
2024-09-24 03:54:25,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=414647.3333333333, ans=0.125
2024-09-24 03:54:37,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=414647.3333333333, ans=0.125
2024-09-24 03:54:42,871 INFO [train.py:1198] (1/4) Epoch 23, batch 3150, loss[loss=0.2162, ctc_loss=0.1446, cr_loss=0.3578, over 17158.00 frames. ], tot_loss[loss=0.2081, ctc_loss=0.1374, cr_loss=0.3533, over 3369793.30 frames. ], batch size: 45, lr: 5.17e-03, grad_scale: 16.0
2024-09-24 03:55:23,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=414787.3333333333, ans=0.2
2024-09-24 03:55:40,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer_ff2.min_abs, batch_count=414834.0, ans=0.1
2024-09-24 03:55:59,846 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.246e+02 1.369e+02 1.500e+02 3.916e+02, threshold=2.738e+02, percent-clipped=1.0
2024-09-24 03:56:03,017 INFO [train.py:1198] (1/4) Epoch 23, batch 3200, loss[loss=0.232, ctc_loss=0.1533, cr_loss=0.3935, over 17150.00 frames. ], tot_loss[loss=0.2098, ctc_loss=0.1386, cr_loss=0.356, over 3367096.43 frames. ], batch size: 48, lr: 5.17e-03, grad_scale: 32.0
2024-09-24 03:56:56,925 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.73 vs. limit=22.5
2024-09-24 03:57:03,419 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.41 vs. limit=15.0
2024-09-24 03:57:21,131 INFO [train.py:1198] (1/4) Epoch 23, batch 3250, loss[loss=0.2841, ctc_loss=0.1948, cr_loss=0.4465, over 12240.00 frames. ], tot_loss[loss=0.2102, ctc_loss=0.1389, cr_loss=0.3565, over 3366605.31 frames. ], batch size: 124, lr: 5.17e-03, grad_scale: 32.0
2024-09-24 03:57:30,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=415160.6666666667, ans=0.125
2024-09-24 03:57:43,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=415207.3333333333, ans=0.2
2024-09-24 03:57:46,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=415207.3333333333, ans=0.2
2024-09-24 03:57:47,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=415207.3333333333, ans=0.1
2024-09-24 03:58:00,275 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.20 vs. limit=6.0
2024-09-24 03:58:12,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=415300.6666666667, ans=0.025
2024-09-24 03:58:16,054 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.95 vs. limit=15.0
2024-09-24 03:58:19,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten.whitening_limit, batch_count=415300.6666666667, ans=15.0
2024-09-24 03:58:23,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=415347.3333333333, ans=0.125
2024-09-24 03:58:36,028 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.250e+02 1.334e+02 1.506e+02 2.556e+02, threshold=2.668e+02, percent-clipped=0.0
2024-09-24 03:58:37,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=415394.0, ans=0.1
2024-09-24 03:58:39,220 INFO [train.py:1198] (1/4) Epoch 23, batch 3300, loss[loss=0.2313, ctc_loss=0.1533, cr_loss=0.3897, over 16387.00 frames. ], tot_loss[loss=0.2107, ctc_loss=0.1392, cr_loss=0.3573, over 3368576.69 frames. ], batch size: 66, lr: 5.17e-03, grad_scale: 32.0
2024-09-24 03:58:45,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=415394.0, ans=0.125
2024-09-24 03:58:58,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=415440.6666666667, ans=0.5
2024-09-24 03:59:28,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=415534.0, ans=0.1
2024-09-24 03:59:32,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=415534.0, ans=0.125
2024-09-24 03:59:34,987 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.65 vs. limit=15.0
2024-09-24 03:59:43,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=415580.6666666667, ans=0.1
2024-09-24 03:59:51,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=415580.6666666667, ans=0.125
2024-09-24 03:59:51,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=415580.6666666667, ans=0.125
2024-09-24 03:59:57,317 INFO [train.py:1198] (1/4) Epoch 23, batch 3350, loss[loss=0.2294, ctc_loss=0.152, cr_loss=0.3867, over 17032.00 frames. ], tot_loss[loss=0.21, ctc_loss=0.1389, cr_loss=0.3558, over 3369462.49 frames. ], batch size: 52, lr: 5.17e-03, grad_scale: 32.0
2024-09-24 03:59:57,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=415627.3333333333, ans=0.125
2024-09-24 04:00:19,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=415674.0, ans=0.125
2024-09-24 04:00:57,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=415767.3333333333, ans=0.1
2024-09-24 04:01:13,684 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.12 vs. limit=15.0
2024-09-24 04:01:14,621 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.296e+02 1.415e+02 1.543e+02 2.020e+02, threshold=2.829e+02, percent-clipped=0.0
2024-09-24 04:01:17,788 INFO [train.py:1198] (1/4) Epoch 23, batch 3400, loss[loss=0.2088, ctc_loss=0.1385, cr_loss=0.3514, over 17215.00 frames. ], tot_loss[loss=0.2109, ctc_loss=0.1395, cr_loss=0.3567, over 3371448.76 frames. ], batch size: 47, lr: 5.17e-03, grad_scale: 32.0
2024-09-24 04:01:24,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=415860.6666666667, ans=0.125
2024-09-24 04:01:24,494 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.14 vs. limit=6.0
2024-09-24 04:02:23,084 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.38 vs. limit=12.0
2024-09-24 04:02:38,113 INFO [train.py:1198] (1/4) Epoch 23, batch 3450, loss[loss=0.2361, ctc_loss=0.1591, cr_loss=0.3849, over 16889.00 frames. ], tot_loss[loss=0.2117, ctc_loss=0.1403, cr_loss=0.3573, over 3359713.69 frames. ], batch size: 58, lr: 5.16e-03, grad_scale: 32.0
2024-09-24 04:02:52,666 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 04:02:54,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=416140.6666666667, ans=0.0
2024-09-24 04:02:58,191 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.64 vs. limit=15.0
2024-09-24 04:03:13,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=416187.3333333333, ans=0.125
2024-09-24 04:03:32,965 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.59 vs. limit=15.0
2024-09-24 04:03:53,800 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.237e+02 1.333e+02 1.427e+02 1.642e+02, threshold=2.667e+02, percent-clipped=0.0
2024-09-24 04:03:57,047 INFO [train.py:1198] (1/4) Epoch 23, batch 3500, loss[loss=0.2042, ctc_loss=0.1351, cr_loss=0.3452, over 17156.00 frames. ], tot_loss[loss=0.21, ctc_loss=0.139, cr_loss=0.3549, over 3353200.68 frames. ], batch size: 45, lr: 5.16e-03, grad_scale: 32.0
2024-09-24 04:04:12,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=416374.0, ans=0.125
2024-09-24 04:04:36,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=416420.6666666667, ans=0.0
2024-09-24 04:04:47,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=416467.3333333333, ans=0.0
2024-09-24 04:04:54,527 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=17.33 vs. limit=22.5
2024-09-24 04:04:58,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=416467.3333333333, ans=0.2
2024-09-24 04:05:15,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=416560.6666666667, ans=0.025
2024-09-24 04:05:17,346 INFO [train.py:1198] (1/4) Epoch 23, batch 3550, loss[loss=0.1775, ctc_loss=0.1183, cr_loss=0.2957, over 17275.00 frames. ], tot_loss[loss=0.2098, ctc_loss=0.139, cr_loss=0.3543, over 3353697.66 frames. ], batch size: 42, lr: 5.16e-03, grad_scale: 32.0
2024-09-24 04:05:30,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=416560.6666666667, ans=0.0
2024-09-24 04:06:00,571 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.36 vs. limit=10.0
2024-09-24 04:06:22,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=416747.3333333333, ans=0.125
2024-09-24 04:06:23,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=416747.3333333333, ans=0.2
2024-09-24 04:06:29,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=416747.3333333333, ans=0.125
2024-09-24 04:06:34,282 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.260e+02 1.366e+02 1.469e+02 1.975e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-24 04:06:34,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=416747.3333333333, ans=0.0
2024-09-24 04:06:37,348 INFO [train.py:1198] (1/4) Epoch 23, batch 3600, loss[loss=0.2356, ctc_loss=0.1599, cr_loss=0.3783, over 17302.00 frames. ], tot_loss[loss=0.209, ctc_loss=0.1383, cr_loss=0.3536, over 3367727.09 frames. ], batch size: 51, lr: 5.16e-03, grad_scale: 32.0
2024-09-24 04:06:46,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=416794.0, ans=0.125
2024-09-24 04:06:59,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=416840.6666666667, ans=0.125
2024-09-24 04:07:07,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=416887.3333333333, ans=0.04949747468305833
2024-09-24 04:07:11,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=416887.3333333333, ans=0.125
2024-09-24 04:07:25,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=416934.0, ans=0.2
2024-09-24 04:07:30,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=416934.0, ans=0.125
2024-09-24 04:07:44,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=416980.6666666667, ans=0.125
2024-09-24 04:07:55,450 INFO [train.py:1198] (1/4) Epoch 23, batch 3650, loss[loss=0.1943, ctc_loss=0.129, cr_loss=0.3261, over 17035.00 frames. ], tot_loss[loss=0.2089, ctc_loss=0.1382, cr_loss=0.3538, over 3363956.17 frames. ], batch size: 52, lr: 5.16e-03, grad_scale: 32.0
2024-09-24 04:07:56,231 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.33 vs. limit=15.0
2024-09-24 04:08:23,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=417074.0, ans=0.1
2024-09-24 04:09:11,666 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.268e+02 1.365e+02 1.526e+02 2.043e+02, threshold=2.731e+02, percent-clipped=0.0
2024-09-24 04:09:13,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=417260.6666666667, ans=0.125
2024-09-24 04:09:14,879 INFO [train.py:1198] (1/4) Epoch 23, batch 3700, loss[loss=0.2134, ctc_loss=0.1415, cr_loss=0.3592, over 16018.00 frames. ], tot_loss[loss=0.2079, ctc_loss=0.1373, cr_loss=0.3531, over 3365213.76 frames. ], batch size: 74, lr: 5.16e-03, grad_scale: 32.0
2024-09-24 04:09:18,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=417260.6666666667, ans=0.0
2024-09-24 04:09:18,762 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.31 vs. limit=22.5
2024-09-24 04:09:26,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=417260.6666666667, ans=0.125
2024-09-24 04:09:44,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=417354.0, ans=0.125
2024-09-24 04:10:13,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.58 vs. limit=15.0
2024-09-24 04:10:33,639 INFO [train.py:1198] (1/4) Epoch 23, batch 3750, loss[loss=0.1888, ctc_loss=0.1203, cr_loss=0.3424, over 17254.00 frames. ], tot_loss[loss=0.2095, ctc_loss=0.1384, cr_loss=0.3551, over 3340110.49 frames. ], batch size: 44, lr: 5.16e-03, grad_scale: 32.0
2024-09-24 04:10:38,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=417494.0, ans=0.1
2024-09-24 04:10:55,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=417540.6666666667, ans=0.125
2024-09-24 04:10:57,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=417540.6666666667, ans=0.125
2024-09-24 04:11:05,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=417587.3333333333, ans=0.1
2024-09-24 04:11:15,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=417587.3333333333, ans=0.2
2024-09-24 04:11:22,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=417634.0, ans=0.09899494936611666
2024-09-24 04:11:49,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=417680.6666666667, ans=0.125
2024-09-24 04:11:50,428 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.304e+02 1.374e+02 1.461e+02 1.993e+02, threshold=2.748e+02, percent-clipped=0.0
2024-09-24 04:11:50,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=417680.6666666667, ans=0.0
2024-09-24 04:11:53,566 INFO [train.py:1198] (1/4) Epoch 23, batch 3800, loss[loss=0.196, ctc_loss=0.1304, cr_loss=0.3283, over 17309.00 frames. ], tot_loss[loss=0.2111, ctc_loss=0.1397, cr_loss=0.3569, over 3327649.51 frames. ], batch size: 51, lr: 5.15e-03, grad_scale: 32.0
2024-09-24 04:12:33,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=417820.6666666667, ans=0.125
2024-09-24 04:12:41,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=417867.3333333333, ans=0.1
2024-09-24 04:12:47,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=417867.3333333333, ans=0.0
2024-09-24 04:12:55,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=417914.0, ans=0.125
2024-09-24 04:13:07,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=417914.0, ans=0.0
2024-09-24 04:13:07,955 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.80 vs. limit=12.0
2024-09-24 04:13:09,450 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 04:13:12,142 INFO [train.py:1198] (1/4) Epoch 23, batch 3850, loss[loss=0.2622, ctc_loss=0.1844, cr_loss=0.3894, over 11882.00 frames. ], tot_loss[loss=0.2134, ctc_loss=0.1415, cr_loss=0.3594, over 3286520.04 frames. ], batch size: 123, lr: 5.15e-03, grad_scale: 32.0
2024-09-24 04:13:38,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=418007.3333333333, ans=0.2
2024-09-24 04:13:57,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=418100.6666666667, ans=0.2
2024-09-24 04:13:57,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=418100.6666666667, ans=0.125
2024-09-24 04:14:02,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=418100.6666666667, ans=0.025
2024-09-24 04:15:15,635 INFO [train.py:1198] (1/4) Epoch 24, batch 0, loss[loss=0.2041, ctc_loss=0.1342, cr_loss=0.3493, over 16961.00 frames. ], tot_loss[loss=0.2041, ctc_loss=0.1342, cr_loss=0.3493, over 16961.00 frames. ], batch size: 42, lr: 5.04e-03, grad_scale: 32.0
2024-09-24 04:15:15,635 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 04:15:28,593 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.0.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.6141, 4.0472, 4.4768, 4.3162], device='cuda:1')
2024-09-24 04:15:33,343 INFO [train.py:1230] (1/4) Epoch 24, validation: loss=0.03789, ctc_loss=0.03789, cr_loss=8.011e-15, over 944034.00 frames. 
2024-09-24 04:15:33,344 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 04:15:36,593 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.389e+02 1.528e+02 1.642e+02 3.495e+02, threshold=3.056e+02, percent-clipped=0.0
2024-09-24 04:16:26,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=418315.3333333333, ans=0.0
2024-09-24 04:16:29,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=418315.3333333333, ans=0.0
2024-09-24 04:16:44,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=418362.0, ans=0.2
2024-09-24 04:16:53,367 INFO [train.py:1198] (1/4) Epoch 24, batch 50, loss[loss=0.1687, ctc_loss=0.1111, cr_loss=0.2881, over 16962.00 frames. ], tot_loss[loss=0.2127, ctc_loss=0.141, cr_loss=0.3583, over 754073.70 frames. ], batch size: 42, lr: 5.04e-03, grad_scale: 32.0
2024-09-24 04:17:26,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=418502.0, ans=0.0
2024-09-24 04:17:32,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=418502.0, ans=0.125
2024-09-24 04:18:03,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=418595.3333333333, ans=0.125
2024-09-24 04:18:15,655 INFO [train.py:1198] (1/4) Epoch 24, batch 100, loss[loss=0.2385, ctc_loss=0.1579, cr_loss=0.403, over 17096.00 frames. ], tot_loss[loss=0.2128, ctc_loss=0.1409, cr_loss=0.3595, over 1324245.74 frames. ], batch size: 49, lr: 5.04e-03, grad_scale: 32.0
2024-09-24 04:18:18,831 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.264e+02 1.318e+02 1.445e+02 2.140e+02, threshold=2.636e+02, percent-clipped=1.0
2024-09-24 04:18:33,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=418688.6666666667, ans=0.05
2024-09-24 04:18:36,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=418688.6666666667, ans=0.125
2024-09-24 04:19:01,174 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.98 vs. limit=6.0
2024-09-24 04:19:38,292 INFO [train.py:1198] (1/4) Epoch 24, batch 150, loss[loss=0.2167, ctc_loss=0.1419, cr_loss=0.3736, over 17071.00 frames. ], tot_loss[loss=0.2121, ctc_loss=0.1404, cr_loss=0.3588, over 1770601.17 frames. ], batch size: 43, lr: 5.04e-03, grad_scale: 32.0
2024-09-24 04:19:46,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=418875.3333333333, ans=0.125
2024-09-24 04:19:53,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=418922.0, ans=0.025
2024-09-24 04:19:56,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=418922.0, ans=0.1
2024-09-24 04:19:57,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=418922.0, ans=0.025
2024-09-24 04:20:02,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=418922.0, ans=0.05
2024-09-24 04:20:10,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=418968.6666666667, ans=0.125
2024-09-24 04:20:57,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=419062.0, ans=0.0
2024-09-24 04:21:03,915 INFO [train.py:1198] (1/4) Epoch 24, batch 200, loss[loss=0.1931, ctc_loss=0.1258, cr_loss=0.3365, over 16701.00 frames. ], tot_loss[loss=0.2109, ctc_loss=0.1395, cr_loss=0.3572, over 2124190.73 frames. ], batch size: 37, lr: 5.04e-03, grad_scale: 32.0
2024-09-24 04:21:07,012 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.247e+02 1.339e+02 1.438e+02 1.930e+02, threshold=2.677e+02, percent-clipped=0.0
2024-09-24 04:21:09,617 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.00 vs. limit=15.0
2024-09-24 04:21:12,552 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.12 vs. limit=12.0
2024-09-24 04:21:15,891 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.14 vs. limit=15.0
2024-09-24 04:21:20,742 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.90 vs. limit=15.0
2024-09-24 04:21:34,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=419202.0, ans=0.125
2024-09-24 04:21:36,782 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.61 vs. limit=15.0
2024-09-24 04:21:44,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=419202.0, ans=0.0
2024-09-24 04:21:45,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=419202.0, ans=0.125
2024-09-24 04:22:06,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=419248.6666666667, ans=0.125
2024-09-24 04:22:26,694 INFO [train.py:1198] (1/4) Epoch 24, batch 250, loss[loss=0.1876, ctc_loss=0.1186, cr_loss=0.3446, over 17271.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1384, cr_loss=0.356, over 2398179.64 frames. ], batch size: 42, lr: 5.03e-03, grad_scale: 32.0
2024-09-24 04:22:36,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=419342.0, ans=0.0
2024-09-24 04:22:41,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=419388.6666666667, ans=0.125
2024-09-24 04:23:00,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=419435.3333333333, ans=0.125
2024-09-24 04:23:11,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=419435.3333333333, ans=0.0
2024-09-24 04:23:19,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=419482.0, ans=0.0
2024-09-24 04:23:46,170 INFO [train.py:1198] (1/4) Epoch 24, batch 300, loss[loss=0.1705, ctc_loss=0.1083, cr_loss=0.3109, over 15834.00 frames. ], tot_loss[loss=0.209, ctc_loss=0.138, cr_loss=0.355, over 2611807.62 frames. ], batch size: 35, lr: 5.03e-03, grad_scale: 32.0
2024-09-24 04:23:49,253 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.252e+02 1.369e+02 1.490e+02 2.368e+02, threshold=2.737e+02, percent-clipped=0.0
2024-09-24 04:23:52,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=419575.3333333333, ans=0.2
2024-09-24 04:23:54,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=419575.3333333333, ans=0.125
2024-09-24 04:24:03,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=419622.0, ans=0.125
2024-09-24 04:24:10,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=419622.0, ans=0.0
2024-09-24 04:24:16,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=419668.6666666667, ans=0.1
2024-09-24 04:25:08,436 INFO [train.py:1198] (1/4) Epoch 24, batch 350, loss[loss=0.2198, ctc_loss=0.1452, cr_loss=0.3731, over 17152.00 frames. ], tot_loss[loss=0.2086, ctc_loss=0.1376, cr_loss=0.3547, over 2783941.19 frames. ], batch size: 45, lr: 5.03e-03, grad_scale: 32.0
2024-09-24 04:25:21,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=419808.6666666667, ans=0.1
2024-09-24 04:25:34,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=419855.3333333333, ans=0.125
2024-09-24 04:25:53,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=419902.0, ans=0.125
2024-09-24 04:26:06,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=419948.6666666667, ans=0.0
2024-09-24 04:26:16,698 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.19 vs. limit=15.0
2024-09-24 04:26:34,306 INFO [train.py:1198] (1/4) Epoch 24, batch 400, loss[loss=0.2305, ctc_loss=0.1506, cr_loss=0.3995, over 17229.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.1373, cr_loss=0.3535, over 2908969.49 frames. ], batch size: 50, lr: 5.03e-03, grad_scale: 32.0
2024-09-24 04:26:37,589 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.274e+02 1.350e+02 1.482e+02 1.874e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-24 04:26:38,100 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=2.632e-03
2024-09-24 04:26:55,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=420088.6666666667, ans=0.1
2024-09-24 04:27:04,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.64 vs. limit=10.0
2024-09-24 04:27:11,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=420135.3333333333, ans=0.1
2024-09-24 04:27:20,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=420135.3333333333, ans=0.125
2024-09-24 04:27:27,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=420182.0, ans=0.1
2024-09-24 04:27:49,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=420228.6666666667, ans=0.125
2024-09-24 04:27:55,075 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.37 vs. limit=6.0
2024-09-24 04:27:57,668 INFO [train.py:1198] (1/4) Epoch 24, batch 450, loss[loss=0.2362, ctc_loss=0.1596, cr_loss=0.3827, over 17008.00 frames. ], tot_loss[loss=0.2094, ctc_loss=0.1383, cr_loss=0.3553, over 3007723.04 frames. ], batch size: 53, lr: 5.03e-03, grad_scale: 32.0
2024-09-24 04:28:17,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=420322.0, ans=0.0
2024-09-24 04:28:34,271 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.35 vs. limit=15.0
2024-09-24 04:29:13,601 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.65 vs. limit=15.0
2024-09-24 04:29:14,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=420462.0, ans=0.125
2024-09-24 04:29:17,789 INFO [train.py:1198] (1/4) Epoch 24, batch 500, loss[loss=0.1841, ctc_loss=0.1224, cr_loss=0.3086, over 17200.00 frames. ], tot_loss[loss=0.2093, ctc_loss=0.1383, cr_loss=0.3553, over 3092212.87 frames. ], batch size: 41, lr: 5.03e-03, grad_scale: 32.0
2024-09-24 04:29:21,053 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.236e+02 1.308e+02 1.375e+02 2.594e+02, threshold=2.616e+02, percent-clipped=0.0
2024-09-24 04:29:30,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=420508.6666666667, ans=0.0
2024-09-24 04:29:48,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_abs, batch_count=420555.3333333333, ans=0.5
2024-09-24 04:29:59,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=420602.0, ans=0.125
2024-09-24 04:30:02,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=420602.0, ans=0.0
2024-09-24 04:30:03,002 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.87 vs. limit=10.0
2024-09-24 04:30:39,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=420695.3333333333, ans=0.0
2024-09-24 04:30:45,504 INFO [train.py:1198] (1/4) Epoch 24, batch 550, loss[loss=0.2177, ctc_loss=0.1421, cr_loss=0.3776, over 16493.00 frames. ], tot_loss[loss=0.2087, ctc_loss=0.1378, cr_loss=0.3547, over 3149523.40 frames. ], batch size: 66, lr: 5.03e-03, grad_scale: 16.0
2024-09-24 04:31:09,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=420788.6666666667, ans=0.1
2024-09-24 04:31:14,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=420788.6666666667, ans=0.2
2024-09-24 04:31:45,363 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=9.34 vs. limit=15.0
2024-09-24 04:32:04,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=420928.6666666667, ans=0.0
2024-09-24 04:32:08,521 INFO [train.py:1198] (1/4) Epoch 24, batch 600, loss[loss=0.2259, ctc_loss=0.1511, cr_loss=0.3742, over 16049.00 frames. ], tot_loss[loss=0.2104, ctc_loss=0.139, cr_loss=0.3568, over 3193262.86 frames. ], batch size: 74, lr: 5.02e-03, grad_scale: 16.0
2024-09-24 04:32:13,195 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.249e+02 1.314e+02 1.426e+02 3.030e+02, threshold=2.628e+02, percent-clipped=1.0
2024-09-24 04:33:02,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=421115.3333333333, ans=0.125
2024-09-24 04:33:07,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=421115.3333333333, ans=0.2
2024-09-24 04:33:17,969 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.76 vs. limit=22.5
2024-09-24 04:33:27,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=421208.6666666667, ans=0.025
2024-09-24 04:33:28,573 INFO [train.py:1198] (1/4) Epoch 24, batch 650, loss[loss=0.1986, ctc_loss=0.1306, cr_loss=0.34, over 17030.00 frames. ], tot_loss[loss=0.2101, ctc_loss=0.1388, cr_loss=0.3563, over 3235397.40 frames. ], batch size: 44, lr: 5.02e-03, grad_scale: 16.0
2024-09-24 04:33:35,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=421208.6666666667, ans=0.125
2024-09-24 04:33:43,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=421255.3333333333, ans=0.125
2024-09-24 04:33:48,254 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 04:34:04,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=421302.0, ans=0.125
2024-09-24 04:34:29,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=421348.6666666667, ans=0.125
2024-09-24 04:34:50,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=421442.0, ans=0.1
2024-09-24 04:34:51,510 INFO [train.py:1198] (1/4) Epoch 24, batch 700, loss[loss=0.2162, ctc_loss=0.143, cr_loss=0.3658, over 17210.00 frames. ], tot_loss[loss=0.2092, ctc_loss=0.1381, cr_loss=0.3552, over 3268143.10 frames. ], batch size: 47, lr: 5.02e-03, grad_scale: 16.0
2024-09-24 04:34:53,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=421442.0, ans=0.0
2024-09-24 04:34:56,334 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.315e+02 1.415e+02 1.543e+02 2.275e+02, threshold=2.830e+02, percent-clipped=0.0
2024-09-24 04:34:58,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=421442.0, ans=0.1
2024-09-24 04:34:59,112 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.90 vs. limit=10.0
2024-09-24 04:35:07,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=421488.6666666667, ans=0.0
2024-09-24 04:35:20,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=421488.6666666667, ans=0.125
2024-09-24 04:35:25,741 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.83 vs. limit=10.0
2024-09-24 04:36:14,457 INFO [train.py:1198] (1/4) Epoch 24, batch 750, loss[loss=0.1949, ctc_loss=0.1291, cr_loss=0.3287, over 17294.00 frames. ], tot_loss[loss=0.2101, ctc_loss=0.1389, cr_loss=0.356, over 3292282.20 frames. ], batch size: 46, lr: 5.02e-03, grad_scale: 16.0
2024-09-24 04:36:42,641 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.46 vs. limit=15.0
2024-09-24 04:37:10,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=421815.3333333333, ans=0.2
2024-09-24 04:37:36,697 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.87 vs. limit=15.0
2024-09-24 04:37:37,164 INFO [train.py:1198] (1/4) Epoch 24, batch 800, loss[loss=0.2329, ctc_loss=0.1547, cr_loss=0.3909, over 16983.00 frames. ], tot_loss[loss=0.2109, ctc_loss=0.1394, cr_loss=0.3573, over 3312001.57 frames. ], batch size: 53, lr: 5.02e-03, grad_scale: 32.0
2024-09-24 04:37:38,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=421908.6666666667, ans=0.0
2024-09-24 04:37:41,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.56 vs. limit=22.5
2024-09-24 04:37:41,996 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.296e+02 1.355e+02 1.500e+02 2.153e+02, threshold=2.711e+02, percent-clipped=0.0
2024-09-24 04:37:42,781 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.57 vs. limit=15.0
2024-09-24 04:38:57,342 INFO [train.py:1198] (1/4) Epoch 24, batch 850, loss[loss=0.1984, ctc_loss=0.1298, cr_loss=0.3427, over 17149.00 frames. ], tot_loss[loss=0.2105, ctc_loss=0.1391, cr_loss=0.3569, over 3332581.73 frames. ], batch size: 45, lr: 5.02e-03, grad_scale: 32.0
2024-09-24 04:39:12,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.11 vs. limit=8.0
2024-09-24 04:39:15,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=422188.6666666667, ans=0.125
2024-09-24 04:39:28,825 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 04:39:33,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=422235.3333333333, ans=0.0
2024-09-24 04:39:43,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=422235.3333333333, ans=0.025
2024-09-24 04:39:49,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=422282.0, ans=0.125
2024-09-24 04:40:25,045 INFO [train.py:1198] (1/4) Epoch 24, batch 900, loss[loss=0.2451, ctc_loss=0.1635, cr_loss=0.4081, over 16807.00 frames. ], tot_loss[loss=0.2101, ctc_loss=0.1388, cr_loss=0.3564, over 3339722.95 frames. ], batch size: 61, lr: 5.02e-03, grad_scale: 32.0
2024-09-24 04:40:29,775 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.271e+02 1.382e+02 1.510e+02 2.333e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-24 04:40:48,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.11 vs. limit=22.5
2024-09-24 04:40:50,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=422422.0, ans=0.0
2024-09-24 04:41:15,269 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.88 vs. limit=10.0
2024-09-24 04:41:21,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=422515.3333333333, ans=0.125
2024-09-24 04:41:43,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=422608.6666666667, ans=0.1
2024-09-24 04:41:45,282 INFO [train.py:1198] (1/4) Epoch 24, batch 950, loss[loss=0.2149, ctc_loss=0.1413, cr_loss=0.3679, over 17357.00 frames. ], tot_loss[loss=0.2086, ctc_loss=0.1377, cr_loss=0.3543, over 3349559.66 frames. ], batch size: 48, lr: 5.01e-03, grad_scale: 32.0
2024-09-24 04:41:48,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=422608.6666666667, ans=0.0
2024-09-24 04:42:00,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=6.52 vs. limit=12.0
2024-09-24 04:42:12,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=422655.3333333333, ans=0.2
2024-09-24 04:42:31,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=422702.0, ans=0.125
2024-09-24 04:43:08,571 INFO [train.py:1198] (1/4) Epoch 24, batch 1000, loss[loss=0.243, ctc_loss=0.1647, cr_loss=0.3914, over 16772.00 frames. ], tot_loss[loss=0.2088, ctc_loss=0.1379, cr_loss=0.3543, over 3338018.35 frames. ], batch size: 61, lr: 5.01e-03, grad_scale: 32.0
2024-09-24 04:43:08,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=422842.0, ans=0.125
2024-09-24 04:43:13,212 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.311e+02 1.412e+02 1.541e+02 1.926e+02, threshold=2.824e+02, percent-clipped=0.0
2024-09-24 04:43:19,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=422842.0, ans=0.0
2024-09-24 04:43:33,163 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.73 vs. limit=15.0
2024-09-24 04:43:46,453 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.52 vs. limit=15.0
2024-09-24 04:43:58,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=422982.0, ans=0.5
2024-09-24 04:44:28,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=423028.6666666667, ans=0.0
2024-09-24 04:44:30,826 INFO [train.py:1198] (1/4) Epoch 24, batch 1050, loss[loss=0.2, ctc_loss=0.1329, cr_loss=0.3356, over 17146.00 frames. ], tot_loss[loss=0.2083, ctc_loss=0.1375, cr_loss=0.354, over 3350198.61 frames. ], batch size: 48, lr: 5.01e-03, grad_scale: 32.0
2024-09-24 04:44:55,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=423122.0, ans=0.125
2024-09-24 04:44:55,983 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.65 vs. limit=15.0
2024-09-24 04:45:24,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=423215.3333333333, ans=0.2
2024-09-24 04:45:26,830 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.56 vs. limit=6.0
2024-09-24 04:45:56,523 INFO [train.py:1198] (1/4) Epoch 24, batch 1100, loss[loss=0.2063, ctc_loss=0.1339, cr_loss=0.3622, over 17016.00 frames. ], tot_loss[loss=0.2091, ctc_loss=0.138, cr_loss=0.3555, over 3349525.04 frames. ], batch size: 44, lr: 5.01e-03, grad_scale: 32.0
2024-09-24 04:46:01,273 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.919e+01 1.239e+02 1.343e+02 1.468e+02 1.769e+02, threshold=2.686e+02, percent-clipped=0.0
2024-09-24 04:46:14,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=423355.3333333333, ans=0.0
2024-09-24 04:46:14,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.88 vs. limit=6.0
2024-09-24 04:46:24,259 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.69 vs. limit=12.0
2024-09-24 04:46:24,314 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=13.76 vs. limit=15.0
2024-09-24 04:46:58,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=423448.6666666667, ans=0.125
2024-09-24 04:47:09,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=423495.3333333333, ans=0.125
2024-09-24 04:47:13,148 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=11.28 vs. limit=22.5
2024-09-24 04:47:18,698 INFO [train.py:1198] (1/4) Epoch 24, batch 1150, loss[loss=0.2229, ctc_loss=0.1468, cr_loss=0.3806, over 17318.00 frames. ], tot_loss[loss=0.2091, ctc_loss=0.138, cr_loss=0.3555, over 3355779.22 frames. ], batch size: 51, lr: 5.01e-03, grad_scale: 32.0
2024-09-24 04:47:34,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=423588.6666666667, ans=0.2
2024-09-24 04:47:36,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=423588.6666666667, ans=0.0
2024-09-24 04:47:41,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=423588.6666666667, ans=0.125
2024-09-24 04:47:54,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.34 vs. limit=6.0
2024-09-24 04:48:20,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.69 vs. limit=15.0
2024-09-24 04:48:29,064 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.38 vs. limit=22.5
2024-09-24 04:48:39,272 INFO [train.py:1198] (1/4) Epoch 24, batch 1200, loss[loss=0.1971, ctc_loss=0.1268, cr_loss=0.3514, over 17315.00 frames. ], tot_loss[loss=0.2095, ctc_loss=0.1383, cr_loss=0.356, over 3360006.42 frames. ], batch size: 49, lr: 5.01e-03, grad_scale: 32.0
2024-09-24 04:48:44,026 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.248e+02 1.325e+02 1.416e+02 2.562e+02, threshold=2.650e+02, percent-clipped=0.0
2024-09-24 04:48:44,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=423775.3333333333, ans=0.2
2024-09-24 04:48:57,821 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.43 vs. limit=22.5
2024-09-24 04:49:00,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=423822.0, ans=0.0
2024-09-24 04:49:20,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=423868.6666666667, ans=0.04949747468305833
2024-09-24 04:49:30,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=423915.3333333333, ans=0.2
2024-09-24 04:49:39,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=423915.3333333333, ans=0.125
2024-09-24 04:49:45,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=423962.0, ans=0.09899494936611666
2024-09-24 04:50:06,367 INFO [train.py:1198] (1/4) Epoch 24, batch 1250, loss[loss=0.2352, ctc_loss=0.1593, cr_loss=0.3793, over 16736.00 frames. ], tot_loss[loss=0.21, ctc_loss=0.1388, cr_loss=0.356, over 3346500.95 frames. ], batch size: 61, lr: 5.01e-03, grad_scale: 32.0
2024-09-24 04:50:15,454 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.68 vs. limit=12.0
2024-09-24 04:50:41,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=424102.0, ans=0.125
2024-09-24 04:50:41,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=424102.0, ans=0.125
2024-09-24 04:50:42,044 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 04:51:05,081 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.24 vs. limit=15.0
2024-09-24 04:51:07,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=424148.6666666667, ans=0.0
2024-09-24 04:51:12,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=424195.3333333333, ans=0.1
2024-09-24 04:51:14,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=424195.3333333333, ans=0.125
2024-09-24 04:51:20,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=424195.3333333333, ans=0.2
2024-09-24 04:51:26,792 INFO [train.py:1198] (1/4) Epoch 24, batch 1300, loss[loss=0.1882, ctc_loss=0.1249, cr_loss=0.3167, over 17146.00 frames. ], tot_loss[loss=0.2096, ctc_loss=0.1385, cr_loss=0.3556, over 3359929.20 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 04:51:31,562 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.254e+02 1.331e+02 1.449e+02 1.850e+02, threshold=2.662e+02, percent-clipped=0.0
2024-09-24 04:52:49,352 INFO [train.py:1198] (1/4) Epoch 24, batch 1350, loss[loss=0.2324, ctc_loss=0.1559, cr_loss=0.3828, over 16550.00 frames. ], tot_loss[loss=0.2093, ctc_loss=0.1382, cr_loss=0.3552, over 3363877.42 frames. ], batch size: 66, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 04:52:54,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=424475.3333333333, ans=0.07
2024-09-24 04:53:13,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=424522.0, ans=0.09899494936611666
2024-09-24 04:53:23,737 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.96 vs. limit=22.5
2024-09-24 04:53:26,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=424568.6666666667, ans=0.0
2024-09-24 04:53:43,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=424615.3333333333, ans=0.125
2024-09-24 04:53:45,757 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.69 vs. limit=15.0
2024-09-24 04:54:11,955 INFO [train.py:1198] (1/4) Epoch 24, batch 1400, loss[loss=0.2163, ctc_loss=0.1421, cr_loss=0.371, over 17129.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1384, cr_loss=0.3565, over 3369077.67 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 04:54:13,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=424708.6666666667, ans=0.0
2024-09-24 04:54:16,812 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.257e+02 1.358e+02 1.495e+02 1.831e+02, threshold=2.716e+02, percent-clipped=0.0
2024-09-24 04:54:23,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=424708.6666666667, ans=0.0
2024-09-24 04:54:25,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=424708.6666666667, ans=0.125
2024-09-24 04:54:28,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.32 vs. limit=10.0
2024-09-24 04:54:51,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=424802.0, ans=0.05
2024-09-24 04:55:02,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=424802.0, ans=0.025
2024-09-24 04:55:37,530 INFO [train.py:1198] (1/4) Epoch 24, batch 1450, loss[loss=0.2016, ctc_loss=0.1341, cr_loss=0.3379, over 17211.00 frames. ], tot_loss[loss=0.2088, ctc_loss=0.1378, cr_loss=0.3548, over 3362818.49 frames. ], batch size: 50, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 04:55:52,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=424988.6666666667, ans=0.125
2024-09-24 04:55:52,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=424988.6666666667, ans=0.125
2024-09-24 04:55:52,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=424988.6666666667, ans=0.125
2024-09-24 04:56:14,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=425035.3333333333, ans=0.125
2024-09-24 04:56:18,449 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.59 vs. limit=6.0
2024-09-24 04:56:37,230 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.39 vs. limit=12.0
2024-09-24 04:56:41,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=425128.6666666667, ans=0.1
2024-09-24 04:56:53,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=425128.6666666667, ans=0.025
2024-09-24 04:56:59,699 INFO [train.py:1198] (1/4) Epoch 24, batch 1500, loss[loss=0.2141, ctc_loss=0.1431, cr_loss=0.3551, over 17211.00 frames. ], tot_loss[loss=0.2086, ctc_loss=0.1377, cr_loss=0.3548, over 3361344.76 frames. ], batch size: 50, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 04:57:03,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=425175.3333333333, ans=0.0
2024-09-24 04:57:04,517 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.246e+02 1.334e+02 1.449e+02 2.075e+02, threshold=2.667e+02, percent-clipped=0.0
2024-09-24 04:57:22,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=425222.0, ans=0.125
2024-09-24 04:57:23,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=425222.0, ans=0.1
2024-09-24 04:57:28,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=425222.0, ans=0.0
2024-09-24 04:57:35,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=425268.6666666667, ans=10.0
2024-09-24 04:57:38,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=425268.6666666667, ans=0.05
2024-09-24 04:58:19,809 INFO [train.py:1198] (1/4) Epoch 24, batch 1550, loss[loss=0.2039, ctc_loss=0.1344, cr_loss=0.3476, over 17026.00 frames. ], tot_loss[loss=0.2073, ctc_loss=0.1367, cr_loss=0.3529, over 3369667.00 frames. ], batch size: 44, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 04:58:40,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=425455.3333333333, ans=0.125
2024-09-24 04:58:52,291 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=425502.0, ans=0.125
2024-09-24 04:59:11,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=425548.6666666667, ans=0.2
2024-09-24 04:59:19,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=425548.6666666667, ans=0.125
2024-09-24 04:59:42,276 INFO [train.py:1198] (1/4) Epoch 24, batch 1600, loss[loss=0.1908, ctc_loss=0.1239, cr_loss=0.3347, over 17200.00 frames. ], tot_loss[loss=0.2075, ctc_loss=0.1368, cr_loss=0.3535, over 3371092.60 frames. ], batch size: 47, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 04:59:47,003 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.230e+02 1.386e+02 1.499e+02 2.034e+02, threshold=2.773e+02, percent-clipped=0.0
2024-09-24 05:00:32,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=425735.3333333333, ans=0.0
2024-09-24 05:00:43,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=425782.0, ans=0.2
2024-09-24 05:00:44,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=425782.0, ans=0.0
2024-09-24 05:01:07,191 INFO [train.py:1198] (1/4) Epoch 24, batch 1650, loss[loss=0.2219, ctc_loss=0.1475, cr_loss=0.3717, over 17341.00 frames. ], tot_loss[loss=0.2087, ctc_loss=0.1377, cr_loss=0.3549, over 3364069.75 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 32.0
2024-09-24 05:01:20,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=425875.3333333333, ans=0.1
2024-09-24 05:01:38,271 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.08 vs. limit=15.0
2024-09-24 05:01:47,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=425968.6666666667, ans=0.0
2024-09-24 05:02:09,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=426015.3333333333, ans=0.125
2024-09-24 05:02:18,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=426062.0, ans=0.125
2024-09-24 05:02:29,580 INFO [train.py:1198] (1/4) Epoch 24, batch 1700, loss[loss=0.1817, ctc_loss=0.1189, cr_loss=0.314, over 17100.00 frames. ], tot_loss[loss=0.2082, ctc_loss=0.1374, cr_loss=0.3542, over 3367723.73 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 32.0
2024-09-24 05:02:34,431 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.248e+02 1.319e+02 1.421e+02 3.276e+02, threshold=2.637e+02, percent-clipped=2.0
2024-09-24 05:02:49,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=426155.3333333333, ans=0.2
2024-09-24 05:02:58,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=426155.3333333333, ans=0.1
2024-09-24 05:03:07,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=10.62 vs. limit=15.0
2024-09-24 05:03:23,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=426248.6666666667, ans=0.1
2024-09-24 05:03:27,342 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=15.60 vs. limit=22.5
2024-09-24 05:03:50,635 INFO [train.py:1198] (1/4) Epoch 24, batch 1750, loss[loss=0.2377, ctc_loss=0.157, cr_loss=0.4038, over 16554.00 frames. ], tot_loss[loss=0.2078, ctc_loss=0.1372, cr_loss=0.3532, over 3367236.21 frames. ], batch size: 66, lr: 4.99e-03, grad_scale: 32.0
2024-09-24 05:03:57,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=426342.0, ans=0.0
2024-09-24 05:04:02,746 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.41 vs. limit=15.0
2024-09-24 05:04:24,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=426435.3333333333, ans=0.2
2024-09-24 05:05:17,716 INFO [train.py:1198] (1/4) Epoch 24, batch 1800, loss[loss=0.2009, ctc_loss=0.1311, cr_loss=0.349, over 17057.00 frames. ], tot_loss[loss=0.2078, ctc_loss=0.137, cr_loss=0.3536, over 3363130.17 frames. ], batch size: 46, lr: 4.99e-03, grad_scale: 32.0
2024-09-24 05:05:22,469 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.235e+02 1.322e+02 1.422e+02 1.827e+02, threshold=2.643e+02, percent-clipped=0.0
2024-09-24 05:05:27,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=426575.3333333333, ans=0.5
2024-09-24 05:05:53,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=426668.6666666667, ans=0.125
2024-09-24 05:05:57,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=426668.6666666667, ans=0.125
2024-09-24 05:05:57,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=426668.6666666667, ans=0.125
2024-09-24 05:05:59,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=426668.6666666667, ans=0.0
2024-09-24 05:06:20,877 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.49 vs. limit=15.0
2024-09-24 05:06:31,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=426762.0, ans=0.125
2024-09-24 05:06:37,695 INFO [train.py:1198] (1/4) Epoch 24, batch 1850, loss[loss=0.2391, ctc_loss=0.159, cr_loss=0.4005, over 17035.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1363, cr_loss=0.3527, over 3373209.33 frames. ], batch size: 52, lr: 4.99e-03, grad_scale: 32.0
2024-09-24 05:06:55,764 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.10 vs. limit=15.0
2024-09-24 05:07:13,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=426902.0, ans=0.125
2024-09-24 05:07:31,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=426948.6666666667, ans=0.1
2024-09-24 05:07:32,420 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.19 vs. limit=22.5
2024-09-24 05:07:33,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=426948.6666666667, ans=0.035
2024-09-24 05:07:57,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=426995.3333333333, ans=0.0
2024-09-24 05:08:00,018 INFO [train.py:1198] (1/4) Epoch 24, batch 1900, loss[loss=0.2051, ctc_loss=0.1357, cr_loss=0.3471, over 17292.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1363, cr_loss=0.3524, over 3375326.68 frames. ], batch size: 49, lr: 4.99e-03, grad_scale: 16.0
2024-09-24 05:08:05,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=427042.0, ans=0.125
2024-09-24 05:08:06,244 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.237e+02 1.306e+02 1.387e+02 1.778e+02, threshold=2.611e+02, percent-clipped=0.0
2024-09-24 05:08:48,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=427182.0, ans=0.125
2024-09-24 05:09:22,692 INFO [train.py:1198] (1/4) Epoch 24, batch 1950, loss[loss=0.2508, ctc_loss=0.1752, cr_loss=0.3782, over 15253.00 frames. ], tot_loss[loss=0.2085, ctc_loss=0.1377, cr_loss=0.3542, over 3366222.12 frames. ], batch size: 89, lr: 4.99e-03, grad_scale: 16.0
2024-09-24 05:09:50,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=427322.0, ans=0.125
2024-09-24 05:10:08,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=427368.6666666667, ans=0.125
2024-09-24 05:10:36,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=427462.0, ans=0.125
2024-09-24 05:10:47,724 INFO [train.py:1198] (1/4) Epoch 24, batch 2000, loss[loss=0.2066, ctc_loss=0.1347, cr_loss=0.3598, over 17216.00 frames. ], tot_loss[loss=0.2089, ctc_loss=0.1379, cr_loss=0.3549, over 3362293.20 frames. ], batch size: 50, lr: 4.99e-03, grad_scale: 16.0
2024-09-24 05:10:51,467 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 05:10:55,710 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.252e+02 1.338e+02 1.434e+02 1.849e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-24 05:11:00,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=427508.6666666667, ans=0.1
2024-09-24 05:11:02,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=427555.3333333333, ans=0.2
2024-09-24 05:11:10,732 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.38 vs. limit=15.0
2024-09-24 05:11:13,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=427555.3333333333, ans=0.125
2024-09-24 05:11:50,644 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.48 vs. limit=15.0
2024-09-24 05:11:53,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=427695.3333333333, ans=0.2
2024-09-24 05:12:09,511 INFO [train.py:1198] (1/4) Epoch 24, batch 2050, loss[loss=0.2395, ctc_loss=0.1637, cr_loss=0.379, over 14867.00 frames. ], tot_loss[loss=0.2099, ctc_loss=0.1387, cr_loss=0.3562, over 3357692.96 frames. ], batch size: 89, lr: 4.98e-03, grad_scale: 16.0
2024-09-24 05:12:10,780 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.87 vs. limit=5.0
2024-09-24 05:12:40,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=427835.3333333333, ans=0.1
2024-09-24 05:12:57,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=427882.0, ans=0.125
2024-09-24 05:12:57,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=427882.0, ans=0.0
2024-09-24 05:13:01,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=427882.0, ans=0.125
2024-09-24 05:13:01,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=427882.0, ans=0.0
2024-09-24 05:13:29,441 INFO [train.py:1198] (1/4) Epoch 24, batch 2100, loss[loss=0.2358, ctc_loss=0.1545, cr_loss=0.4065, over 16970.00 frames. ], tot_loss[loss=0.2118, ctc_loss=0.1399, cr_loss=0.3592, over 3355840.02 frames. ], batch size: 53, lr: 4.98e-03, grad_scale: 16.0
2024-09-24 05:13:31,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=427975.3333333333, ans=0.1
2024-09-24 05:13:32,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=427975.3333333333, ans=0.125
2024-09-24 05:13:37,476 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.278e+02 1.335e+02 1.481e+02 2.167e+02, threshold=2.670e+02, percent-clipped=0.0
2024-09-24 05:13:47,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=428022.0, ans=0.0
2024-09-24 05:13:47,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=428022.0, ans=0.0
2024-09-24 05:14:19,258 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.95 vs. limit=15.0
2024-09-24 05:14:33,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=428115.3333333333, ans=0.0
2024-09-24 05:14:54,809 INFO [train.py:1198] (1/4) Epoch 24, batch 2150, loss[loss=0.2176, ctc_loss=0.143, cr_loss=0.3733, over 17299.00 frames. ], tot_loss[loss=0.2108, ctc_loss=0.1392, cr_loss=0.3581, over 3363412.89 frames. ], batch size: 49, lr: 4.98e-03, grad_scale: 16.0
2024-09-24 05:15:13,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=428255.3333333333, ans=0.0
2024-09-24 05:15:50,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=428348.6666666667, ans=0.0
2024-09-24 05:15:56,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=428348.6666666667, ans=0.1
2024-09-24 05:16:17,768 INFO [train.py:1198] (1/4) Epoch 24, batch 2200, loss[loss=0.2364, ctc_loss=0.1549, cr_loss=0.4075, over 17013.00 frames. ], tot_loss[loss=0.21, ctc_loss=0.1385, cr_loss=0.3572, over 3357898.41 frames. ], batch size: 53, lr: 4.98e-03, grad_scale: 16.0
2024-09-24 05:16:25,767 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.245e+02 1.329e+02 1.411e+02 2.102e+02, threshold=2.657e+02, percent-clipped=0.0
2024-09-24 05:16:31,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=428442.0, ans=0.0
2024-09-24 05:16:39,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=428488.6666666667, ans=0.0
2024-09-24 05:16:44,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=428488.6666666667, ans=0.125
2024-09-24 05:17:04,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=428535.3333333333, ans=0.125
2024-09-24 05:17:06,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.35 vs. limit=15.0
2024-09-24 05:17:29,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=428628.6666666667, ans=0.125
2024-09-24 05:17:31,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=428628.6666666667, ans=0.1
2024-09-24 05:17:40,961 INFO [train.py:1198] (1/4) Epoch 24, batch 2250, loss[loss=0.2312, ctc_loss=0.1517, cr_loss=0.3974, over 16633.00 frames. ], tot_loss[loss=0.2099, ctc_loss=0.1384, cr_loss=0.3572, over 3355329.79 frames. ], batch size: 66, lr: 4.98e-03, grad_scale: 16.0
2024-09-24 05:18:17,278 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.54 vs. limit=12.0
2024-09-24 05:18:26,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=428768.6666666667, ans=0.1
2024-09-24 05:18:46,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten.whitening_limit, batch_count=428862.0, ans=15.0
2024-09-24 05:18:48,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=428862.0, ans=0.0
2024-09-24 05:19:01,240 INFO [train.py:1198] (1/4) Epoch 24, batch 2300, loss[loss=0.2034, ctc_loss=0.133, cr_loss=0.3521, over 17205.00 frames. ], tot_loss[loss=0.2093, ctc_loss=0.1381, cr_loss=0.3562, over 3356195.22 frames. ], batch size: 55, lr: 4.98e-03, grad_scale: 16.0
2024-09-24 05:19:11,767 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.244e+02 1.345e+02 1.429e+02 2.009e+02, threshold=2.690e+02, percent-clipped=0.0
2024-09-24 05:19:25,484 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.64 vs. limit=12.0
2024-09-24 05:20:04,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.06 vs. limit=22.5
2024-09-24 05:20:28,878 INFO [train.py:1198] (1/4) Epoch 24, batch 2350, loss[loss=0.1939, ctc_loss=0.1248, cr_loss=0.3453, over 17214.00 frames. ], tot_loss[loss=0.2106, ctc_loss=0.139, cr_loss=0.3579, over 3355807.01 frames. ], batch size: 41, lr: 4.98e-03, grad_scale: 16.0
2024-09-24 05:20:36,243 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.71 vs. limit=15.0
2024-09-24 05:20:47,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=429188.6666666667, ans=0.125
2024-09-24 05:20:52,938 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=9.80 vs. limit=15.0
2024-09-24 05:21:03,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=429235.3333333333, ans=0.125
2024-09-24 05:21:12,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=429235.3333333333, ans=0.125
2024-09-24 05:21:16,372 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.39 vs. limit=15.0
2024-09-24 05:21:54,430 INFO [train.py:1198] (1/4) Epoch 24, batch 2400, loss[loss=0.221, ctc_loss=0.1469, cr_loss=0.3703, over 17282.00 frames. ], tot_loss[loss=0.2091, ctc_loss=0.138, cr_loss=0.3556, over 3357137.19 frames. ], batch size: 49, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:22:02,374 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.270e+02 1.408e+02 1.564e+02 2.432e+02, threshold=2.816e+02, percent-clipped=0.0
2024-09-24 05:22:18,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=429422.0, ans=0.125
2024-09-24 05:22:33,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.90 vs. limit=15.0
2024-09-24 05:22:36,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=429468.6666666667, ans=0.125
2024-09-24 05:22:44,447 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=14.36 vs. limit=15.0
2024-09-24 05:22:52,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=429515.3333333333, ans=0.125
2024-09-24 05:23:06,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=429562.0, ans=0.025
2024-09-24 05:23:07,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=429562.0, ans=0.0
2024-09-24 05:23:13,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=429608.6666666667, ans=0.2
2024-09-24 05:23:14,641 INFO [train.py:1198] (1/4) Epoch 24, batch 2450, loss[loss=0.2251, ctc_loss=0.1531, cr_loss=0.3599, over 15150.00 frames. ], tot_loss[loss=0.209, ctc_loss=0.1378, cr_loss=0.3559, over 3352336.00 frames. ], batch size: 89, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:23:19,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=429608.6666666667, ans=0.025
2024-09-24 05:23:33,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.85 vs. limit=10.0
2024-09-24 05:24:12,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=429748.6666666667, ans=0.035
2024-09-24 05:24:25,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=429795.3333333333, ans=0.2
2024-09-24 05:24:37,351 INFO [train.py:1198] (1/4) Epoch 24, batch 2500, loss[loss=0.2385, ctc_loss=0.1593, cr_loss=0.3958, over 16480.00 frames. ], tot_loss[loss=0.2099, ctc_loss=0.1385, cr_loss=0.3569, over 3353927.06 frames. ], batch size: 66, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:24:48,051 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.250e+02 1.362e+02 1.457e+02 2.366e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-24 05:25:10,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=429888.6666666667, ans=0.125
2024-09-24 05:25:11,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=429888.6666666667, ans=0.125
2024-09-24 05:25:25,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=429935.3333333333, ans=0.1
2024-09-24 05:25:29,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=429982.0, ans=0.125
2024-09-24 05:25:33,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=429982.0, ans=0.0
2024-09-24 05:25:35,265 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.40 vs. limit=22.5
2024-09-24 05:25:42,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=429982.0, ans=0.125
2024-09-24 05:26:02,839 INFO [train.py:1198] (1/4) Epoch 24, batch 2550, loss[loss=0.1974, ctc_loss=0.1304, cr_loss=0.3351, over 17300.00 frames. ], tot_loss[loss=0.209, ctc_loss=0.1379, cr_loss=0.3555, over 3363382.37 frames. ], batch size: 46, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:26:09,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=430075.3333333333, ans=0.07
2024-09-24 05:26:28,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=430122.0, ans=0.025
2024-09-24 05:26:37,464 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.69 vs. limit=15.0
2024-09-24 05:26:56,188 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.08 vs. limit=15.0
2024-09-24 05:26:58,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=430215.3333333333, ans=0.125
2024-09-24 05:26:59,446 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.24 vs. limit=6.0
2024-09-24 05:27:09,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=430262.0, ans=0.125
2024-09-24 05:27:25,867 INFO [train.py:1198] (1/4) Epoch 24, batch 2600, loss[loss=0.2256, ctc_loss=0.1491, cr_loss=0.3829, over 16999.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.1373, cr_loss=0.3538, over 3354632.21 frames. ], batch size: 53, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:27:33,811 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.253e+02 1.345e+02 1.495e+02 2.149e+02, threshold=2.690e+02, percent-clipped=0.0
2024-09-24 05:27:36,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=430308.6666666667, ans=6.0
2024-09-24 05:27:38,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=430308.6666666667, ans=0.125
2024-09-24 05:27:56,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=430402.0, ans=0.125
2024-09-24 05:28:32,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=430495.3333333333, ans=0.125
2024-09-24 05:28:44,974 INFO [train.py:1198] (1/4) Epoch 24, batch 2650, loss[loss=0.2363, ctc_loss=0.1595, cr_loss=0.3839, over 15073.00 frames. ], tot_loss[loss=0.2086, ctc_loss=0.1377, cr_loss=0.3546, over 3341959.94 frames. ], batch size: 89, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:28:49,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=9.58 vs. limit=15.0
2024-09-24 05:29:03,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=430588.6666666667, ans=0.1
2024-09-24 05:29:06,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.93 vs. limit=15.0
2024-09-24 05:29:21,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=430635.3333333333, ans=0.025
2024-09-24 05:29:21,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=430635.3333333333, ans=0.2
2024-09-24 05:29:31,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=430635.3333333333, ans=0.125
2024-09-24 05:29:41,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=430682.0, ans=0.125
2024-09-24 05:29:50,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=430682.0, ans=0.2
2024-09-24 05:30:12,483 INFO [train.py:1198] (1/4) Epoch 24, batch 2700, loss[loss=0.1793, ctc_loss=0.1164, cr_loss=0.3147, over 16964.00 frames. ], tot_loss[loss=0.2089, ctc_loss=0.138, cr_loss=0.3545, over 3350584.17 frames. ], batch size: 42, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:30:14,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=430775.3333333333, ans=0.0
2024-09-24 05:30:20,466 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.268e+02 1.356e+02 1.521e+02 3.128e+02, threshold=2.712e+02, percent-clipped=1.0
2024-09-24 05:30:33,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=430822.0, ans=0.2
2024-09-24 05:30:37,347 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.84 vs. limit=22.5
2024-09-24 05:30:49,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=430868.6666666667, ans=0.125
2024-09-24 05:30:57,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=430868.6666666667, ans=0.125
2024-09-24 05:31:11,818 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 05:31:28,379 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.53 vs. limit=6.0
2024-09-24 05:31:30,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=431008.6666666667, ans=0.125
2024-09-24 05:31:32,249 INFO [train.py:1198] (1/4) Epoch 24, batch 2750, loss[loss=0.1828, ctc_loss=0.1201, cr_loss=0.3132, over 17007.00 frames. ], tot_loss[loss=0.2089, ctc_loss=0.1379, cr_loss=0.3546, over 3359907.12 frames. ], batch size: 51, lr: 4.97e-03, grad_scale: 32.0
2024-09-24 05:32:04,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=431055.3333333333, ans=0.1
2024-09-24 05:32:07,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=431102.0, ans=0.1
2024-09-24 05:32:30,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=431148.6666666667, ans=0.125
2024-09-24 05:32:55,658 INFO [train.py:1198] (1/4) Epoch 24, batch 2800, loss[loss=0.2457, ctc_loss=0.1605, cr_loss=0.426, over 17024.00 frames. ], tot_loss[loss=0.2103, ctc_loss=0.1391, cr_loss=0.3562, over 3355030.91 frames. ], batch size: 52, lr: 4.96e-03, grad_scale: 32.0
2024-09-24 05:33:03,460 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.249e+02 1.392e+02 1.537e+02 2.011e+02, threshold=2.784e+02, percent-clipped=0.0
2024-09-24 05:33:09,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=431288.6666666667, ans=0.0
2024-09-24 05:33:18,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=431288.6666666667, ans=0.2
2024-09-24 05:33:31,618 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.61 vs. limit=12.0
2024-09-24 05:33:43,831 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 05:33:59,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=431382.0, ans=0.125
2024-09-24 05:34:18,046 INFO [train.py:1198] (1/4) Epoch 24, batch 2850, loss[loss=0.1848, ctc_loss=0.1181, cr_loss=0.3331, over 17275.00 frames. ], tot_loss[loss=0.2102, ctc_loss=0.1389, cr_loss=0.3565, over 3358430.56 frames. ], batch size: 42, lr: 4.96e-03, grad_scale: 32.0
2024-09-24 05:34:23,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=431475.3333333333, ans=0.125
2024-09-24 05:34:57,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=431568.6666666667, ans=0.09899494936611666
2024-09-24 05:35:08,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=431568.6666666667, ans=0.125
2024-09-24 05:35:10,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=431615.3333333333, ans=0.125
2024-09-24 05:35:13,627 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=6.44 vs. limit=12.0
2024-09-24 05:35:43,353 INFO [train.py:1198] (1/4) Epoch 24, batch 2900, loss[loss=0.1987, ctc_loss=0.1309, cr_loss=0.3388, over 17249.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1386, cr_loss=0.3558, over 3359165.73 frames. ], batch size: 44, lr: 4.96e-03, grad_scale: 16.0
2024-09-24 05:35:52,964 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.236e+02 1.333e+02 1.474e+02 3.420e+02, threshold=2.666e+02, percent-clipped=1.0
2024-09-24 05:36:00,073 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.16 vs. limit=15.0
2024-09-24 05:36:35,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.81 vs. limit=12.0
2024-09-24 05:36:40,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=431848.6666666667, ans=0.125
2024-09-24 05:36:41,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=431848.6666666667, ans=0.0
2024-09-24 05:36:45,890 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.20 vs. limit=8.0
2024-09-24 05:37:02,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=431895.3333333333, ans=0.025
2024-09-24 05:37:06,444 INFO [train.py:1198] (1/4) Epoch 24, batch 2950, loss[loss=0.261, ctc_loss=0.187, cr_loss=0.3704, over 11499.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1385, cr_loss=0.3558, over 3354081.94 frames. ], batch size: 123, lr: 4.96e-03, grad_scale: 16.0
2024-09-24 05:37:54,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=432082.0, ans=0.0
2024-09-24 05:38:03,364 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.84 vs. limit=15.0
2024-09-24 05:38:20,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=432128.6666666667, ans=0.0
2024-09-24 05:38:26,284 INFO [train.py:1198] (1/4) Epoch 24, batch 3000, loss[loss=0.2055, ctc_loss=0.1333, cr_loss=0.3609, over 17320.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1385, cr_loss=0.356, over 3355375.12 frames. ], batch size: 51, lr: 4.96e-03, grad_scale: 16.0
2024-09-24 05:38:26,284 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 05:38:42,491 INFO [train.py:1230] (1/4) Epoch 24, validation: loss=0.03786, ctc_loss=0.03786, cr_loss=8.617e-15, over 944034.00 frames. 
2024-09-24 05:38:42,492 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 05:38:51,890 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.040e+02 1.243e+02 1.342e+02 1.455e+02 1.995e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-24 05:38:58,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=432222.0, ans=0.2
2024-09-24 05:39:11,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=432222.0, ans=0.05
2024-09-24 05:39:51,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=432362.0, ans=0.125
2024-09-24 05:40:03,068 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.44 vs. limit=22.5
2024-09-24 05:40:03,771 INFO [train.py:1198] (1/4) Epoch 24, batch 3050, loss[loss=0.1735, ctc_loss=0.1116, cr_loss=0.3095, over 17107.00 frames. ], tot_loss[loss=0.2098, ctc_loss=0.1386, cr_loss=0.3562, over 3364597.71 frames. ], batch size: 43, lr: 4.96e-03, grad_scale: 16.0
2024-09-24 05:40:18,234 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 05:40:23,512 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.51 vs. limit=15.0
2024-09-24 05:40:54,502 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.38 vs. limit=15.0
2024-09-24 05:41:22,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=432595.3333333333, ans=0.0
2024-09-24 05:41:26,720 INFO [train.py:1198] (1/4) Epoch 24, batch 3100, loss[loss=0.2143, ctc_loss=0.1405, cr_loss=0.3692, over 17260.00 frames. ], tot_loss[loss=0.209, ctc_loss=0.1381, cr_loss=0.3549, over 3361162.38 frames. ], batch size: 44, lr: 4.96e-03, grad_scale: 16.0
2024-09-24 05:41:35,818 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.243e+02 1.328e+02 1.464e+02 2.073e+02, threshold=2.656e+02, percent-clipped=0.0
2024-09-24 05:42:15,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=432782.0, ans=0.1
2024-09-24 05:42:18,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=432782.0, ans=0.0
2024-09-24 05:42:23,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=432782.0, ans=0.125
2024-09-24 05:42:26,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=432782.0, ans=0.0
2024-09-24 05:42:30,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=432828.6666666667, ans=0.1
2024-09-24 05:42:40,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=432828.6666666667, ans=0.125
2024-09-24 05:42:44,646 INFO [train.py:1198] (1/4) Epoch 24, batch 3150, loss[loss=0.2112, ctc_loss=0.1371, cr_loss=0.3705, over 17028.00 frames. ], tot_loss[loss=0.2101, ctc_loss=0.1389, cr_loss=0.356, over 3356476.92 frames. ], batch size: 44, lr: 4.95e-03, grad_scale: 16.0
2024-09-24 05:42:57,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=432875.3333333333, ans=0.0
2024-09-24 05:43:20,082 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.31 vs. limit=15.0
2024-09-24 05:43:24,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=432968.6666666667, ans=0.125
2024-09-24 05:43:24,760 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.03 vs. limit=6.0
2024-09-24 05:44:02,936 INFO [train.py:1198] (1/4) Epoch 24, batch 3200, loss[loss=0.2283, ctc_loss=0.1513, cr_loss=0.3847, over 16529.00 frames. ], tot_loss[loss=0.21, ctc_loss=0.1388, cr_loss=0.3559, over 3353153.60 frames. ], batch size: 66, lr: 4.95e-03, grad_scale: 32.0
2024-09-24 05:44:10,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=433108.6666666667, ans=0.04949747468305833
2024-09-24 05:44:12,035 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.252e+02 1.364e+02 1.478e+02 2.406e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-24 05:44:27,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=433155.3333333333, ans=0.0
2024-09-24 05:45:02,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=433248.6666666667, ans=0.125
2024-09-24 05:45:20,627 INFO [train.py:1198] (1/4) Epoch 24, batch 3250, loss[loss=0.2236, ctc_loss=0.1478, cr_loss=0.379, over 17098.00 frames. ], tot_loss[loss=0.2097, ctc_loss=0.1387, cr_loss=0.3551, over 3344580.37 frames. ], batch size: 49, lr: 4.95e-03, grad_scale: 32.0
2024-09-24 05:45:27,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=433342.0, ans=0.125
2024-09-24 05:45:43,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=433388.6666666667, ans=0.025
2024-09-24 05:45:51,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=433388.6666666667, ans=0.0
2024-09-24 05:46:05,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=433435.3333333333, ans=0.05
2024-09-24 05:46:10,689 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.79 vs. limit=15.0
2024-09-24 05:46:36,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=433528.6666666667, ans=0.0
2024-09-24 05:46:36,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=433528.6666666667, ans=0.125
2024-09-24 05:46:39,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=433575.3333333333, ans=0.125
2024-09-24 05:46:40,881 INFO [train.py:1198] (1/4) Epoch 24, batch 3300, loss[loss=0.1739, ctc_loss=0.1132, cr_loss=0.3035, over 16948.00 frames. ], tot_loss[loss=0.2102, ctc_loss=0.1389, cr_loss=0.3562, over 3348497.88 frames. ], batch size: 42, lr: 4.95e-03, grad_scale: 32.0
2024-09-24 05:46:45,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=433575.3333333333, ans=0.0
2024-09-24 05:46:49,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=433575.3333333333, ans=0.0
2024-09-24 05:46:50,400 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.267e+02 1.337e+02 1.528e+02 2.027e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-24 05:46:58,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=433622.0, ans=0.2
2024-09-24 05:47:09,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=433622.0, ans=0.1
2024-09-24 05:47:41,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=433762.0, ans=0.125
2024-09-24 05:47:50,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=433762.0, ans=0.125
2024-09-24 05:47:58,603 INFO [train.py:1198] (1/4) Epoch 24, batch 3350, loss[loss=0.2026, ctc_loss=0.1341, cr_loss=0.3426, over 17163.00 frames. ], tot_loss[loss=0.2095, ctc_loss=0.1384, cr_loss=0.3558, over 3347765.15 frames. ], batch size: 45, lr: 4.95e-03, grad_scale: 32.0
2024-09-24 05:48:14,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=433855.3333333333, ans=0.125
2024-09-24 05:48:20,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=433855.3333333333, ans=0.125
2024-09-24 05:48:31,269 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.89 vs. limit=5.0
2024-09-24 05:48:53,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=433948.6666666667, ans=0.2
2024-09-24 05:48:59,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=433995.3333333333, ans=0.1
2024-09-24 05:49:03,787 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.35 vs. limit=22.5
2024-09-24 05:49:05,234 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.59 vs. limit=15.0
2024-09-24 05:49:16,884 INFO [train.py:1198] (1/4) Epoch 24, batch 3400, loss[loss=0.219, ctc_loss=0.1443, cr_loss=0.3732, over 17035.00 frames. ], tot_loss[loss=0.2085, ctc_loss=0.1377, cr_loss=0.3541, over 3336660.10 frames. ], batch size: 52, lr: 4.95e-03, grad_scale: 32.0
2024-09-24 05:49:18,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=434042.0, ans=0.0
2024-09-24 05:49:26,483 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.267e+02 1.361e+02 1.503e+02 2.338e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-24 05:49:49,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=434135.3333333333, ans=0.125
2024-09-24 05:49:50,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=434135.3333333333, ans=0.125
2024-09-24 05:49:50,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=434135.3333333333, ans=0.125
2024-09-24 05:50:02,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=434135.3333333333, ans=0.0
2024-09-24 05:50:15,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=434182.0, ans=0.125
2024-09-24 05:50:18,760 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.49 vs. limit=15.0
2024-09-24 05:50:36,791 INFO [train.py:1198] (1/4) Epoch 24, batch 3450, loss[loss=0.2261, ctc_loss=0.1494, cr_loss=0.3836, over 17069.00 frames. ], tot_loss[loss=0.2085, ctc_loss=0.1377, cr_loss=0.3541, over 3340506.50 frames. ], batch size: 52, lr: 4.95e-03, grad_scale: 32.0
2024-09-24 05:51:05,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=434322.0, ans=0.2
2024-09-24 05:51:26,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=434415.3333333333, ans=0.025
2024-09-24 05:51:30,513 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.36 vs. limit=12.0
2024-09-24 05:51:59,813 INFO [train.py:1198] (1/4) Epoch 24, batch 3500, loss[loss=0.2125, ctc_loss=0.1389, cr_loss=0.3679, over 17238.00 frames. ], tot_loss[loss=0.2082, ctc_loss=0.1374, cr_loss=0.3539, over 3346031.79 frames. ], batch size: 55, lr: 4.95e-03, grad_scale: 32.0
2024-09-24 05:52:00,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=434508.6666666667, ans=0.025
2024-09-24 05:52:10,825 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.254e+02 1.358e+02 1.511e+02 3.142e+02, threshold=2.715e+02, percent-clipped=1.0
2024-09-24 05:52:14,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=434555.3333333333, ans=0.125
2024-09-24 05:52:24,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=434555.3333333333, ans=0.0
2024-09-24 05:52:39,763 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.63 vs. limit=15.0
2024-09-24 05:52:44,811 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.51 vs. limit=15.0
2024-09-24 05:52:53,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=434648.6666666667, ans=0.125
2024-09-24 05:52:58,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=434648.6666666667, ans=0.125
2024-09-24 05:53:00,215 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.40 vs. limit=10.0
2024-09-24 05:53:04,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=434695.3333333333, ans=0.07
2024-09-24 05:53:10,026 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.22 vs. limit=6.0
2024-09-24 05:53:18,583 INFO [train.py:1198] (1/4) Epoch 24, batch 3550, loss[loss=0.1758, ctc_loss=0.1123, cr_loss=0.3175, over 17257.00 frames. ], tot_loss[loss=0.2082, ctc_loss=0.1373, cr_loss=0.3545, over 3351820.73 frames. ], batch size: 44, lr: 4.94e-03, grad_scale: 16.0
2024-09-24 05:53:35,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=434788.6666666667, ans=0.035
2024-09-24 05:54:18,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=434882.0, ans=0.125
2024-09-24 05:54:25,294 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.89 vs. limit=15.0
2024-09-24 05:54:36,017 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.00 vs. limit=15.0
2024-09-24 05:54:36,759 INFO [train.py:1198] (1/4) Epoch 24, batch 3600, loss[loss=0.2106, ctc_loss=0.1353, cr_loss=0.3766, over 17317.00 frames. ], tot_loss[loss=0.2083, ctc_loss=0.1374, cr_loss=0.3543, over 3350580.44 frames. ], batch size: 51, lr: 4.94e-03, grad_scale: 32.0
2024-09-24 05:54:47,631 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.266e+02 1.361e+02 1.484e+02 1.804e+02, threshold=2.723e+02, percent-clipped=0.0
2024-09-24 05:55:24,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=435115.3333333333, ans=0.0
2024-09-24 05:55:33,621 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=435115.3333333333, ans=0.125
2024-09-24 05:55:51,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=435162.0, ans=0.025
2024-09-24 05:55:57,298 INFO [train.py:1198] (1/4) Epoch 24, batch 3650, loss[loss=0.1806, ctc_loss=0.1159, cr_loss=0.3234, over 16949.00 frames. ], tot_loss[loss=0.2081, ctc_loss=0.1374, cr_loss=0.3535, over 3345364.66 frames. ], batch size: 42, lr: 4.94e-03, grad_scale: 32.0
2024-09-24 05:56:15,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=435255.3333333333, ans=0.125
2024-09-24 05:56:16,932 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=16.50 vs. limit=22.5
2024-09-24 05:56:35,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=435302.0, ans=0.0
2024-09-24 05:56:46,191 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.86 vs. limit=6.0
2024-09-24 05:56:46,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=435348.6666666667, ans=0.5
2024-09-24 05:57:16,735 INFO [train.py:1198] (1/4) Epoch 24, batch 3700, loss[loss=0.1553, ctc_loss=0.1016, cr_loss=0.2688, over 17053.00 frames. ], tot_loss[loss=0.2071, ctc_loss=0.1367, cr_loss=0.352, over 3354997.03 frames. ], batch size: 39, lr: 4.94e-03, grad_scale: 16.0
2024-09-24 05:57:17,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=435442.0, ans=0.2
2024-09-24 05:57:29,293 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.270e+02 1.350e+02 1.462e+02 1.892e+02, threshold=2.701e+02, percent-clipped=0.0
2024-09-24 05:57:37,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=435488.6666666667, ans=0.125
2024-09-24 05:57:43,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=435488.6666666667, ans=0.0
2024-09-24 05:57:50,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=10.44 vs. limit=15.0
2024-09-24 05:58:32,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=18.26 vs. limit=22.5
2024-09-24 05:58:35,097 INFO [train.py:1198] (1/4) Epoch 24, batch 3750, loss[loss=0.1976, ctc_loss=0.1287, cr_loss=0.3445, over 17355.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1365, cr_loss=0.3513, over 3355304.81 frames. ], batch size: 48, lr: 4.94e-03, grad_scale: 16.0
2024-09-24 05:59:09,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=435768.6666666667, ans=0.2
2024-09-24 05:59:10,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=435768.6666666667, ans=0.0
2024-09-24 05:59:43,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=435862.0, ans=0.125
2024-09-24 05:59:54,384 INFO [train.py:1198] (1/4) Epoch 24, batch 3800, loss[loss=0.2244, ctc_loss=0.1511, cr_loss=0.3663, over 17163.00 frames. ], tot_loss[loss=0.2087, ctc_loss=0.1381, cr_loss=0.3533, over 3314888.41 frames. ], batch size: 45, lr: 4.94e-03, grad_scale: 16.0
2024-09-24 06:00:02,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=435908.6666666667, ans=0.1
2024-09-24 06:00:05,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=435908.6666666667, ans=0.125
2024-09-24 06:00:07,111 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.258e+02 1.341e+02 1.479e+02 2.397e+02, threshold=2.682e+02, percent-clipped=0.0
2024-09-24 06:00:16,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=435955.3333333333, ans=0.2
2024-09-24 06:00:29,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=436002.0, ans=0.125
2024-09-24 06:00:35,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=436002.0, ans=0.125
2024-09-24 06:00:41,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=436048.6666666667, ans=0.1
2024-09-24 06:00:45,061 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.16 vs. limit=15.0
2024-09-24 06:00:47,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=436048.6666666667, ans=0.125
2024-09-24 06:01:01,087 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.80 vs. limit=15.0
2024-09-24 06:01:03,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=436095.3333333333, ans=0.1
2024-09-24 06:01:14,193 INFO [train.py:1198] (1/4) Epoch 24, batch 3850, loss[loss=0.2206, ctc_loss=0.147, cr_loss=0.3684, over 17004.00 frames. ], tot_loss[loss=0.2083, ctc_loss=0.1378, cr_loss=0.3526, over 3303465.97 frames. ], batch size: 53, lr: 4.94e-03, grad_scale: 16.0
2024-09-24 06:01:22,722 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.72 vs. limit=10.0
2024-09-24 06:01:37,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=436188.6666666667, ans=0.125
2024-09-24 06:01:48,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=436235.3333333333, ans=0.025
2024-09-24 06:01:52,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=436235.3333333333, ans=0.1
2024-09-24 06:01:54,219 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=436235.3333333333, ans=0.2
2024-09-24 06:01:58,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=436282.0, ans=0.125
2024-09-24 06:02:03,744 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=10.78 vs. limit=12.0
2024-09-24 06:02:09,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=436282.0, ans=0.125
2024-09-24 06:02:20,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=436328.6666666667, ans=0.125
2024-09-24 06:03:16,546 INFO [train.py:1198] (1/4) Epoch 25, batch 0, loss[loss=0.2085, ctc_loss=0.1361, cr_loss=0.3623, over 17223.00 frames. ], tot_loss[loss=0.2085, ctc_loss=0.1361, cr_loss=0.3623, over 17223.00 frames. ], batch size: 50, lr: 4.83e-03, grad_scale: 32.0
2024-09-24 06:03:16,546 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 06:03:31,928 INFO [train.py:1230] (1/4) Epoch 25, validation: loss=0.03759, ctc_loss=0.03759, cr_loss=8.067e-15, over 944034.00 frames. 
2024-09-24 06:03:31,929 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 06:03:51,062 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.314e+02 1.430e+02 1.672e+02 2.033e+02, threshold=2.861e+02, percent-clipped=0.0
2024-09-24 06:04:17,000 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.22 vs. limit=15.0
2024-09-24 06:04:18,505 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.28 vs. limit=22.5
2024-09-24 06:04:54,468 INFO [train.py:1198] (1/4) Epoch 25, batch 50, loss[loss=0.2448, ctc_loss=0.1711, cr_loss=0.3684, over 11932.00 frames. ], tot_loss[loss=0.2069, ctc_loss=0.1362, cr_loss=0.3532, over 758459.14 frames. ], batch size: 123, lr: 4.83e-03, grad_scale: 32.0
2024-09-24 06:04:59,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=436590.0, ans=0.5
2024-09-24 06:05:10,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=436636.6666666667, ans=0.0
2024-09-24 06:06:19,638 INFO [train.py:1198] (1/4) Epoch 25, batch 100, loss[loss=0.1818, ctc_loss=0.1176, cr_loss=0.321, over 17008.00 frames. ], tot_loss[loss=0.2064, ctc_loss=0.136, cr_loss=0.352, over 1335339.56 frames. ], batch size: 39, lr: 4.83e-03, grad_scale: 16.0
2024-09-24 06:06:27,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=436823.3333333333, ans=0.125
2024-09-24 06:06:40,252 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.292e+02 1.373e+02 1.497e+02 2.148e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-24 06:07:17,664 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.85 vs. limit=10.0
2024-09-24 06:07:42,085 INFO [train.py:1198] (1/4) Epoch 25, batch 150, loss[loss=0.1877, ctc_loss=0.1207, cr_loss=0.3348, over 17212.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.135, cr_loss=0.3504, over 1788309.54 frames. ], batch size: 41, lr: 4.83e-03, grad_scale: 16.0
2024-09-24 06:07:42,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=437056.6666666667, ans=0.125
2024-09-24 06:08:03,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=437103.3333333333, ans=0.2
2024-09-24 06:08:08,118 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.38 vs. limit=12.0
2024-09-24 06:08:14,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.75 vs. limit=12.0
2024-09-24 06:08:18,134 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.74 vs. limit=15.0
2024-09-24 06:08:20,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=437150.0, ans=0.04949747468305833
2024-09-24 06:08:36,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer_na.min_abs, batch_count=437196.6666666667, ans=0.02
2024-09-24 06:08:39,858 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.67 vs. limit=15.0
2024-09-24 06:08:53,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=437243.3333333333, ans=0.125
2024-09-24 06:09:04,124 INFO [train.py:1198] (1/4) Epoch 25, batch 200, loss[loss=0.2014, ctc_loss=0.1331, cr_loss=0.3415, over 17269.00 frames. ], tot_loss[loss=0.206, ctc_loss=0.1357, cr_loss=0.3518, over 2140024.48 frames. ], batch size: 44, lr: 4.83e-03, grad_scale: 8.0
2024-09-24 06:09:14,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=437290.0, ans=0.09899494936611666
2024-09-24 06:09:26,586 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.223e+02 1.322e+02 1.442e+02 1.903e+02, threshold=2.645e+02, percent-clipped=0.0
2024-09-24 06:09:28,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=437336.6666666667, ans=0.5
2024-09-24 06:09:31,997 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.00 vs. limit=12.0
2024-09-24 06:09:46,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=437383.3333333333, ans=0.0
2024-09-24 06:10:08,474 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=437476.6666666667, ans=0.0
2024-09-24 06:10:18,328 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.30 vs. limit=15.0
2024-09-24 06:10:24,168 INFO [train.py:1198] (1/4) Epoch 25, batch 250, loss[loss=0.1961, ctc_loss=0.1265, cr_loss=0.3482, over 17273.00 frames. ], tot_loss[loss=0.2054, ctc_loss=0.1352, cr_loss=0.3509, over 2414388.80 frames. ], batch size: 42, lr: 4.83e-03, grad_scale: 8.0
2024-09-24 06:10:31,302 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.86 vs. limit=15.0
2024-09-24 06:10:41,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=437570.0, ans=0.125
2024-09-24 06:10:46,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=437570.0, ans=0.125
2024-09-24 06:11:12,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=437616.6666666667, ans=0.035
2024-09-24 06:11:14,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=437616.6666666667, ans=0.0
2024-09-24 06:11:24,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=437663.3333333333, ans=0.125
2024-09-24 06:11:27,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=437663.3333333333, ans=0.0
2024-09-24 06:11:49,771 INFO [train.py:1198] (1/4) Epoch 25, batch 300, loss[loss=0.2058, ctc_loss=0.135, cr_loss=0.354, over 17203.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.135, cr_loss=0.3511, over 2630134.32 frames. ], batch size: 55, lr: 4.83e-03, grad_scale: 8.0
2024-09-24 06:12:05,074 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.13 vs. limit=15.0
2024-09-24 06:12:12,131 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.257e+02 1.339e+02 1.438e+02 1.926e+02, threshold=2.678e+02, percent-clipped=0.0
2024-09-24 06:13:08,011 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.52 vs. limit=15.0
2024-09-24 06:13:09,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=437943.3333333333, ans=0.125
2024-09-24 06:13:12,245 INFO [train.py:1198] (1/4) Epoch 25, batch 350, loss[loss=0.1827, ctc_loss=0.1213, cr_loss=0.3071, over 16272.00 frames. ], tot_loss[loss=0.2072, ctc_loss=0.1364, cr_loss=0.354, over 2792654.70 frames. ], batch size: 36, lr: 4.82e-03, grad_scale: 8.0
2024-09-24 06:13:28,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=438036.6666666667, ans=0.025
2024-09-24 06:13:34,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=438036.6666666667, ans=0.125
2024-09-24 06:13:35,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=438036.6666666667, ans=0.125
2024-09-24 06:14:06,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=438130.0, ans=0.125
2024-09-24 06:14:07,056 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.84 vs. limit=6.0
2024-09-24 06:14:35,220 INFO [train.py:1198] (1/4) Epoch 25, batch 400, loss[loss=0.216, ctc_loss=0.1443, cr_loss=0.3582, over 17023.00 frames. ], tot_loss[loss=0.2064, ctc_loss=0.1358, cr_loss=0.3526, over 2928001.87 frames. ], batch size: 51, lr: 4.82e-03, grad_scale: 16.0
2024-09-24 06:14:43,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=438223.3333333333, ans=0.125
2024-09-24 06:14:47,554 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.10 vs. limit=12.0
2024-09-24 06:14:57,568 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.241e+02 1.339e+02 1.521e+02 2.224e+02, threshold=2.677e+02, percent-clipped=0.0
2024-09-24 06:15:24,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=438363.3333333333, ans=0.125
2024-09-24 06:15:34,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=438363.3333333333, ans=0.1
2024-09-24 06:15:57,605 INFO [train.py:1198] (1/4) Epoch 25, batch 450, loss[loss=0.1932, ctc_loss=0.1255, cr_loss=0.3383, over 17241.00 frames. ], tot_loss[loss=0.2072, ctc_loss=0.1364, cr_loss=0.3541, over 3033912.51 frames. ], batch size: 42, lr: 4.82e-03, grad_scale: 16.0
2024-09-24 06:16:16,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=438503.3333333333, ans=0.125
2024-09-24 06:16:20,286 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.46 vs. limit=15.0
2024-09-24 06:17:01,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=438596.6666666667, ans=0.125
2024-09-24 06:17:20,718 INFO [train.py:1198] (1/4) Epoch 25, batch 500, loss[loss=0.2092, ctc_loss=0.1367, cr_loss=0.3624, over 17308.00 frames. ], tot_loss[loss=0.2074, ctc_loss=0.1365, cr_loss=0.3545, over 3110348.77 frames. ], batch size: 49, lr: 4.82e-03, grad_scale: 16.0
2024-09-24 06:17:46,419 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.265e+02 1.364e+02 1.484e+02 2.816e+02, threshold=2.728e+02, percent-clipped=1.0
2024-09-24 06:17:53,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=438736.6666666667, ans=0.125
2024-09-24 06:18:06,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=438783.3333333333, ans=0.125
2024-09-24 06:18:32,690 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.96 vs. limit=15.0
2024-09-24 06:18:38,804 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.76 vs. limit=10.0
2024-09-24 06:18:44,382 INFO [train.py:1198] (1/4) Epoch 25, batch 550, loss[loss=0.1917, ctc_loss=0.1247, cr_loss=0.335, over 17142.00 frames. ], tot_loss[loss=0.2073, ctc_loss=0.1365, cr_loss=0.3542, over 3167300.84 frames. ], batch size: 48, lr: 4.82e-03, grad_scale: 16.0
2024-09-24 06:18:58,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=438923.3333333333, ans=0.2
2024-09-24 06:19:06,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=438970.0, ans=0.125
2024-09-24 06:19:09,750 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 06:19:27,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=439016.6666666667, ans=0.07
2024-09-24 06:19:44,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=439063.3333333333, ans=0.1
2024-09-24 06:20:01,486 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.13 vs. limit=15.0
2024-09-24 06:20:06,966 INFO [train.py:1198] (1/4) Epoch 25, batch 600, loss[loss=0.2067, ctc_loss=0.1342, cr_loss=0.3623, over 17020.00 frames. ], tot_loss[loss=0.207, ctc_loss=0.1361, cr_loss=0.3543, over 3214069.84 frames. ], batch size: 44, lr: 4.82e-03, grad_scale: 16.0
2024-09-24 06:20:29,578 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.242e+02 1.338e+02 1.453e+02 1.774e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-24 06:20:41,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=439250.0, ans=0.09899494936611666
2024-09-24 06:20:41,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=439250.0, ans=0.0
2024-09-24 06:20:43,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.06 vs. limit=15.0
2024-09-24 06:20:45,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=439250.0, ans=0.0
2024-09-24 06:21:13,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=439296.6666666667, ans=0.125
2024-09-24 06:21:15,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=439343.3333333333, ans=0.2
2024-09-24 06:21:15,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=439343.3333333333, ans=0.2
2024-09-24 06:21:18,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=439343.3333333333, ans=0.125
2024-09-24 06:21:27,265 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.48 vs. limit=15.0
2024-09-24 06:21:29,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=439343.3333333333, ans=0.125
2024-09-24 06:21:32,783 INFO [train.py:1198] (1/4) Epoch 25, batch 650, loss[loss=0.211, ctc_loss=0.1397, cr_loss=0.3565, over 16992.00 frames. ], tot_loss[loss=0.2062, ctc_loss=0.1356, cr_loss=0.3531, over 3252898.53 frames. ], batch size: 53, lr: 4.82e-03, grad_scale: 16.0
2024-09-24 06:21:52,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=439436.6666666667, ans=0.125
2024-09-24 06:22:26,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=439530.0, ans=0.125
2024-09-24 06:22:52,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=439576.6666666667, ans=0.025
2024-09-24 06:22:54,951 INFO [train.py:1198] (1/4) Epoch 25, batch 700, loss[loss=0.1972, ctc_loss=0.126, cr_loss=0.3559, over 16639.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.1348, cr_loss=0.3515, over 3279565.45 frames. ], batch size: 37, lr: 4.82e-03, grad_scale: 16.0
2024-09-24 06:23:01,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=439623.3333333333, ans=0.2
2024-09-24 06:23:17,410 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.239e+02 1.345e+02 1.493e+02 2.005e+02, threshold=2.689e+02, percent-clipped=0.0
2024-09-24 06:23:23,323 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=4.25 vs. limit=15.0
2024-09-24 06:23:31,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=439716.6666666667, ans=0.0
2024-09-24 06:23:46,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=439763.3333333333, ans=0.0
2024-09-24 06:23:52,232 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 06:24:01,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=439810.0, ans=0.0
2024-09-24 06:24:08,655 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.22 vs. limit=15.0
2024-09-24 06:24:17,399 INFO [train.py:1198] (1/4) Epoch 25, batch 750, loss[loss=0.1853, ctc_loss=0.1202, cr_loss=0.3256, over 17147.00 frames. ], tot_loss[loss=0.2055, ctc_loss=0.1352, cr_loss=0.3516, over 3297137.39 frames. ], batch size: 48, lr: 4.81e-03, grad_scale: 16.0
2024-09-24 06:24:17,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=439856.6666666667, ans=0.0
2024-09-24 06:24:25,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=439856.6666666667, ans=0.5
2024-09-24 06:24:46,260 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 06:24:56,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=439950.0, ans=0.125
2024-09-24 06:24:59,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=439950.0, ans=0.125
2024-09-24 06:25:02,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=439950.0, ans=0.125
2024-09-24 06:25:08,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=439996.6666666667, ans=0.125
2024-09-24 06:25:19,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=440043.3333333333, ans=0.0
2024-09-24 06:25:24,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=440043.3333333333, ans=0.1
2024-09-24 06:25:28,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=440043.3333333333, ans=0.0
2024-09-24 06:25:37,713 INFO [train.py:1198] (1/4) Epoch 25, batch 800, loss[loss=0.1578, ctc_loss=0.09909, cr_loss=0.2933, over 17079.00 frames. ], tot_loss[loss=0.2067, ctc_loss=0.1361, cr_loss=0.3528, over 3313445.60 frames. ], batch size: 40, lr: 4.81e-03, grad_scale: 32.0
2024-09-24 06:25:49,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=440090.0, ans=0.05
2024-09-24 06:25:50,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=440090.0, ans=0.025
2024-09-24 06:26:02,615 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.255e+02 1.320e+02 1.414e+02 2.395e+02, threshold=2.641e+02, percent-clipped=0.0
2024-09-24 06:26:02,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=440136.6666666667, ans=0.1
2024-09-24 06:26:15,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=440183.3333333333, ans=0.125
2024-09-24 06:26:31,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=440230.0, ans=0.1
2024-09-24 06:26:37,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=440230.0, ans=0.1
2024-09-24 06:26:47,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=440276.6666666667, ans=0.0
2024-09-24 06:26:50,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=440276.6666666667, ans=0.125
2024-09-24 06:27:01,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=440323.3333333333, ans=0.0
2024-09-24 06:27:03,008 INFO [train.py:1198] (1/4) Epoch 25, batch 850, loss[loss=0.2111, ctc_loss=0.1372, cr_loss=0.3696, over 17215.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1357, cr_loss=0.352, over 3335553.19 frames. ], batch size: 50, lr: 4.81e-03, grad_scale: 32.0
2024-09-24 06:27:11,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=440323.3333333333, ans=0.2
2024-09-24 06:27:24,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=440370.0, ans=0.125
2024-09-24 06:28:18,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=440510.0, ans=0.125
2024-09-24 06:28:19,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=440510.0, ans=0.0
2024-09-24 06:28:26,004 INFO [train.py:1198] (1/4) Epoch 25, batch 900, loss[loss=0.2192, ctc_loss=0.1447, cr_loss=0.3725, over 17317.00 frames. ], tot_loss[loss=0.2066, ctc_loss=0.136, cr_loss=0.3529, over 3338594.45 frames. ], batch size: 49, lr: 4.81e-03, grad_scale: 32.0
2024-09-24 06:28:50,994 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.308e+02 1.404e+02 1.529e+02 2.023e+02, threshold=2.807e+02, percent-clipped=0.0
2024-09-24 06:28:56,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=440603.3333333333, ans=0.125
2024-09-24 06:29:47,841 INFO [train.py:1198] (1/4) Epoch 25, batch 950, loss[loss=0.1831, ctc_loss=0.119, cr_loss=0.3205, over 17098.00 frames. ], tot_loss[loss=0.2065, ctc_loss=0.1359, cr_loss=0.3526, over 3352170.52 frames. ], batch size: 43, lr: 4.81e-03, grad_scale: 16.0
2024-09-24 06:30:17,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.11 vs. limit=15.0
2024-09-24 06:31:09,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=440976.6666666667, ans=0.035
2024-09-24 06:31:12,904 INFO [train.py:1198] (1/4) Epoch 25, batch 1000, loss[loss=0.2317, ctc_loss=0.1536, cr_loss=0.3906, over 17300.00 frames. ], tot_loss[loss=0.2066, ctc_loss=0.136, cr_loss=0.353, over 3357967.47 frames. ], batch size: 51, lr: 4.81e-03, grad_scale: 16.0
2024-09-24 06:31:36,533 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.287e+02 1.404e+02 1.496e+02 1.832e+02, threshold=2.807e+02, percent-clipped=0.0
2024-09-24 06:31:55,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.56 vs. limit=15.0
2024-09-24 06:31:56,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=441116.6666666667, ans=0.125
2024-09-24 06:32:01,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=441163.3333333333, ans=0.125
2024-09-24 06:32:06,127 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.37 vs. limit=15.0
2024-09-24 06:32:21,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=441210.0, ans=0.07
2024-09-24 06:32:29,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=441210.0, ans=0.1
2024-09-24 06:32:32,883 INFO [train.py:1198] (1/4) Epoch 25, batch 1050, loss[loss=0.1918, ctc_loss=0.1265, cr_loss=0.3266, over 17170.00 frames. ], tot_loss[loss=0.2066, ctc_loss=0.1362, cr_loss=0.3522, over 3351975.82 frames. ], batch size: 45, lr: 4.81e-03, grad_scale: 16.0
2024-09-24 06:32:46,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=441256.6666666667, ans=0.2
2024-09-24 06:32:48,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=441256.6666666667, ans=0.2
2024-09-24 06:33:38,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=441443.3333333333, ans=0.0
2024-09-24 06:33:41,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=441443.3333333333, ans=0.0
2024-09-24 06:33:58,445 INFO [train.py:1198] (1/4) Epoch 25, batch 1100, loss[loss=0.2025, ctc_loss=0.1316, cr_loss=0.3546, over 17092.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1362, cr_loss=0.3531, over 3362821.04 frames. ], batch size: 49, lr: 4.81e-03, grad_scale: 16.0
2024-09-24 06:34:22,528 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.251e+02 1.334e+02 1.435e+02 1.725e+02, threshold=2.668e+02, percent-clipped=0.0
2024-09-24 06:34:40,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=441583.3333333333, ans=0.2
2024-09-24 06:35:11,325 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.13 vs. limit=15.0
2024-09-24 06:35:15,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=441676.6666666667, ans=0.125
2024-09-24 06:35:18,623 INFO [train.py:1198] (1/4) Epoch 25, batch 1150, loss[loss=0.2397, ctc_loss=0.1603, cr_loss=0.3971, over 16509.00 frames. ], tot_loss[loss=0.2076, ctc_loss=0.1368, cr_loss=0.354, over 3361457.86 frames. ], batch size: 66, lr: 4.80e-03, grad_scale: 16.0
2024-09-24 06:35:23,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=441723.3333333333, ans=0.2
2024-09-24 06:35:36,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=441770.0, ans=0.0
2024-09-24 06:35:39,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=441770.0, ans=0.0
2024-09-24 06:35:49,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=441816.6666666667, ans=0.125
2024-09-24 06:36:20,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=441863.3333333333, ans=0.0
2024-09-24 06:36:22,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=441863.3333333333, ans=0.0
2024-09-24 06:36:29,617 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.96 vs. limit=15.0
2024-09-24 06:36:42,700 INFO [train.py:1198] (1/4) Epoch 25, batch 1200, loss[loss=0.1661, ctc_loss=0.1093, cr_loss=0.284, over 16240.00 frames. ], tot_loss[loss=0.2062, ctc_loss=0.1358, cr_loss=0.352, over 3365907.16 frames. ], batch size: 36, lr: 4.80e-03, grad_scale: 32.0
2024-09-24 06:37:06,722 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.253e+02 1.340e+02 1.429e+02 1.909e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-24 06:37:37,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=442096.6666666667, ans=0.125
2024-09-24 06:37:38,977 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.85 vs. limit=10.0
2024-09-24 06:37:41,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=442096.6666666667, ans=0.125
2024-09-24 06:37:43,978 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.51 vs. limit=15.0
2024-09-24 06:38:05,401 INFO [train.py:1198] (1/4) Epoch 25, batch 1250, loss[loss=0.2505, ctc_loss=0.168, cr_loss=0.4121, over 16423.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1362, cr_loss=0.3529, over 3361723.67 frames. ], batch size: 66, lr: 4.80e-03, grad_scale: 32.0
2024-09-24 06:38:29,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=442236.6666666667, ans=0.04949747468305833
2024-09-24 06:38:44,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=442283.3333333333, ans=0.0
2024-09-24 06:39:00,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=442330.0, ans=0.125
2024-09-24 06:39:27,925 INFO [train.py:1198] (1/4) Epoch 25, batch 1300, loss[loss=0.1696, ctc_loss=0.108, cr_loss=0.3081, over 16945.00 frames. ], tot_loss[loss=0.2067, ctc_loss=0.136, cr_loss=0.3533, over 3367134.39 frames. ], batch size: 42, lr: 4.80e-03, grad_scale: 32.0
2024-09-24 06:39:45,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=442470.0, ans=0.0
2024-09-24 06:39:53,305 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.277e+02 1.396e+02 1.533e+02 2.196e+02, threshold=2.791e+02, percent-clipped=0.0
2024-09-24 06:39:53,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=442470.0, ans=0.1
2024-09-24 06:40:30,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=442610.0, ans=0.2
2024-09-24 06:40:30,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=442610.0, ans=0.0
2024-09-24 06:40:30,723 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.78 vs. limit=15.0
2024-09-24 06:40:36,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=442610.0, ans=0.025
2024-09-24 06:40:46,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=442656.6666666667, ans=0.2
2024-09-24 06:40:47,418 INFO [train.py:1198] (1/4) Epoch 25, batch 1350, loss[loss=0.1926, ctc_loss=0.124, cr_loss=0.3433, over 17031.00 frames. ], tot_loss[loss=0.2055, ctc_loss=0.1353, cr_loss=0.351, over 3363293.89 frames. ], batch size: 44, lr: 4.80e-03, grad_scale: 16.0
2024-09-24 06:40:52,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=442656.6666666667, ans=0.025
2024-09-24 06:41:05,623 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 06:41:18,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=442703.3333333333, ans=0.125
2024-09-24 06:41:48,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=442796.6666666667, ans=0.0
2024-09-24 06:41:51,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=442796.6666666667, ans=0.0
2024-09-24 06:41:52,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=442796.6666666667, ans=0.125
2024-09-24 06:41:53,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=442796.6666666667, ans=0.125
2024-09-24 06:41:58,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=442843.3333333333, ans=0.125
2024-09-24 06:42:10,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=442890.0, ans=0.1
2024-09-24 06:42:12,268 INFO [train.py:1198] (1/4) Epoch 25, batch 1400, loss[loss=0.2014, ctc_loss=0.1358, cr_loss=0.3281, over 17017.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1355, cr_loss=0.3518, over 3371671.21 frames. ], batch size: 51, lr: 4.80e-03, grad_scale: 16.0
2024-09-24 06:42:31,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=442936.6666666667, ans=0.0
2024-09-24 06:42:37,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=442936.6666666667, ans=0.125
2024-09-24 06:42:37,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=442936.6666666667, ans=0.1
2024-09-24 06:42:40,062 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.265e+02 1.378e+02 1.497e+02 2.360e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-24 06:42:41,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=442936.6666666667, ans=0.05
2024-09-24 06:42:57,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=442983.3333333333, ans=0.0
2024-09-24 06:42:57,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=442983.3333333333, ans=0.0
2024-09-24 06:43:04,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=443030.0, ans=0.1
2024-09-24 06:43:04,544 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.97 vs. limit=15.0
2024-09-24 06:43:15,398 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 06:43:18,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=443076.6666666667, ans=0.0
2024-09-24 06:43:29,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=443076.6666666667, ans=0.1
2024-09-24 06:43:36,562 INFO [train.py:1198] (1/4) Epoch 25, batch 1450, loss[loss=0.2201, ctc_loss=0.1431, cr_loss=0.385, over 16991.00 frames. ], tot_loss[loss=0.2071, ctc_loss=0.1366, cr_loss=0.3527, over 3354924.56 frames. ], batch size: 53, lr: 4.80e-03, grad_scale: 16.0
2024-09-24 06:43:38,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=443123.3333333333, ans=0.125
2024-09-24 06:43:49,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=443123.3333333333, ans=0.0
2024-09-24 06:43:54,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=443170.0, ans=0.125
2024-09-24 06:44:17,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=443216.6666666667, ans=0.125
2024-09-24 06:44:32,876 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.06 vs. limit=22.5
2024-09-24 06:44:55,969 INFO [train.py:1198] (1/4) Epoch 25, batch 1500, loss[loss=0.2136, ctc_loss=0.1383, cr_loss=0.3764, over 17021.00 frames. ], tot_loss[loss=0.2066, ctc_loss=0.1361, cr_loss=0.3523, over 3361648.03 frames. ], batch size: 44, lr: 4.80e-03, grad_scale: 16.0
2024-09-24 06:44:59,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=443356.6666666667, ans=0.125
2024-09-24 06:45:18,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=443403.3333333333, ans=0.1
2024-09-24 06:45:21,557 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.238e+02 1.333e+02 1.437e+02 1.693e+02, threshold=2.665e+02, percent-clipped=0.0
2024-09-24 06:45:39,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=443450.0, ans=0.125
2024-09-24 06:45:39,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.45 vs. limit=15.0
2024-09-24 06:46:21,157 INFO [train.py:1198] (1/4) Epoch 25, batch 1550, loss[loss=0.1824, ctc_loss=0.1177, cr_loss=0.3239, over 17191.00 frames. ], tot_loss[loss=0.2078, ctc_loss=0.137, cr_loss=0.354, over 3358846.70 frames. ], batch size: 41, lr: 4.79e-03, grad_scale: 16.0
2024-09-24 06:47:06,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=443683.3333333333, ans=0.125
2024-09-24 06:47:16,100 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=443730.0, ans=0.025
2024-09-24 06:47:43,800 INFO [train.py:1198] (1/4) Epoch 25, batch 1600, loss[loss=0.2049, ctc_loss=0.1343, cr_loss=0.3531, over 17152.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.1372, cr_loss=0.3539, over 3369768.70 frames. ], batch size: 48, lr: 4.79e-03, grad_scale: 32.0
2024-09-24 06:48:05,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=443870.0, ans=0.0
2024-09-24 06:48:09,684 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.242e+02 1.313e+02 1.472e+02 2.185e+02, threshold=2.626e+02, percent-clipped=0.0
2024-09-24 06:49:06,445 INFO [train.py:1198] (1/4) Epoch 25, batch 1650, loss[loss=0.1884, ctc_loss=0.1268, cr_loss=0.3082, over 17195.00 frames. ], tot_loss[loss=0.2072, ctc_loss=0.1366, cr_loss=0.3529, over 3372950.24 frames. ], batch size: 41, lr: 4.79e-03, grad_scale: 32.0
2024-09-24 06:49:13,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=444056.6666666667, ans=0.125
2024-09-24 06:49:22,021 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.63 vs. limit=15.0
2024-09-24 06:50:16,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=444243.3333333333, ans=0.0
2024-09-24 06:50:26,223 INFO [train.py:1198] (1/4) Epoch 25, batch 1700, loss[loss=0.1643, ctc_loss=0.1062, cr_loss=0.2904, over 16287.00 frames. ], tot_loss[loss=0.2066, ctc_loss=0.136, cr_loss=0.3527, over 3369809.42 frames. ], batch size: 36, lr: 4.79e-03, grad_scale: 32.0
2024-09-24 06:50:42,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=444336.6666666667, ans=0.125
2024-09-24 06:50:54,231 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.301e+02 1.388e+02 1.513e+02 1.890e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-24 06:51:51,033 INFO [train.py:1198] (1/4) Epoch 25, batch 1750, loss[loss=0.2252, ctc_loss=0.1489, cr_loss=0.3813, over 17300.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1349, cr_loss=0.3513, over 3378490.88 frames. ], batch size: 49, lr: 4.79e-03, grad_scale: 32.0
2024-09-24 06:52:16,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=444570.0, ans=0.025
2024-09-24 06:52:18,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=444570.0, ans=0.0
2024-09-24 06:52:55,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=444710.0, ans=0.0
2024-09-24 06:53:12,788 INFO [train.py:1198] (1/4) Epoch 25, batch 1800, loss[loss=0.1944, ctc_loss=0.124, cr_loss=0.3519, over 17025.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.134, cr_loss=0.3497, over 3379362.02 frames. ], batch size: 51, lr: 4.79e-03, grad_scale: 32.0
2024-09-24 06:53:21,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=444756.6666666667, ans=0.125
2024-09-24 06:53:40,996 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.233e+02 1.324e+02 1.428e+02 1.805e+02, threshold=2.648e+02, percent-clipped=0.0
2024-09-24 06:54:35,386 INFO [train.py:1198] (1/4) Epoch 25, batch 1850, loss[loss=0.1894, ctc_loss=0.1224, cr_loss=0.3348, over 17278.00 frames. ], tot_loss[loss=0.2044, ctc_loss=0.1344, cr_loss=0.3501, over 3378038.34 frames. ], batch size: 42, lr: 4.79e-03, grad_scale: 32.0
2024-09-24 06:54:43,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=444990.0, ans=0.025
2024-09-24 06:54:59,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=445036.6666666667, ans=0.125
2024-09-24 06:55:44,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=445176.6666666667, ans=0.125
2024-09-24 06:56:01,116 INFO [train.py:1198] (1/4) Epoch 25, batch 1900, loss[loss=0.1908, ctc_loss=0.1244, cr_loss=0.3316, over 17016.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.1345, cr_loss=0.3505, over 3384434.88 frames. ], batch size: 44, lr: 4.79e-03, grad_scale: 32.0
2024-09-24 06:56:09,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=445223.3333333333, ans=0.125
2024-09-24 06:56:21,311 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.78 vs. limit=15.0
2024-09-24 06:56:26,824 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.232e+02 1.312e+02 1.425e+02 2.956e+02, threshold=2.624e+02, percent-clipped=1.0
2024-09-24 06:56:42,382 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.45 vs. limit=22.5
2024-09-24 06:56:55,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=445363.3333333333, ans=0.0
2024-09-24 06:56:55,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=445363.3333333333, ans=0.1
2024-09-24 06:57:06,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=445410.0, ans=0.125
2024-09-24 06:57:20,853 INFO [train.py:1198] (1/4) Epoch 25, batch 1950, loss[loss=0.2018, ctc_loss=0.1339, cr_loss=0.3398, over 17031.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.1336, cr_loss=0.3484, over 3381305.07 frames. ], batch size: 52, lr: 4.78e-03, grad_scale: 32.0
2024-09-24 06:57:31,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=445456.6666666667, ans=0.125
2024-09-24 06:57:32,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=445456.6666666667, ans=0.1
2024-09-24 06:57:32,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=445456.6666666667, ans=0.1
2024-09-24 06:57:48,210 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.34 vs. limit=10.0
2024-09-24 06:57:51,670 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.01 vs. limit=12.0
2024-09-24 06:58:03,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=445550.0, ans=0.1
2024-09-24 06:58:35,374 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 06:58:38,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=445643.3333333333, ans=0.125
2024-09-24 06:58:40,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=445643.3333333333, ans=0.0
2024-09-24 06:58:46,270 INFO [train.py:1198] (1/4) Epoch 25, batch 2000, loss[loss=0.1885, ctc_loss=0.1233, cr_loss=0.3258, over 17023.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.1343, cr_loss=0.3499, over 3373657.11 frames. ], batch size: 44, lr: 4.78e-03, grad_scale: 32.0
2024-09-24 06:58:48,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=445690.0, ans=0.0
2024-09-24 06:59:11,910 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.272e+02 1.363e+02 1.482e+02 1.870e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-24 06:59:39,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=445830.0, ans=0.0
2024-09-24 07:00:04,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=445923.3333333333, ans=0.125
2024-09-24 07:00:06,179 INFO [train.py:1198] (1/4) Epoch 25, batch 2050, loss[loss=0.2296, ctc_loss=0.1524, cr_loss=0.386, over 17030.00 frames. ], tot_loss[loss=0.205, ctc_loss=0.1348, cr_loss=0.351, over 3376073.92 frames. ], batch size: 52, lr: 4.78e-03, grad_scale: 32.0
2024-09-24 07:00:06,905 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.08 vs. limit=15.0
2024-09-24 07:00:28,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.min_positive, batch_count=445970.0, ans=0.05
2024-09-24 07:00:50,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=446016.6666666667, ans=0.0
2024-09-24 07:00:57,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=446063.3333333333, ans=0.0
2024-09-24 07:01:07,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=446063.3333333333, ans=0.125
2024-09-24 07:01:09,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=446063.3333333333, ans=0.125
2024-09-24 07:01:11,311 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.28 vs. limit=22.5
2024-09-24 07:01:31,628 INFO [train.py:1198] (1/4) Epoch 25, batch 2100, loss[loss=0.2089, ctc_loss=0.1339, cr_loss=0.3752, over 17209.00 frames. ], tot_loss[loss=0.2063, ctc_loss=0.1357, cr_loss=0.353, over 3368267.30 frames. ], batch size: 47, lr: 4.78e-03, grad_scale: 32.0
2024-09-24 07:01:56,975 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.289e+02 1.372e+02 1.523e+02 2.426e+02, threshold=2.744e+02, percent-clipped=0.0
2024-09-24 07:02:37,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=446343.3333333333, ans=0.0
2024-09-24 07:02:44,978 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=446343.3333333333, ans=0.125
2024-09-24 07:02:54,490 INFO [train.py:1198] (1/4) Epoch 25, batch 2150, loss[loss=0.2081, ctc_loss=0.1377, cr_loss=0.352, over 17135.00 frames. ], tot_loss[loss=0.206, ctc_loss=0.1355, cr_loss=0.3525, over 3370234.40 frames. ], batch size: 48, lr: 4.78e-03, grad_scale: 32.0
2024-09-24 07:03:06,262 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=6.63 vs. limit=12.0
2024-09-24 07:03:45,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=446530.0, ans=0.025
2024-09-24 07:03:53,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=446530.0, ans=0.125
2024-09-24 07:04:05,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=446576.6666666667, ans=0.05
2024-09-24 07:04:08,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=446576.6666666667, ans=15.0
2024-09-24 07:04:17,541 INFO [train.py:1198] (1/4) Epoch 25, batch 2200, loss[loss=0.1833, ctc_loss=0.1165, cr_loss=0.3338, over 17203.00 frames. ], tot_loss[loss=0.2055, ctc_loss=0.1351, cr_loss=0.3522, over 3376071.73 frames. ], batch size: 41, lr: 4.78e-03, grad_scale: 16.0
2024-09-24 07:04:20,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=446623.3333333333, ans=0.125
2024-09-24 07:04:27,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=446623.3333333333, ans=0.0
2024-09-24 07:04:40,135 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:04:43,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=446670.0, ans=0.0
2024-09-24 07:04:44,644 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.273e+02 1.353e+02 1.559e+02 2.204e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-24 07:04:50,343 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.63 vs. limit=15.0
2024-09-24 07:04:57,067 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.52 vs. limit=6.0
2024-09-24 07:05:28,600 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:05:37,840 INFO [train.py:1198] (1/4) Epoch 25, batch 2250, loss[loss=0.1884, ctc_loss=0.122, cr_loss=0.3318, over 17121.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1349, cr_loss=0.3514, over 3372309.32 frames. ], batch size: 40, lr: 4.78e-03, grad_scale: 16.0
2024-09-24 07:07:03,314 INFO [train.py:1198] (1/4) Epoch 25, batch 2300, loss[loss=0.2175, ctc_loss=0.1401, cr_loss=0.3871, over 17300.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.1348, cr_loss=0.3513, over 3374286.93 frames. ], batch size: 49, lr: 4.78e-03, grad_scale: 16.0
2024-09-24 07:07:10,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=447090.0, ans=0.125
2024-09-24 07:07:30,511 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.028e+02 1.259e+02 1.322e+02 1.446e+02 1.963e+02, threshold=2.643e+02, percent-clipped=0.0
2024-09-24 07:08:28,419 INFO [train.py:1198] (1/4) Epoch 25, batch 2350, loss[loss=0.2166, ctc_loss=0.1442, cr_loss=0.3621, over 17072.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1356, cr_loss=0.3516, over 3365288.27 frames. ], batch size: 56, lr: 4.77e-03, grad_scale: 16.0
2024-09-24 07:08:30,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=447323.3333333333, ans=0.0
2024-09-24 07:08:35,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=447323.3333333333, ans=0.0
2024-09-24 07:08:36,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=447323.3333333333, ans=0.0
2024-09-24 07:08:46,473 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.99 vs. limit=22.5
2024-09-24 07:08:49,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=447370.0, ans=0.125
2024-09-24 07:08:57,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=447370.0, ans=0.125
2024-09-24 07:08:58,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=447416.6666666667, ans=0.125
2024-09-24 07:09:27,565 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:09:47,646 INFO [train.py:1198] (1/4) Epoch 25, batch 2400, loss[loss=0.1898, ctc_loss=0.1232, cr_loss=0.3328, over 17001.00 frames. ], tot_loss[loss=0.2079, ctc_loss=0.137, cr_loss=0.3545, over 3360262.22 frames. ], batch size: 44, lr: 4.77e-03, grad_scale: 32.0
2024-09-24 07:10:11,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=447603.3333333333, ans=0.1
2024-09-24 07:10:14,558 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.299e+02 1.395e+02 1.509e+02 2.801e+02, threshold=2.791e+02, percent-clipped=1.0
2024-09-24 07:10:26,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=447650.0, ans=0.125
2024-09-24 07:10:35,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=447696.6666666667, ans=0.04949747468305833
2024-09-24 07:10:45,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.72 vs. limit=15.0
2024-09-24 07:10:50,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=447696.6666666667, ans=0.125
2024-09-24 07:11:00,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=447743.3333333333, ans=0.025
2024-09-24 07:11:11,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=447790.0, ans=0.07
2024-09-24 07:11:12,576 INFO [train.py:1198] (1/4) Epoch 25, batch 2450, loss[loss=0.236, ctc_loss=0.1584, cr_loss=0.3881, over 16555.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.1371, cr_loss=0.3546, over 3357182.57 frames. ], batch size: 66, lr: 4.77e-03, grad_scale: 32.0
2024-09-24 07:11:22,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=447790.0, ans=0.125
2024-09-24 07:11:44,209 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.58 vs. limit=15.0
2024-09-24 07:11:58,437 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.73 vs. limit=10.0
2024-09-24 07:12:16,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=447976.6666666667, ans=0.125
2024-09-24 07:12:20,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=447976.6666666667, ans=0.125
2024-09-24 07:12:33,240 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.36 vs. limit=15.0
2024-09-24 07:12:36,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=448023.3333333333, ans=0.125
2024-09-24 07:12:37,501 INFO [train.py:1198] (1/4) Epoch 25, batch 2500, loss[loss=0.1917, ctc_loss=0.1249, cr_loss=0.3343, over 17357.00 frames. ], tot_loss[loss=0.2082, ctc_loss=0.1372, cr_loss=0.3549, over 3351509.77 frames. ], batch size: 48, lr: 4.77e-03, grad_scale: 32.0
2024-09-24 07:12:37,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=448023.3333333333, ans=0.125
2024-09-24 07:12:42,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=448023.3333333333, ans=0.0
2024-09-24 07:12:44,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=448023.3333333333, ans=0.125
2024-09-24 07:13:04,743 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.243e+02 1.319e+02 1.410e+02 2.410e+02, threshold=2.638e+02, percent-clipped=0.0
2024-09-24 07:13:11,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=448116.6666666667, ans=0.0
2024-09-24 07:13:13,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=448116.6666666667, ans=0.1
2024-09-24 07:13:29,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=448163.3333333333, ans=0.1
2024-09-24 07:13:38,148 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.45 vs. limit=22.5
2024-09-24 07:13:54,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=448210.0, ans=0.0
2024-09-24 07:13:58,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.39 vs. limit=10.0
2024-09-24 07:13:59,333 INFO [train.py:1198] (1/4) Epoch 25, batch 2550, loss[loss=0.2171, ctc_loss=0.1425, cr_loss=0.3726, over 17300.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1361, cr_loss=0.3536, over 3358840.35 frames. ], batch size: 46, lr: 4.77e-03, grad_scale: 32.0
2024-09-24 07:14:02,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=448256.6666666667, ans=0.125
2024-09-24 07:14:18,727 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:14:28,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=448303.3333333333, ans=0.1
2024-09-24 07:14:44,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=448350.0, ans=0.0
2024-09-24 07:14:49,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=448396.6666666667, ans=0.95
2024-09-24 07:14:50,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=448396.6666666667, ans=0.0
2024-09-24 07:15:08,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=448443.3333333333, ans=0.0
2024-09-24 07:15:10,488 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.16 vs. limit=15.0
2024-09-24 07:15:19,294 INFO [train.py:1198] (1/4) Epoch 25, batch 2600, loss[loss=0.1999, ctc_loss=0.1332, cr_loss=0.3333, over 17036.00 frames. ], tot_loss[loss=0.2074, ctc_loss=0.1365, cr_loss=0.3547, over 3353000.19 frames. ], batch size: 56, lr: 4.77e-03, grad_scale: 32.0
2024-09-24 07:15:28,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2.whitening_limit, batch_count=448490.0, ans=15.0
2024-09-24 07:15:29,985 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=7.97 vs. limit=22.5
2024-09-24 07:15:51,574 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.258e+02 1.378e+02 1.492e+02 2.205e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-24 07:16:06,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=448583.3333333333, ans=0.125
2024-09-24 07:16:27,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.21 vs. limit=12.0
2024-09-24 07:16:44,543 INFO [train.py:1198] (1/4) Epoch 25, batch 2650, loss[loss=0.1993, ctc_loss=0.1309, cr_loss=0.342, over 17072.00 frames. ], tot_loss[loss=0.2072, ctc_loss=0.1364, cr_loss=0.3542, over 3353429.11 frames. ], batch size: 46, lr: 4.77e-03, grad_scale: 32.0
2024-09-24 07:16:46,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=448723.3333333333, ans=0.125
2024-09-24 07:17:35,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=448863.3333333333, ans=0.125
2024-09-24 07:17:37,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=448863.3333333333, ans=0.1
2024-09-24 07:17:59,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=448910.0, ans=0.125
2024-09-24 07:18:10,103 INFO [train.py:1198] (1/4) Epoch 25, batch 2700, loss[loss=0.2143, ctc_loss=0.1386, cr_loss=0.3786, over 17150.00 frames. ], tot_loss[loss=0.2066, ctc_loss=0.1359, cr_loss=0.3531, over 3351897.79 frames. ], batch size: 48, lr: 4.77e-03, grad_scale: 32.0
2024-09-24 07:18:10,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer_ff3.min_abs, batch_count=448956.6666666667, ans=0.2
2024-09-24 07:18:13,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=448956.6666666667, ans=0.0
2024-09-24 07:18:24,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=449003.3333333333, ans=0.125
2024-09-24 07:18:31,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=449003.3333333333, ans=0.125
2024-09-24 07:18:37,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.249e+02 1.338e+02 1.409e+02 1.725e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-24 07:18:47,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=449050.0, ans=0.125
2024-09-24 07:18:48,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=449050.0, ans=0.0
2024-09-24 07:18:58,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=449096.6666666667, ans=0.0
2024-09-24 07:19:29,777 INFO [train.py:1198] (1/4) Epoch 25, batch 2750, loss[loss=0.2026, ctc_loss=0.1345, cr_loss=0.3404, over 17185.00 frames. ], tot_loss[loss=0.2073, ctc_loss=0.1366, cr_loss=0.3533, over 3348342.72 frames. ], batch size: 45, lr: 4.76e-03, grad_scale: 16.0
2024-09-24 07:19:39,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=449190.0, ans=0.125
2024-09-24 07:19:47,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=449236.6666666667, ans=0.125
2024-09-24 07:19:52,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=449236.6666666667, ans=0.1
2024-09-24 07:20:05,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=449283.3333333333, ans=0.04949747468305833
2024-09-24 07:20:55,049 INFO [train.py:1198] (1/4) Epoch 25, batch 2800, loss[loss=0.2132, ctc_loss=0.1402, cr_loss=0.365, over 17237.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1356, cr_loss=0.3523, over 3355836.42 frames. ], batch size: 44, lr: 4.76e-03, grad_scale: 32.0
2024-09-24 07:21:04,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=449423.3333333333, ans=0.125
2024-09-24 07:21:09,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=449470.0, ans=0.125
2024-09-24 07:21:19,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.37 vs. limit=15.0
2024-09-24 07:21:21,441 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.47 vs. limit=15.0
2024-09-24 07:21:23,916 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.252e+02 1.365e+02 1.477e+02 3.816e+02, threshold=2.730e+02, percent-clipped=1.0
2024-09-24 07:21:54,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=449563.3333333333, ans=0.2
2024-09-24 07:22:04,755 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.41 vs. limit=15.0
2024-09-24 07:22:06,713 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.84 vs. limit=15.0
2024-09-24 07:22:17,826 INFO [train.py:1198] (1/4) Epoch 25, batch 2850, loss[loss=0.1932, ctc_loss=0.1243, cr_loss=0.3442, over 17084.00 frames. ], tot_loss[loss=0.2062, ctc_loss=0.1357, cr_loss=0.3526, over 3356867.42 frames. ], batch size: 46, lr: 4.76e-03, grad_scale: 32.0
2024-09-24 07:22:18,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=449656.6666666667, ans=0.0
2024-09-24 07:22:26,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=449656.6666666667, ans=0.0
2024-09-24 07:22:27,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=449656.6666666667, ans=0.2
2024-09-24 07:22:43,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.min_positive, batch_count=449703.3333333333, ans=0.025
2024-09-24 07:23:40,782 INFO [train.py:1198] (1/4) Epoch 25, batch 2900, loss[loss=0.2138, ctc_loss=0.1381, cr_loss=0.3788, over 17216.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1361, cr_loss=0.3531, over 3344159.79 frames. ], batch size: 55, lr: 4.76e-03, grad_scale: 32.0
2024-09-24 07:23:44,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=449890.0, ans=0.1
2024-09-24 07:24:06,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=449936.6666666667, ans=0.1
2024-09-24 07:24:09,619 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.270e+02 1.360e+02 1.492e+02 4.410e+02, threshold=2.720e+02, percent-clipped=1.0
2024-09-24 07:24:19,971 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.11 vs. limit=12.0
2024-09-24 07:24:20,470 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.93 vs. limit=5.0
2024-09-24 07:25:00,612 INFO [train.py:1198] (1/4) Epoch 25, batch 2950, loss[loss=0.215, ctc_loss=0.1438, cr_loss=0.3561, over 17196.00 frames. ], tot_loss[loss=0.2071, ctc_loss=0.1364, cr_loss=0.3531, over 3349835.48 frames. ], batch size: 55, lr: 4.76e-03, grad_scale: 32.0
2024-09-24 07:25:02,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=450123.3333333333, ans=0.2
2024-09-24 07:25:15,841 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.66 vs. limit=12.0
2024-09-24 07:25:29,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=450170.0, ans=0.025
2024-09-24 07:25:46,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=450216.6666666667, ans=0.1
2024-09-24 07:26:00,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=450263.3333333333, ans=0.0
2024-09-24 07:26:06,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=450263.3333333333, ans=0.0
2024-09-24 07:26:08,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=450310.0, ans=0.0
2024-09-24 07:26:24,989 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.15 vs. limit=15.0
2024-09-24 07:26:25,451 INFO [train.py:1198] (1/4) Epoch 25, batch 3000, loss[loss=0.1743, ctc_loss=0.1108, cr_loss=0.3174, over 16943.00 frames. ], tot_loss[loss=0.2062, ctc_loss=0.1358, cr_loss=0.352, over 3350722.75 frames. ], batch size: 42, lr: 4.76e-03, grad_scale: 32.0
2024-09-24 07:26:25,452 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 07:26:41,273 INFO [train.py:1230] (1/4) Epoch 25, validation: loss=0.03749, ctc_loss=0.03749, cr_loss=8.201e-15, over 944034.00 frames. 
2024-09-24 07:26:41,274 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 07:26:41,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=450356.6666666667, ans=0.125
2024-09-24 07:26:45,578 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.34 vs. limit=22.5
2024-09-24 07:26:49,010 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.83 vs. limit=6.0
2024-09-24 07:26:49,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=450356.6666666667, ans=0.125
2024-09-24 07:27:09,644 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.216e+02 1.332e+02 1.454e+02 2.331e+02, threshold=2.665e+02, percent-clipped=0.0
2024-09-24 07:27:23,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=450450.0, ans=0.125
2024-09-24 07:27:38,385 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.70 vs. limit=15.0
2024-09-24 07:28:02,385 INFO [train.py:1198] (1/4) Epoch 25, batch 3050, loss[loss=0.1864, ctc_loss=0.1209, cr_loss=0.3273, over 17026.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1357, cr_loss=0.352, over 3348673.99 frames. ], batch size: 39, lr: 4.76e-03, grad_scale: 32.0
2024-09-24 07:28:12,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=450590.0, ans=0.125
2024-09-24 07:28:13,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=450590.0, ans=0.1
2024-09-24 07:28:30,601 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:29:03,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=450776.6666666667, ans=0.0
2024-09-24 07:29:16,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=450776.6666666667, ans=0.125
2024-09-24 07:29:21,016 INFO [train.py:1198] (1/4) Epoch 25, batch 3100, loss[loss=0.201, ctc_loss=0.1315, cr_loss=0.3478, over 17110.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1357, cr_loss=0.352, over 3350951.30 frames. ], batch size: 49, lr: 4.76e-03, grad_scale: 32.0
2024-09-24 07:29:42,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=450870.0, ans=0.125
2024-09-24 07:29:51,454 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.263e+02 1.349e+02 1.464e+02 5.566e+02, threshold=2.698e+02, percent-clipped=1.0
2024-09-24 07:29:57,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=450916.6666666667, ans=0.2
2024-09-24 07:30:19,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=450963.3333333333, ans=0.125
2024-09-24 07:30:38,552 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.89 vs. limit=12.0
2024-09-24 07:30:41,210 INFO [train.py:1198] (1/4) Epoch 25, batch 3150, loss[loss=0.2409, ctc_loss=0.1568, cr_loss=0.4204, over 17167.00 frames. ], tot_loss[loss=0.206, ctc_loss=0.1355, cr_loss=0.3525, over 3362651.93 frames. ], batch size: 45, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:31:12,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=451150.0, ans=0.125
2024-09-24 07:31:17,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=451150.0, ans=0.0
2024-09-24 07:31:40,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=451196.6666666667, ans=0.125
2024-09-24 07:31:42,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=451243.3333333333, ans=0.025
2024-09-24 07:31:58,814 INFO [train.py:1198] (1/4) Epoch 25, batch 3200, loss[loss=0.1955, ctc_loss=0.1276, cr_loss=0.3393, over 17035.00 frames. ], tot_loss[loss=0.2062, ctc_loss=0.1356, cr_loss=0.3531, over 3366222.40 frames. ], batch size: 44, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:32:05,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=451290.0, ans=0.0
2024-09-24 07:32:26,942 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.256e+02 1.371e+02 1.503e+02 1.985e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-24 07:32:36,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=451383.3333333333, ans=0.125
2024-09-24 07:32:39,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=451383.3333333333, ans=0.025
2024-09-24 07:33:03,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=451476.6666666667, ans=0.2
2024-09-24 07:33:16,639 INFO [train.py:1198] (1/4) Epoch 25, batch 3250, loss[loss=0.2217, ctc_loss=0.1442, cr_loss=0.3874, over 17220.00 frames. ], tot_loss[loss=0.2062, ctc_loss=0.1356, cr_loss=0.3532, over 3373040.63 frames. ], batch size: 50, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:33:29,842 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.14 vs. limit=22.5
2024-09-24 07:33:52,958 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.86 vs. limit=15.0
2024-09-24 07:33:57,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=451616.6666666667, ans=0.2
2024-09-24 07:34:04,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.23 vs. limit=15.0
2024-09-24 07:34:08,543 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.21 vs. limit=6.0
2024-09-24 07:34:34,603 INFO [train.py:1198] (1/4) Epoch 25, batch 3300, loss[loss=0.2188, ctc_loss=0.1416, cr_loss=0.3862, over 17073.00 frames. ], tot_loss[loss=0.2074, ctc_loss=0.1365, cr_loss=0.3545, over 3357866.46 frames. ], batch size: 46, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:34:51,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=451803.3333333333, ans=0.2
2024-09-24 07:35:04,442 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.271e+02 1.364e+02 1.475e+02 2.205e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-24 07:35:27,320 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.78 vs. limit=15.0
2024-09-24 07:35:47,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=451943.3333333333, ans=0.1
2024-09-24 07:35:56,274 INFO [train.py:1198] (1/4) Epoch 25, batch 3350, loss[loss=0.2441, ctc_loss=0.1634, cr_loss=0.4035, over 15995.00 frames. ], tot_loss[loss=0.2079, ctc_loss=0.1369, cr_loss=0.3549, over 3357897.89 frames. ], batch size: 74, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:36:04,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=451990.0, ans=0.125
2024-09-24 07:36:15,299 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=452036.6666666667, ans=0.0
2024-09-24 07:36:34,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=452083.3333333333, ans=0.125
2024-09-24 07:37:01,625 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.51 vs. limit=15.0
2024-09-24 07:37:14,295 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.90 vs. limit=15.0
2024-09-24 07:37:15,193 INFO [train.py:1198] (1/4) Epoch 25, batch 3400, loss[loss=0.2112, ctc_loss=0.1424, cr_loss=0.3437, over 16733.00 frames. ], tot_loss[loss=0.2078, ctc_loss=0.1368, cr_loss=0.3551, over 3357163.67 frames. ], batch size: 61, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:37:26,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=452223.3333333333, ans=0.125
2024-09-24 07:37:32,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=452270.0, ans=0.0
2024-09-24 07:37:34,326 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:37:39,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=452270.0, ans=0.125
2024-09-24 07:37:43,410 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.052e+02 1.278e+02 1.359e+02 1.525e+02 3.338e+02, threshold=2.719e+02, percent-clipped=1.0
2024-09-24 07:37:59,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=452316.6666666667, ans=0.125
2024-09-24 07:38:05,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=452363.3333333333, ans=0.1
2024-09-24 07:38:09,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=452363.3333333333, ans=0.0
2024-09-24 07:38:28,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=452410.0, ans=0.0
2024-09-24 07:38:35,006 INFO [train.py:1198] (1/4) Epoch 25, batch 3450, loss[loss=0.183, ctc_loss=0.1193, cr_loss=0.3185, over 17291.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.137, cr_loss=0.355, over 3359192.85 frames. ], batch size: 46, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:38:43,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=452456.6666666667, ans=0.0
2024-09-24 07:39:41,992 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.78 vs. limit=22.5
2024-09-24 07:39:45,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=452643.3333333333, ans=0.0
2024-09-24 07:39:55,459 INFO [train.py:1198] (1/4) Epoch 25, batch 3500, loss[loss=0.2313, ctc_loss=0.1552, cr_loss=0.3807, over 17202.00 frames. ], tot_loss[loss=0.2076, ctc_loss=0.1367, cr_loss=0.3548, over 3365993.75 frames. ], batch size: 47, lr: 4.75e-03, grad_scale: 32.0
2024-09-24 07:40:08,353 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:40:16,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=452736.6666666667, ans=0.025
2024-09-24 07:40:23,617 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.253e+02 1.349e+02 1.455e+02 2.168e+02, threshold=2.697e+02, percent-clipped=1.0
2024-09-24 07:40:26,057 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.24 vs. limit=15.0
2024-09-24 07:40:29,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.43 vs. limit=15.0
2024-09-24 07:40:41,529 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=8.15 vs. limit=15.0
2024-09-24 07:41:12,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=452923.3333333333, ans=0.2
2024-09-24 07:41:13,559 INFO [train.py:1198] (1/4) Epoch 25, batch 3550, loss[loss=0.1794, ctc_loss=0.1165, cr_loss=0.3143, over 17009.00 frames. ], tot_loss[loss=0.2073, ctc_loss=0.1365, cr_loss=0.3543, over 3366690.01 frames. ], batch size: 44, lr: 4.74e-03, grad_scale: 16.0
2024-09-24 07:41:35,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=452970.0, ans=0.125
2024-09-24 07:41:38,933 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.88 vs. limit=6.0
2024-09-24 07:41:44,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=453016.6666666667, ans=0.125
2024-09-24 07:41:55,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=453016.6666666667, ans=0.125
2024-09-24 07:42:09,536 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:42:13,549 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.56 vs. limit=15.0
2024-09-24 07:42:22,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=453110.0, ans=0.1
2024-09-24 07:42:25,930 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.78 vs. limit=15.0
2024-09-24 07:42:26,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=453110.0, ans=0.125
2024-09-24 07:42:31,018 INFO [train.py:1198] (1/4) Epoch 25, batch 3600, loss[loss=0.2342, ctc_loss=0.1553, cr_loss=0.3945, over 17040.00 frames. ], tot_loss[loss=0.2073, ctc_loss=0.1365, cr_loss=0.3538, over 3365551.99 frames. ], batch size: 53, lr: 4.74e-03, grad_scale: 32.0
2024-09-24 07:42:32,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=453156.6666666667, ans=0.0
2024-09-24 07:42:35,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=453156.6666666667, ans=0.1
2024-09-24 07:42:40,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=453156.6666666667, ans=0.125
2024-09-24 07:42:54,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=453203.3333333333, ans=0.125
2024-09-24 07:42:57,953 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.71 vs. limit=22.5
2024-09-24 07:43:00,636 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.246e+02 1.327e+02 1.467e+02 2.954e+02, threshold=2.655e+02, percent-clipped=1.0
2024-09-24 07:43:08,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=453250.0, ans=0.1
2024-09-24 07:43:38,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=453343.3333333333, ans=0.2
2024-09-24 07:43:41,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=453343.3333333333, ans=0.0
2024-09-24 07:43:41,707 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.82 vs. limit=15.0
2024-09-24 07:43:48,679 INFO [train.py:1198] (1/4) Epoch 25, batch 3650, loss[loss=0.2245, ctc_loss=0.1479, cr_loss=0.3829, over 17044.00 frames. ], tot_loss[loss=0.2063, ctc_loss=0.1358, cr_loss=0.3529, over 3370972.56 frames. ], batch size: 52, lr: 4.74e-03, grad_scale: 32.0
2024-09-24 07:44:09,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=453436.6666666667, ans=0.125
2024-09-24 07:44:17,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=453436.6666666667, ans=0.125
2024-09-24 07:44:54,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.min_positive, batch_count=453576.6666666667, ans=0.05
2024-09-24 07:45:04,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=453576.6666666667, ans=0.0
2024-09-24 07:45:11,866 INFO [train.py:1198] (1/4) Epoch 25, batch 3700, loss[loss=0.2442, ctc_loss=0.1708, cr_loss=0.367, over 11799.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1355, cr_loss=0.352, over 3363809.02 frames. ], batch size: 123, lr: 4.74e-03, grad_scale: 32.0
2024-09-24 07:45:41,667 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.237e+02 1.315e+02 1.374e+02 1.764e+02, threshold=2.629e+02, percent-clipped=0.0
2024-09-24 07:45:56,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=453716.6666666667, ans=0.025
2024-09-24 07:46:09,495 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=6.25 vs. limit=12.0
2024-09-24 07:46:15,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=453810.0, ans=0.0
2024-09-24 07:46:25,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=453810.0, ans=0.2
2024-09-24 07:46:30,215 INFO [train.py:1198] (1/4) Epoch 25, batch 3750, loss[loss=0.176, ctc_loss=0.1167, cr_loss=0.2964, over 16968.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1358, cr_loss=0.3514, over 3341331.91 frames. ], batch size: 42, lr: 4.74e-03, grad_scale: 32.0
2024-09-24 07:46:39,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=453856.6666666667, ans=0.0
2024-09-24 07:46:44,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=453903.3333333333, ans=0.125
2024-09-24 07:46:46,257 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 07:46:53,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=453903.3333333333, ans=0.125
2024-09-24 07:47:08,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=453950.0, ans=0.2
2024-09-24 07:47:08,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.57 vs. limit=6.0
2024-09-24 07:47:31,543 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.97 vs. limit=22.5
2024-09-24 07:47:47,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=454090.0, ans=0.125
2024-09-24 07:47:49,006 INFO [train.py:1198] (1/4) Epoch 25, batch 3800, loss[loss=0.1767, ctc_loss=0.1152, cr_loss=0.3078, over 17059.00 frames. ], tot_loss[loss=0.2067, ctc_loss=0.1363, cr_loss=0.3523, over 3334940.05 frames. ], batch size: 39, lr: 4.74e-03, grad_scale: 32.0
2024-09-24 07:48:08,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=454136.6666666667, ans=0.125
2024-09-24 07:48:11,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=454136.6666666667, ans=0.025
2024-09-24 07:48:11,874 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.44 vs. limit=6.0
2024-09-24 07:48:15,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=454136.6666666667, ans=0.125
2024-09-24 07:48:18,770 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.251e+02 1.351e+02 1.487e+02 2.041e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-24 07:48:19,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=454183.3333333333, ans=0.2
2024-09-24 07:48:20,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=454183.3333333333, ans=0.5
2024-09-24 07:48:22,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=454183.3333333333, ans=0.0
2024-09-24 07:48:33,902 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=13.77 vs. limit=15.0
2024-09-24 07:48:39,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=454230.0, ans=0.1
2024-09-24 07:48:41,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=454230.0, ans=0.125
2024-09-24 07:48:58,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=454276.6666666667, ans=0.0
2024-09-24 07:49:07,749 INFO [train.py:1198] (1/4) Epoch 25, batch 3850, loss[loss=0.2127, ctc_loss=0.1443, cr_loss=0.3419, over 16873.00 frames. ], tot_loss[loss=0.208, ctc_loss=0.1373, cr_loss=0.3535, over 3289408.59 frames. ], batch size: 58, lr: 4.74e-03, grad_scale: 16.0
2024-09-24 07:49:12,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=454323.3333333333, ans=0.0
2024-09-24 07:49:20,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=454323.3333333333, ans=0.0
2024-09-24 07:49:28,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=454370.0, ans=0.0
2024-09-24 07:49:41,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=454416.6666666667, ans=0.125
2024-09-24 07:49:43,795 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=10.25 vs. limit=15.0
2024-09-24 07:50:07,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=454510.0, ans=0.125
2024-09-24 07:51:09,081 INFO [train.py:1198] (1/4) Epoch 26, batch 0, loss[loss=0.1782, ctc_loss=0.1131, cr_loss=0.3256, over 16958.00 frames. ], tot_loss[loss=0.1782, ctc_loss=0.1131, cr_loss=0.3256, over 16958.00 frames. ], batch size: 42, lr: 4.64e-03, grad_scale: 32.0
2024-09-24 07:51:09,081 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 07:51:25,123 INFO [train.py:1230] (1/4) Epoch 26, validation: loss=0.03743, ctc_loss=0.03743, cr_loss=8.662e-15, over 944034.00 frames. 
2024-09-24 07:51:25,124 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 07:51:28,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=454538.0, ans=0.125
2024-09-24 07:51:36,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=454538.0, ans=0.125
2024-09-24 07:51:41,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=454584.6666666667, ans=0.1
2024-09-24 07:51:41,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=454584.6666666667, ans=0.0
2024-09-24 07:51:46,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=454584.6666666667, ans=0.025
2024-09-24 07:51:53,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=454584.6666666667, ans=0.125
2024-09-24 07:51:54,108 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.68 vs. limit=15.0
2024-09-24 07:52:01,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=454631.3333333333, ans=0.0
2024-09-24 07:52:06,030 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.313e+02 1.484e+02 1.654e+02 2.315e+02, threshold=2.969e+02, percent-clipped=0.0
2024-09-24 07:52:50,422 INFO [train.py:1198] (1/4) Epoch 26, batch 50, loss[loss=0.2065, ctc_loss=0.1367, cr_loss=0.3487, over 16195.00 frames. ], tot_loss[loss=0.2068, ctc_loss=0.1357, cr_loss=0.3556, over 759068.92 frames. ], batch size: 74, lr: 4.64e-03, grad_scale: 32.0
2024-09-24 07:53:35,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=454864.6666666667, ans=0.1
2024-09-24 07:53:39,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=454911.3333333333, ans=0.125
2024-09-24 07:53:40,205 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.23 vs. limit=15.0
2024-09-24 07:54:07,495 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.07 vs. limit=12.0
2024-09-24 07:54:08,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=454958.0, ans=0.125
2024-09-24 07:54:11,278 INFO [train.py:1198] (1/4) Epoch 26, batch 100, loss[loss=0.1828, ctc_loss=0.1182, cr_loss=0.3227, over 17248.00 frames. ], tot_loss[loss=0.2082, ctc_loss=0.1368, cr_loss=0.357, over 1341157.25 frames. ], batch size: 44, lr: 4.64e-03, grad_scale: 32.0
2024-09-24 07:54:38,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=455051.3333333333, ans=0.125
2024-09-24 07:54:51,770 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.228e+02 1.285e+02 1.398e+02 1.660e+02, threshold=2.570e+02, percent-clipped=0.0
2024-09-24 07:54:55,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=455098.0, ans=0.125
2024-09-24 07:55:20,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=455191.3333333333, ans=0.0
2024-09-24 07:55:27,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=455191.3333333333, ans=0.0
2024-09-24 07:55:31,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=455238.0, ans=10.0
2024-09-24 07:55:33,074 INFO [train.py:1198] (1/4) Epoch 26, batch 150, loss[loss=0.1893, ctc_loss=0.1245, cr_loss=0.3243, over 17154.00 frames. ], tot_loss[loss=0.207, ctc_loss=0.136, cr_loss=0.355, over 1781327.71 frames. ], batch size: 45, lr: 4.64e-03, grad_scale: 32.0
2024-09-24 07:55:47,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=455284.6666666667, ans=0.0
2024-09-24 07:55:49,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=455284.6666666667, ans=0.125
2024-09-24 07:55:51,476 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.47 vs. limit=22.5
2024-09-24 07:56:51,814 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.79 vs. limit=22.5
2024-09-24 07:56:55,948 INFO [train.py:1198] (1/4) Epoch 26, batch 200, loss[loss=0.225, ctc_loss=0.149, cr_loss=0.38, over 17344.00 frames. ], tot_loss[loss=0.2055, ctc_loss=0.1348, cr_loss=0.3533, over 2134882.57 frames. ], batch size: 48, lr: 4.64e-03, grad_scale: 32.0
2024-09-24 07:57:01,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=455471.3333333333, ans=0.1
2024-09-24 07:57:17,514 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.87 vs. limit=6.0
2024-09-24 07:57:37,053 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.254e+02 1.327e+02 1.395e+02 2.472e+02, threshold=2.655e+02, percent-clipped=0.0
2024-09-24 07:57:48,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=455611.3333333333, ans=0.125
2024-09-24 07:57:50,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=455611.3333333333, ans=0.07
2024-09-24 07:57:50,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.48 vs. limit=15.0
2024-09-24 07:57:58,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=455611.3333333333, ans=0.025
2024-09-24 07:58:03,512 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.31 vs. limit=15.0
2024-09-24 07:58:03,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.94 vs. limit=15.0
2024-09-24 07:58:12,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=455658.0, ans=0.125
2024-09-24 07:58:18,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=455658.0, ans=0.125
2024-09-24 07:58:21,704 INFO [train.py:1198] (1/4) Epoch 26, batch 250, loss[loss=0.2318, ctc_loss=0.1533, cr_loss=0.3925, over 17222.00 frames. ], tot_loss[loss=0.2065, ctc_loss=0.1356, cr_loss=0.3542, over 2393802.12 frames. ], batch size: 55, lr: 4.64e-03, grad_scale: 32.0
2024-09-24 07:58:23,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=455704.6666666667, ans=0.125
2024-09-24 07:59:05,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=455798.0, ans=0.125
2024-09-24 07:59:08,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=455844.6666666667, ans=0.025
2024-09-24 07:59:20,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=455844.6666666667, ans=0.0
2024-09-24 07:59:23,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=455844.6666666667, ans=0.125
2024-09-24 07:59:41,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=455891.3333333333, ans=0.125
2024-09-24 07:59:44,018 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=18.71 vs. limit=22.5
2024-09-24 07:59:44,774 INFO [train.py:1198] (1/4) Epoch 26, batch 300, loss[loss=0.1687, ctc_loss=0.1085, cr_loss=0.301, over 16950.00 frames. ], tot_loss[loss=0.2055, ctc_loss=0.1351, cr_loss=0.3522, over 2615614.71 frames. ], batch size: 42, lr: 4.64e-03, grad_scale: 32.0
2024-09-24 07:59:48,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=455938.0, ans=15.0
2024-09-24 07:59:56,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer_na.min_abs, batch_count=455938.0, ans=0.02
2024-09-24 08:00:22,960 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.258e+02 1.353e+02 1.431e+02 1.919e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-24 08:00:26,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=456031.3333333333, ans=0.0
2024-09-24 08:00:36,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=456078.0, ans=0.2
2024-09-24 08:00:56,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=456124.6666666667, ans=0.125
2024-09-24 08:01:04,576 INFO [train.py:1198] (1/4) Epoch 26, batch 350, loss[loss=0.2119, ctc_loss=0.1369, cr_loss=0.3748, over 17222.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.1342, cr_loss=0.3505, over 2788004.51 frames. ], batch size: 47, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:01:20,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=456218.0, ans=0.125
2024-09-24 08:01:27,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=456218.0, ans=0.125
2024-09-24 08:01:37,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten.whitening_limit, batch_count=456264.6666666667, ans=15.0
2024-09-24 08:01:57,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=456311.3333333333, ans=0.0
2024-09-24 08:02:16,429 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.33 vs. limit=22.5
2024-09-24 08:02:30,049 INFO [train.py:1198] (1/4) Epoch 26, batch 400, loss[loss=0.2257, ctc_loss=0.1485, cr_loss=0.3861, over 17231.00 frames. ], tot_loss[loss=0.2056, ctc_loss=0.135, cr_loss=0.353, over 2923329.65 frames. ], batch size: 50, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:02:36,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=456404.6666666667, ans=0.125
2024-09-24 08:03:11,992 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.93 vs. limit=15.0
2024-09-24 08:03:12,625 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.258e+02 1.321e+02 1.410e+02 2.001e+02, threshold=2.643e+02, percent-clipped=0.0
2024-09-24 08:03:14,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=456498.0, ans=0.125
2024-09-24 08:03:17,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=456498.0, ans=0.0
2024-09-24 08:03:52,933 INFO [train.py:1198] (1/4) Epoch 26, batch 450, loss[loss=0.1819, ctc_loss=0.1189, cr_loss=0.3152, over 17173.00 frames. ], tot_loss[loss=0.2049, ctc_loss=0.1345, cr_loss=0.352, over 3008133.40 frames. ], batch size: 45, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:03:53,669 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.82 vs. limit=22.5
2024-09-24 08:04:35,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=456731.3333333333, ans=0.125
2024-09-24 08:05:06,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=456824.6666666667, ans=0.1
2024-09-24 08:05:14,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.68 vs. limit=12.0
2024-09-24 08:05:15,525 INFO [train.py:1198] (1/4) Epoch 26, batch 500, loss[loss=0.2204, ctc_loss=0.1465, cr_loss=0.3699, over 17004.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1344, cr_loss=0.3509, over 3088689.17 frames. ], batch size: 51, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:05:22,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=456871.3333333333, ans=0.1
2024-09-24 08:05:26,107 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.63 vs. limit=15.0
2024-09-24 08:05:44,686 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=456918.0, ans=0.0
2024-09-24 08:05:55,348 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.259e+02 1.357e+02 1.518e+02 2.705e+02, threshold=2.714e+02, percent-clipped=1.0
2024-09-24 08:05:55,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=456964.6666666667, ans=0.125
2024-09-24 08:06:19,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=457058.0, ans=0.05
2024-09-24 08:06:37,535 INFO [train.py:1198] (1/4) Epoch 26, batch 550, loss[loss=0.2043, ctc_loss=0.133, cr_loss=0.3564, over 17211.00 frames. ], tot_loss[loss=0.2054, ctc_loss=0.135, cr_loss=0.3518, over 3143496.55 frames. ], batch size: 47, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:06:45,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=457104.6666666667, ans=0.125
2024-09-24 08:06:57,444 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:07:00,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.max_positive, batch_count=457151.3333333333, ans=0.95
2024-09-24 08:07:28,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=457244.6666666667, ans=0.0
2024-09-24 08:07:56,204 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.25 vs. limit=12.0
2024-09-24 08:08:00,421 INFO [train.py:1198] (1/4) Epoch 26, batch 600, loss[loss=0.2262, ctc_loss=0.1516, cr_loss=0.3728, over 17205.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1355, cr_loss=0.3529, over 3196661.04 frames. ], batch size: 50, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:08:21,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=457384.6666666667, ans=0.125
2024-09-24 08:08:35,445 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:08:42,969 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.231e+02 1.314e+02 1.456e+02 1.940e+02, threshold=2.629e+02, percent-clipped=0.0
2024-09-24 08:08:46,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=457431.3333333333, ans=0.0
2024-09-24 08:08:46,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=457431.3333333333, ans=0.125
2024-09-24 08:08:57,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=457478.0, ans=0.125
2024-09-24 08:09:02,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=457478.0, ans=0.125
2024-09-24 08:09:07,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=457524.6666666667, ans=0.125
2024-09-24 08:09:19,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=457524.6666666667, ans=0.125
2024-09-24 08:09:26,035 INFO [train.py:1198] (1/4) Epoch 26, batch 650, loss[loss=0.2329, ctc_loss=0.1551, cr_loss=0.3893, over 15928.00 frames. ], tot_loss[loss=0.2058, ctc_loss=0.1352, cr_loss=0.3531, over 3232791.53 frames. ], batch size: 74, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:09:44,380 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.61 vs. limit=15.0
2024-09-24 08:09:52,727 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.73 vs. limit=15.0
2024-09-24 08:09:58,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=457664.6666666667, ans=0.0
2024-09-24 08:10:06,602 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:10:46,655 INFO [train.py:1198] (1/4) Epoch 26, batch 700, loss[loss=0.2252, ctc_loss=0.1511, cr_loss=0.3705, over 17149.00 frames. ], tot_loss[loss=0.2063, ctc_loss=0.1357, cr_loss=0.353, over 3240672.58 frames. ], batch size: 48, lr: 4.63e-03, grad_scale: 32.0
2024-09-24 08:11:18,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=457898.0, ans=0.025
2024-09-24 08:11:28,400 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.269e+02 1.380e+02 1.511e+02 2.346e+02, threshold=2.761e+02, percent-clipped=0.0
2024-09-24 08:11:42,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=457944.6666666667, ans=0.125
2024-09-24 08:12:03,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=457991.3333333333, ans=0.125
2024-09-24 08:12:11,811 INFO [train.py:1198] (1/4) Epoch 26, batch 750, loss[loss=0.1814, ctc_loss=0.1163, cr_loss=0.3255, over 17061.00 frames. ], tot_loss[loss=0.2057, ctc_loss=0.1353, cr_loss=0.3518, over 3260874.78 frames. ], batch size: 46, lr: 4.63e-03, grad_scale: 16.0
2024-09-24 08:12:35,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=458084.6666666667, ans=0.125
2024-09-24 08:12:41,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.70 vs. limit=15.0
2024-09-24 08:12:49,453 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.51 vs. limit=15.0
2024-09-24 08:13:04,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=458178.0, ans=0.0
2024-09-24 08:13:12,612 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:13:15,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=458178.0, ans=0.125
2024-09-24 08:13:34,684 INFO [train.py:1198] (1/4) Epoch 26, batch 800, loss[loss=0.156, ctc_loss=0.1004, cr_loss=0.2782, over 17192.00 frames. ], tot_loss[loss=0.2071, ctc_loss=0.1364, cr_loss=0.3535, over 3277757.43 frames. ], batch size: 41, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:13:38,284 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:13:46,657 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.07 vs. limit=22.5
2024-09-24 08:14:04,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=458318.0, ans=0.0
2024-09-24 08:14:11,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.whiten.whitening_limit, batch_count=458364.6666666667, ans=15.0
2024-09-24 08:14:16,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=458364.6666666667, ans=0.125
2024-09-24 08:14:18,899 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.021e+02 1.286e+02 1.386e+02 1.460e+02 2.197e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-24 08:14:19,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=458364.6666666667, ans=0.125
2024-09-24 08:14:20,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=458364.6666666667, ans=0.05
2024-09-24 08:14:47,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=458458.0, ans=0.125
2024-09-24 08:14:57,079 INFO [train.py:1198] (1/4) Epoch 26, batch 850, loss[loss=0.2139, ctc_loss=0.1458, cr_loss=0.3405, over 11813.00 frames. ], tot_loss[loss=0.2066, ctc_loss=0.136, cr_loss=0.3531, over 3281708.68 frames. ], batch size: 123, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:14:59,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=458504.6666666667, ans=0.025
2024-09-24 08:15:13,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=458551.3333333333, ans=0.1
2024-09-24 08:15:47,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=458644.6666666667, ans=0.0
2024-09-24 08:16:17,230 INFO [train.py:1198] (1/4) Epoch 26, batch 900, loss[loss=0.2076, ctc_loss=0.1341, cr_loss=0.3676, over 17245.00 frames. ], tot_loss[loss=0.2056, ctc_loss=0.1352, cr_loss=0.352, over 3307402.85 frames. ], batch size: 44, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:16:29,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.14 vs. limit=15.0
2024-09-24 08:16:32,004 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:16:53,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=458831.3333333333, ans=0.125
2024-09-24 08:16:58,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=458831.3333333333, ans=0.125
2024-09-24 08:17:01,223 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.265e+02 1.347e+02 1.465e+02 1.812e+02, threshold=2.693e+02, percent-clipped=0.0
2024-09-24 08:17:05,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=458831.3333333333, ans=0.09899494936611666
2024-09-24 08:17:24,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.max_abs, batch_count=458924.6666666667, ans=10.0
2024-09-24 08:17:41,524 INFO [train.py:1198] (1/4) Epoch 26, batch 950, loss[loss=0.2144, ctc_loss=0.139, cr_loss=0.377, over 17091.00 frames. ], tot_loss[loss=0.2057, ctc_loss=0.1352, cr_loss=0.3523, over 3324498.01 frames. ], batch size: 49, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:18:21,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=459064.6666666667, ans=0.2
2024-09-24 08:19:04,838 INFO [train.py:1198] (1/4) Epoch 26, batch 1000, loss[loss=0.2113, ctc_loss=0.1392, cr_loss=0.3607, over 17295.00 frames. ], tot_loss[loss=0.2063, ctc_loss=0.1357, cr_loss=0.3529, over 3330301.99 frames. ], batch size: 51, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:19:06,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=459204.6666666667, ans=0.0
2024-09-24 08:19:23,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=459251.3333333333, ans=0.125
2024-09-24 08:19:25,716 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.93 vs. limit=22.5
2024-09-24 08:19:48,775 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.281e+02 1.338e+02 1.458e+02 2.157e+02, threshold=2.677e+02, percent-clipped=0.0
2024-09-24 08:20:27,260 INFO [train.py:1198] (1/4) Epoch 26, batch 1050, loss[loss=0.1528, ctc_loss=0.0963, cr_loss=0.2827, over 17172.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1353, cr_loss=0.3528, over 3341008.80 frames. ], batch size: 41, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:20:27,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=459438.0, ans=0.0
2024-09-24 08:20:32,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=459438.0, ans=0.125
2024-09-24 08:21:26,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=459578.0, ans=0.0
2024-09-24 08:21:37,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=459624.6666666667, ans=0.0
2024-09-24 08:21:50,046 INFO [train.py:1198] (1/4) Epoch 26, batch 1100, loss[loss=0.1997, ctc_loss=0.1317, cr_loss=0.3403, over 17229.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.1341, cr_loss=0.351, over 3350219.99 frames. ], batch size: 50, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:21:56,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=459671.3333333333, ans=0.125
2024-09-24 08:22:08,407 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.07 vs. limit=15.0
2024-09-24 08:22:28,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=459764.6666666667, ans=0.0
2024-09-24 08:22:34,488 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.255e+02 1.355e+02 1.464e+02 2.621e+02, threshold=2.711e+02, percent-clipped=0.0
2024-09-24 08:22:46,524 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.59 vs. limit=15.0
2024-09-24 08:23:02,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=459858.0, ans=0.025
2024-09-24 08:23:07,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=459858.0, ans=0.0
2024-09-24 08:23:11,449 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.37 vs. limit=15.0
2024-09-24 08:23:13,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=459904.6666666667, ans=0.1
2024-09-24 08:23:15,199 INFO [train.py:1198] (1/4) Epoch 26, batch 1150, loss[loss=0.1577, ctc_loss=0.1022, cr_loss=0.2779, over 17040.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1333, cr_loss=0.3494, over 3361738.27 frames. ], batch size: 39, lr: 4.62e-03, grad_scale: 32.0
2024-09-24 08:24:01,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=460044.6666666667, ans=0.2
2024-09-24 08:24:06,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=460044.6666666667, ans=0.0
2024-09-24 08:24:37,840 INFO [train.py:1198] (1/4) Epoch 26, batch 1200, loss[loss=0.2224, ctc_loss=0.1471, cr_loss=0.3761, over 17234.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.1339, cr_loss=0.3504, over 3358978.22 frames. ], batch size: 55, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:24:42,140 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.90 vs. limit=15.0
2024-09-24 08:24:51,876 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.71 vs. limit=15.0
2024-09-24 08:24:59,352 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.34 vs. limit=22.5
2024-09-24 08:25:13,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=460231.3333333333, ans=0.125
2024-09-24 08:25:19,715 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.237e+02 1.345e+02 1.469e+02 1.862e+02, threshold=2.691e+02, percent-clipped=0.0
2024-09-24 08:25:39,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=460278.0, ans=0.125
2024-09-24 08:25:45,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=460324.6666666667, ans=0.2
2024-09-24 08:25:54,380 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.71 vs. limit=15.0
2024-09-24 08:25:58,189 INFO [train.py:1198] (1/4) Epoch 26, batch 1250, loss[loss=0.2079, ctc_loss=0.1354, cr_loss=0.3627, over 17019.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.1348, cr_loss=0.3514, over 3353940.06 frames. ], batch size: 52, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:26:00,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=460371.3333333333, ans=0.025
2024-09-24 08:26:04,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=460371.3333333333, ans=0.0
2024-09-24 08:26:39,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=460464.6666666667, ans=0.125
2024-09-24 08:26:59,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=460511.3333333333, ans=0.125
2024-09-24 08:27:09,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=460558.0, ans=0.0
2024-09-24 08:27:17,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=460558.0, ans=0.07
2024-09-24 08:27:23,277 INFO [train.py:1198] (1/4) Epoch 26, batch 1300, loss[loss=0.1749, ctc_loss=0.112, cr_loss=0.3143, over 17081.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.135, cr_loss=0.3508, over 3348377.95 frames. ], batch size: 43, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:28:07,145 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.260e+02 1.346e+02 1.486e+02 2.192e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-24 08:28:10,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=460698.0, ans=0.035
2024-09-24 08:28:16,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=460744.6666666667, ans=0.125
2024-09-24 08:28:20,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=460744.6666666667, ans=0.0
2024-09-24 08:28:27,568 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.94 vs. limit=15.0
2024-09-24 08:28:33,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=460791.3333333333, ans=0.1
2024-09-24 08:28:45,545 INFO [train.py:1198] (1/4) Epoch 26, batch 1350, loss[loss=0.223, ctc_loss=0.1451, cr_loss=0.3898, over 17040.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.1349, cr_loss=0.3507, over 3341496.86 frames. ], batch size: 52, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:28:58,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=460838.0, ans=0.1
2024-09-24 08:29:21,520 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.70 vs. limit=15.0
2024-09-24 08:29:23,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=460931.3333333333, ans=0.0
2024-09-24 08:29:49,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=460978.0, ans=0.04949747468305833
2024-09-24 08:29:51,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=461024.6666666667, ans=0.125
2024-09-24 08:30:08,671 INFO [train.py:1198] (1/4) Epoch 26, batch 1400, loss[loss=0.2234, ctc_loss=0.1482, cr_loss=0.3759, over 17018.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.135, cr_loss=0.351, over 3343209.12 frames. ], batch size: 53, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:30:35,474 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.43 vs. limit=15.0
2024-09-24 08:30:40,240 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.36 vs. limit=15.0
2024-09-24 08:30:50,706 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.251e+02 1.327e+02 1.445e+02 2.357e+02, threshold=2.654e+02, percent-clipped=0.0
2024-09-24 08:30:52,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=461164.6666666667, ans=0.1
2024-09-24 08:31:18,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=461258.0, ans=0.125
2024-09-24 08:31:31,440 INFO [train.py:1198] (1/4) Epoch 26, batch 1450, loss[loss=0.2098, ctc_loss=0.1398, cr_loss=0.35, over 17220.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1355, cr_loss=0.3518, over 3342850.88 frames. ], batch size: 50, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:31:36,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=461304.6666666667, ans=0.05
2024-09-24 08:31:47,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=461351.3333333333, ans=0.05
2024-09-24 08:32:13,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=461398.0, ans=0.2
2024-09-24 08:32:18,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=461398.0, ans=0.1
2024-09-24 08:32:20,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=461444.6666666667, ans=0.0
2024-09-24 08:32:21,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=461444.6666666667, ans=0.0
2024-09-24 08:32:38,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.whiten.whitening_limit, batch_count=461491.3333333333, ans=12.0
2024-09-24 08:32:42,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=461491.3333333333, ans=0.0
2024-09-24 08:32:54,989 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.36 vs. limit=15.0
2024-09-24 08:32:55,983 INFO [train.py:1198] (1/4) Epoch 26, batch 1500, loss[loss=0.1956, ctc_loss=0.1263, cr_loss=0.3463, over 17265.00 frames. ], tot_loss[loss=0.2053, ctc_loss=0.1351, cr_loss=0.3508, over 3349142.50 frames. ], batch size: 42, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:33:05,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=461538.0, ans=0.1
2024-09-24 08:33:13,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=461584.6666666667, ans=0.07
2024-09-24 08:33:37,703 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.270e+02 1.352e+02 1.469e+02 1.916e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-24 08:33:46,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=461678.0, ans=0.0
2024-09-24 08:34:07,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=461724.6666666667, ans=0.0
2024-09-24 08:34:09,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=461724.6666666667, ans=0.125
2024-09-24 08:34:19,043 INFO [train.py:1198] (1/4) Epoch 26, batch 1550, loss[loss=0.1822, ctc_loss=0.1203, cr_loss=0.3094, over 17083.00 frames. ], tot_loss[loss=0.2048, ctc_loss=0.1347, cr_loss=0.3504, over 3354980.23 frames. ], batch size: 43, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:34:19,611 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=18.71 vs. limit=22.5
2024-09-24 08:34:20,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=461771.3333333333, ans=0.125
2024-09-24 08:34:41,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=461818.0, ans=0.0
2024-09-24 08:34:43,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=461818.0, ans=0.0
2024-09-24 08:34:56,279 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.81 vs. limit=6.0
2024-09-24 08:35:07,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=461911.3333333333, ans=0.125
2024-09-24 08:35:10,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=461911.3333333333, ans=15.0
2024-09-24 08:35:38,843 INFO [train.py:1198] (1/4) Epoch 26, batch 1600, loss[loss=0.2251, ctc_loss=0.1491, cr_loss=0.38, over 17049.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1346, cr_loss=0.3495, over 3350041.74 frames. ], batch size: 52, lr: 4.61e-03, grad_scale: 32.0
2024-09-24 08:35:44,349 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.94 vs. limit=15.0
2024-09-24 08:35:58,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=462051.3333333333, ans=0.0
2024-09-24 08:36:02,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=462051.3333333333, ans=0.125
2024-09-24 08:36:03,331 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.86 vs. limit=6.0
2024-09-24 08:36:22,545 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.247e+02 1.306e+02 1.406e+02 2.052e+02, threshold=2.612e+02, percent-clipped=0.0
2024-09-24 08:36:51,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=462191.3333333333, ans=0.125
2024-09-24 08:36:56,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=462191.3333333333, ans=0.125
2024-09-24 08:37:03,804 INFO [train.py:1198] (1/4) Epoch 26, batch 1650, loss[loss=0.1771, ctc_loss=0.1155, cr_loss=0.308, over 17191.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.1349, cr_loss=0.3508, over 3357714.56 frames. ], batch size: 41, lr: 4.60e-03, grad_scale: 32.0
2024-09-24 08:37:39,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=462331.3333333333, ans=0.125
2024-09-24 08:38:10,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=462424.6666666667, ans=0.2
2024-09-24 08:38:26,354 INFO [train.py:1198] (1/4) Epoch 26, batch 1700, loss[loss=0.2184, ctc_loss=0.1458, cr_loss=0.3632, over 17103.00 frames. ], tot_loss[loss=0.2044, ctc_loss=0.1344, cr_loss=0.35, over 3367044.53 frames. ], batch size: 49, lr: 4.60e-03, grad_scale: 32.0
2024-09-24 08:38:53,994 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.42 vs. limit=22.5
2024-09-24 08:39:10,601 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.222e+02 1.318e+02 1.444e+02 2.333e+02, threshold=2.637e+02, percent-clipped=0.0
2024-09-24 08:39:10,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=462564.6666666667, ans=0.125
2024-09-24 08:39:27,445 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.56 vs. limit=22.5
2024-09-24 08:39:31,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=462658.0, ans=0.0
2024-09-24 08:39:48,985 INFO [train.py:1198] (1/4) Epoch 26, batch 1750, loss[loss=0.2122, ctc_loss=0.1413, cr_loss=0.3542, over 17092.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.1345, cr_loss=0.3504, over 3363636.31 frames. ], batch size: 49, lr: 4.60e-03, grad_scale: 32.0
2024-09-24 08:39:50,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=462704.6666666667, ans=0.04949747468305833
2024-09-24 08:40:16,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=462751.3333333333, ans=0.2
2024-09-24 08:40:44,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=462844.6666666667, ans=0.125
2024-09-24 08:41:08,744 INFO [train.py:1198] (1/4) Epoch 26, batch 1800, loss[loss=0.1873, ctc_loss=0.1216, cr_loss=0.3284, over 17032.00 frames. ], tot_loss[loss=0.2053, ctc_loss=0.135, cr_loss=0.3516, over 3359563.44 frames. ], batch size: 44, lr: 4.60e-03, grad_scale: 32.0
2024-09-24 08:41:09,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=462938.0, ans=0.1
2024-09-24 08:41:18,294 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.60 vs. limit=22.5
2024-09-24 08:41:21,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=462938.0, ans=0.125
2024-09-24 08:41:31,361 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.72 vs. limit=15.0
2024-09-24 08:41:40,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=462984.6666666667, ans=0.125
2024-09-24 08:41:55,219 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.293e+02 1.406e+02 1.567e+02 1.918e+02, threshold=2.813e+02, percent-clipped=0.0
2024-09-24 08:42:30,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=463124.6666666667, ans=0.0
2024-09-24 08:42:33,470 INFO [train.py:1198] (1/4) Epoch 26, batch 1850, loss[loss=0.266, ctc_loss=0.1798, cr_loss=0.4312, over 15190.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1354, cr_loss=0.3528, over 3365946.34 frames. ], batch size: 89, lr: 4.60e-03, grad_scale: 16.0
2024-09-24 08:42:38,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=463171.3333333333, ans=0.125
2024-09-24 08:42:46,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=463171.3333333333, ans=0.125
2024-09-24 08:42:58,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=463218.0, ans=0.125
2024-09-24 08:43:06,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=463264.6666666667, ans=0.2
2024-09-24 08:43:12,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=463264.6666666667, ans=0.125
2024-09-24 08:43:23,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=463311.3333333333, ans=0.125
2024-09-24 08:43:28,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=463311.3333333333, ans=0.0
2024-09-24 08:43:37,297 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.17 vs. limit=15.0
2024-09-24 08:43:55,586 INFO [train.py:1198] (1/4) Epoch 26, batch 1900, loss[loss=0.2096, ctc_loss=0.1381, cr_loss=0.3578, over 17315.00 frames. ], tot_loss[loss=0.2058, ctc_loss=0.1353, cr_loss=0.3525, over 3362467.20 frames. ], batch size: 51, lr: 4.60e-03, grad_scale: 16.0
2024-09-24 08:44:09,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=463404.6666666667, ans=0.125
2024-09-24 08:44:23,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=463451.3333333333, ans=0.125
2024-09-24 08:44:25,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=463451.3333333333, ans=0.1
2024-09-24 08:44:29,036 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.24 vs. limit=6.0
2024-09-24 08:44:41,026 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.265e+02 1.341e+02 1.479e+02 3.030e+02, threshold=2.683e+02, percent-clipped=1.0
2024-09-24 08:44:43,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=463498.0, ans=0.0
2024-09-24 08:45:17,987 INFO [train.py:1198] (1/4) Epoch 26, batch 1950, loss[loss=0.2164, ctc_loss=0.1417, cr_loss=0.3736, over 17295.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1349, cr_loss=0.3517, over 3365538.14 frames. ], batch size: 46, lr: 4.60e-03, grad_scale: 16.0
2024-09-24 08:45:20,357 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.54 vs. limit=15.0
2024-09-24 08:46:08,008 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:46:40,693 INFO [train.py:1198] (1/4) Epoch 26, batch 2000, loss[loss=0.1871, ctc_loss=0.1222, cr_loss=0.3247, over 17171.00 frames. ], tot_loss[loss=0.2057, ctc_loss=0.1353, cr_loss=0.352, over 3359396.66 frames. ], batch size: 45, lr: 4.60e-03, grad_scale: 32.0
2024-09-24 08:46:46,303 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.80 vs. limit=15.0
2024-09-24 08:47:04,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=463918.0, ans=0.125
2024-09-24 08:47:26,726 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.293e+02 1.366e+02 1.489e+02 2.036e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-24 08:47:30,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=464011.3333333333, ans=0.1
2024-09-24 08:47:38,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=464011.3333333333, ans=0.0
2024-09-24 08:47:38,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=464011.3333333333, ans=0.0
2024-09-24 08:47:54,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=464058.0, ans=0.95
2024-09-24 08:48:06,082 INFO [train.py:1198] (1/4) Epoch 26, batch 2050, loss[loss=0.2037, ctc_loss=0.1311, cr_loss=0.3629, over 17304.00 frames. ], tot_loss[loss=0.2057, ctc_loss=0.1353, cr_loss=0.352, over 3367808.28 frames. ], batch size: 51, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:48:14,465 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:48:45,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=464198.0, ans=0.0
2024-09-24 08:48:45,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=464198.0, ans=0.0
2024-09-24 08:48:51,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=464198.0, ans=0.2
2024-09-24 08:48:53,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=464244.6666666667, ans=0.025
2024-09-24 08:49:28,856 INFO [train.py:1198] (1/4) Epoch 26, batch 2100, loss[loss=0.2026, ctc_loss=0.132, cr_loss=0.3529, over 16929.00 frames. ], tot_loss[loss=0.2056, ctc_loss=0.1351, cr_loss=0.3522, over 3365283.74 frames. ], batch size: 42, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:49:34,356 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.22 vs. limit=15.0
2024-09-24 08:49:53,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=464384.6666666667, ans=0.125
2024-09-24 08:50:12,345 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.252e+02 1.369e+02 1.506e+02 2.347e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-24 08:50:48,960 INFO [train.py:1198] (1/4) Epoch 26, batch 2150, loss[loss=0.2042, ctc_loss=0.136, cr_loss=0.3409, over 17137.00 frames. ], tot_loss[loss=0.2053, ctc_loss=0.1349, cr_loss=0.3522, over 3361131.26 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:50:50,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=464571.3333333333, ans=0.1
2024-09-24 08:50:52,783 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.87 vs. limit=22.5
2024-09-24 08:50:58,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=464571.3333333333, ans=0.125
2024-09-24 08:51:05,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=464618.0, ans=0.0
2024-09-24 08:51:09,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=464618.0, ans=0.125
2024-09-24 08:51:09,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=464618.0, ans=0.125
2024-09-24 08:51:25,186 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.49 vs. limit=15.0
2024-09-24 08:51:42,582 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 08:52:02,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=464758.0, ans=0.125
2024-09-24 08:52:13,761 INFO [train.py:1198] (1/4) Epoch 26, batch 2200, loss[loss=0.2346, ctc_loss=0.1554, cr_loss=0.3962, over 16954.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1348, cr_loss=0.3519, over 3349944.93 frames. ], batch size: 58, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:52:18,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=464804.6666666667, ans=0.125
2024-09-24 08:52:32,109 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.85 vs. limit=10.0
2024-09-24 08:52:37,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=464851.3333333333, ans=0.025
2024-09-24 08:52:44,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=464898.0, ans=0.1
2024-09-24 08:52:56,972 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.271e+02 1.361e+02 1.483e+02 2.576e+02, threshold=2.723e+02, percent-clipped=0.0
2024-09-24 08:53:07,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=464944.6666666667, ans=0.125
2024-09-24 08:53:12,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=464944.6666666667, ans=0.025
2024-09-24 08:53:14,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=464944.6666666667, ans=0.125
2024-09-24 08:53:14,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.72 vs. limit=15.0
2024-09-24 08:53:36,463 INFO [train.py:1198] (1/4) Epoch 26, batch 2250, loss[loss=0.1947, ctc_loss=0.1298, cr_loss=0.3247, over 17001.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1353, cr_loss=0.353, over 3355186.36 frames. ], batch size: 44, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:53:36,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=465038.0, ans=0.0
2024-09-24 08:54:06,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=465084.6666666667, ans=0.125
2024-09-24 08:54:20,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=465131.3333333333, ans=0.025
2024-09-24 08:54:22,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=465131.3333333333, ans=0.125
2024-09-24 08:54:52,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=465224.6666666667, ans=0.125
2024-09-24 08:54:59,008 INFO [train.py:1198] (1/4) Epoch 26, batch 2300, loss[loss=0.2197, ctc_loss=0.1462, cr_loss=0.3675, over 17313.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.1347, cr_loss=0.3517, over 3358155.78 frames. ], batch size: 49, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:55:31,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=465364.6666666667, ans=0.125
2024-09-24 08:55:38,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=465364.6666666667, ans=0.2
2024-09-24 08:55:39,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=465364.6666666667, ans=0.125
2024-09-24 08:55:42,650 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.269e+02 1.366e+02 1.477e+02 2.036e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-24 08:56:01,436 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.62 vs. limit=15.0
2024-09-24 08:56:22,145 INFO [train.py:1198] (1/4) Epoch 26, batch 2350, loss[loss=0.2025, ctc_loss=0.1296, cr_loss=0.3642, over 17209.00 frames. ], tot_loss[loss=0.2034, ctc_loss=0.1334, cr_loss=0.35, over 3358880.08 frames. ], batch size: 47, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:56:31,043 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.48 vs. limit=12.0
2024-09-24 08:56:40,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=465551.3333333333, ans=0.0
2024-09-24 08:56:58,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=465598.0, ans=0.0
2024-09-24 08:57:13,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=465644.6666666667, ans=0.0
2024-09-24 08:57:35,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=465691.3333333333, ans=0.1
2024-09-24 08:57:45,358 INFO [train.py:1198] (1/4) Epoch 26, batch 2400, loss[loss=0.2223, ctc_loss=0.1488, cr_loss=0.3674, over 15195.00 frames. ], tot_loss[loss=0.2042, ctc_loss=0.1341, cr_loss=0.3505, over 3357810.63 frames. ], batch size: 89, lr: 4.59e-03, grad_scale: 32.0
2024-09-24 08:58:12,763 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.43 vs. limit=12.0
2024-09-24 08:58:17,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=465784.6666666667, ans=0.025
2024-09-24 08:58:20,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=465831.3333333333, ans=0.1
2024-09-24 08:58:32,543 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.217e+02 1.274e+02 1.391e+02 1.998e+02, threshold=2.548e+02, percent-clipped=0.0
2024-09-24 08:58:34,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=465878.0, ans=0.125
2024-09-24 08:58:35,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=465878.0, ans=0.125
2024-09-24 08:58:56,069 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.80 vs. limit=15.0
2024-09-24 08:58:56,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=465924.6666666667, ans=0.0
2024-09-24 08:59:10,331 INFO [train.py:1198] (1/4) Epoch 26, batch 2450, loss[loss=0.1975, ctc_loss=0.1294, cr_loss=0.3409, over 17370.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1348, cr_loss=0.3521, over 3356010.70 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 16.0
2024-09-24 08:59:20,527 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.39 vs. limit=15.0
2024-09-24 08:59:30,426 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.77 vs. limit=10.0
2024-09-24 08:59:48,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=466064.6666666667, ans=0.2
2024-09-24 09:00:23,109 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.87 vs. limit=6.0
2024-09-24 09:00:30,249 INFO [train.py:1198] (1/4) Epoch 26, batch 2500, loss[loss=0.2205, ctc_loss=0.1447, cr_loss=0.3788, over 17252.00 frames. ], tot_loss[loss=0.2059, ctc_loss=0.1353, cr_loss=0.3532, over 3354929.12 frames. ], batch size: 44, lr: 4.58e-03, grad_scale: 16.0
2024-09-24 09:00:30,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=466204.6666666667, ans=0.2
2024-09-24 09:00:51,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=466251.3333333333, ans=0.0
2024-09-24 09:00:54,728 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 09:01:17,878 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.260e+02 1.339e+02 1.444e+02 2.110e+02, threshold=2.678e+02, percent-clipped=0.0
2024-09-24 09:01:18,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=466298.0, ans=0.025
2024-09-24 09:01:24,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=466344.6666666667, ans=0.125
2024-09-24 09:01:26,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=466344.6666666667, ans=0.0
2024-09-24 09:01:43,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=466391.3333333333, ans=0.125
2024-09-24 09:01:56,197 INFO [train.py:1198] (1/4) Epoch 26, batch 2550, loss[loss=0.1939, ctc_loss=0.1242, cr_loss=0.3487, over 17287.00 frames. ], tot_loss[loss=0.2065, ctc_loss=0.1357, cr_loss=0.3538, over 3353306.19 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 16.0
2024-09-24 09:02:18,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=466484.6666666667, ans=0.0
2024-09-24 09:02:34,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=466531.3333333333, ans=0.125
2024-09-24 09:02:39,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=466531.3333333333, ans=0.125
2024-09-24 09:02:39,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=466531.3333333333, ans=0.2
2024-09-24 09:02:41,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=466531.3333333333, ans=0.125
2024-09-24 09:03:20,622 INFO [train.py:1198] (1/4) Epoch 26, batch 2600, loss[loss=0.229, ctc_loss=0.1536, cr_loss=0.3769, over 16448.00 frames. ], tot_loss[loss=0.2075, ctc_loss=0.1364, cr_loss=0.3552, over 3360074.57 frames. ], batch size: 66, lr: 4.58e-03, grad_scale: 16.0
2024-09-24 09:03:46,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=466718.0, ans=0.125
2024-09-24 09:04:07,742 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.265e+02 1.355e+02 1.499e+02 2.594e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-24 09:04:27,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=466858.0, ans=0.0
2024-09-24 09:04:35,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=466858.0, ans=0.1
2024-09-24 09:04:42,886 INFO [train.py:1198] (1/4) Epoch 26, batch 2650, loss[loss=0.1756, ctc_loss=0.1116, cr_loss=0.3197, over 17186.00 frames. ], tot_loss[loss=0.2057, ctc_loss=0.1351, cr_loss=0.3531, over 3362386.80 frames. ], batch size: 41, lr: 4.58e-03, grad_scale: 16.0
2024-09-24 09:04:50,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=466904.6666666667, ans=0.07
2024-09-24 09:05:19,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=466998.0, ans=0.2
2024-09-24 09:06:00,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=467138.0, ans=0.0
2024-09-24 09:06:04,513 INFO [train.py:1198] (1/4) Epoch 26, batch 2700, loss[loss=0.202, ctc_loss=0.1326, cr_loss=0.3467, over 17298.00 frames. ], tot_loss[loss=0.2044, ctc_loss=0.1341, cr_loss=0.3517, over 3374193.69 frames. ], batch size: 51, lr: 4.58e-03, grad_scale: 16.0
2024-09-24 09:06:48,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=467231.3333333333, ans=0.125
2024-09-24 09:06:49,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=467231.3333333333, ans=0.0
2024-09-24 09:06:51,909 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.262e+02 1.341e+02 1.444e+02 3.624e+02, threshold=2.682e+02, percent-clipped=1.0
2024-09-24 09:07:11,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=467324.6666666667, ans=0.0
2024-09-24 09:07:27,271 INFO [train.py:1198] (1/4) Epoch 26, batch 2750, loss[loss=0.2195, ctc_loss=0.1421, cr_loss=0.3868, over 17297.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.134, cr_loss=0.3516, over 3379887.17 frames. ], batch size: 46, lr: 4.58e-03, grad_scale: 16.0
2024-09-24 09:07:46,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=467418.0, ans=0.125
2024-09-24 09:08:03,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=467464.6666666667, ans=0.0
2024-09-24 09:08:11,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=467464.6666666667, ans=0.125
2024-09-24 09:08:12,102 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.91 vs. limit=10.0
2024-09-24 09:08:15,594 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.30 vs. limit=22.5
2024-09-24 09:08:18,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=467511.3333333333, ans=0.0
2024-09-24 09:08:21,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=467511.3333333333, ans=0.125
2024-09-24 09:08:34,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=467558.0, ans=0.1
2024-09-24 09:08:35,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=467558.0, ans=0.0
2024-09-24 09:08:52,542 INFO [train.py:1198] (1/4) Epoch 26, batch 2800, loss[loss=0.2296, ctc_loss=0.1532, cr_loss=0.3819, over 16446.00 frames. ], tot_loss[loss=0.2047, ctc_loss=0.1343, cr_loss=0.3524, over 3381534.53 frames. ], batch size: 66, lr: 4.58e-03, grad_scale: 32.0
2024-09-24 09:09:37,848 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.261e+02 1.357e+02 1.478e+02 1.924e+02, threshold=2.714e+02, percent-clipped=0.0
2024-09-24 09:10:06,630 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=467791.3333333333, ans=0.0
2024-09-24 09:10:11,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=467838.0, ans=10.0
2024-09-24 09:10:12,861 INFO [train.py:1198] (1/4) Epoch 26, batch 2850, loss[loss=0.2004, ctc_loss=0.1331, cr_loss=0.3365, over 17023.00 frames. ], tot_loss[loss=0.2049, ctc_loss=0.1345, cr_loss=0.3519, over 3371858.97 frames. ], batch size: 51, lr: 4.58e-03, grad_scale: 32.0
2024-09-24 09:10:14,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=467838.0, ans=0.0
2024-09-24 09:10:14,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=467838.0, ans=0.125
2024-09-24 09:10:36,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=467884.6666666667, ans=0.125
2024-09-24 09:10:38,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=467884.6666666667, ans=0.1
2024-09-24 09:10:38,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=467884.6666666667, ans=10.0
2024-09-24 09:10:40,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=467884.6666666667, ans=0.05
2024-09-24 09:10:49,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=467931.3333333333, ans=0.1
2024-09-24 09:11:13,977 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.44 vs. limit=15.0
2024-09-24 09:11:14,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=467978.0, ans=0.09899494936611666
2024-09-24 09:11:24,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=468024.6666666667, ans=0.0
2024-09-24 09:11:25,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=468024.6666666667, ans=0.0
2024-09-24 09:11:35,003 INFO [train.py:1198] (1/4) Epoch 26, batch 2900, loss[loss=0.1915, ctc_loss=0.1237, cr_loss=0.3389, over 17169.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1342, cr_loss=0.3513, over 3363671.37 frames. ], batch size: 41, lr: 4.58e-03, grad_scale: 32.0
2024-09-24 09:11:48,096 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.46 vs. limit=15.0
2024-09-24 09:11:55,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=468118.0, ans=0.1
2024-09-24 09:12:06,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=468118.0, ans=0.2
2024-09-24 09:12:08,877 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.76 vs. limit=22.5
2024-09-24 09:12:22,358 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.321e+02 1.382e+02 1.512e+02 2.485e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-24 09:12:22,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=468164.6666666667, ans=0.0
2024-09-24 09:12:24,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=468211.3333333333, ans=0.025
2024-09-24 09:13:00,441 INFO [train.py:1198] (1/4) Epoch 26, batch 2950, loss[loss=0.2066, ctc_loss=0.1353, cr_loss=0.3564, over 17296.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1343, cr_loss=0.3512, over 3366091.01 frames. ], batch size: 49, lr: 4.57e-03, grad_scale: 32.0
2024-09-24 09:13:24,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=468351.3333333333, ans=0.1
2024-09-24 09:13:48,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=468398.0, ans=0.025
2024-09-24 09:13:53,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=468444.6666666667, ans=15.0
2024-09-24 09:13:57,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=468444.6666666667, ans=0.125
2024-09-24 09:14:02,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=468444.6666666667, ans=0.125
2024-09-24 09:14:23,227 INFO [train.py:1198] (1/4) Epoch 26, batch 3000, loss[loss=0.2112, ctc_loss=0.1382, cr_loss=0.3649, over 17024.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1336, cr_loss=0.3498, over 3367222.76 frames. ], batch size: 53, lr: 4.57e-03, grad_scale: 32.0
2024-09-24 09:14:23,228 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 09:14:38,572 INFO [train.py:1230] (1/4) Epoch 26, validation: loss=0.03742, ctc_loss=0.03742, cr_loss=8.706e-15, over 944034.00 frames. 
2024-09-24 09:14:38,573 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 09:14:43,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=468538.0, ans=0.09899494936611666
2024-09-24 09:14:55,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=468584.6666666667, ans=0.015
2024-09-24 09:15:02,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=468584.6666666667, ans=0.125
2024-09-24 09:15:22,386 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.270e+02 1.354e+02 1.456e+02 4.080e+02, threshold=2.708e+02, percent-clipped=1.0
2024-09-24 09:15:46,686 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.12 vs. limit=6.0
2024-09-24 09:15:53,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=468724.6666666667, ans=6.0
2024-09-24 09:15:56,790 INFO [train.py:1198] (1/4) Epoch 26, batch 3050, loss[loss=0.203, ctc_loss=0.1318, cr_loss=0.3558, over 16805.00 frames. ], tot_loss[loss=0.2031, ctc_loss=0.1334, cr_loss=0.3489, over 3360770.40 frames. ], batch size: 61, lr: 4.57e-03, grad_scale: 32.0
2024-09-24 09:16:33,526 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=6.80 vs. limit=15.0
2024-09-24 09:16:50,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=468911.3333333333, ans=0.125
2024-09-24 09:16:50,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=468911.3333333333, ans=0.0
2024-09-24 09:17:02,058 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.28 vs. limit=15.0
2024-09-24 09:17:14,927 INFO [train.py:1198] (1/4) Epoch 26, batch 3100, loss[loss=0.188, ctc_loss=0.1222, cr_loss=0.3286, over 17084.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.1334, cr_loss=0.3496, over 3362126.16 frames. ], batch size: 43, lr: 4.57e-03, grad_scale: 32.0
2024-09-24 09:17:21,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=469004.6666666667, ans=0.125
2024-09-24 09:17:30,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=469051.3333333333, ans=0.0
2024-09-24 09:17:31,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=469051.3333333333, ans=0.125
2024-09-24 09:18:01,342 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.241e+02 1.330e+02 1.444e+02 2.208e+02, threshold=2.660e+02, percent-clipped=0.0
2024-09-24 09:18:23,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=469191.3333333333, ans=0.125
2024-09-24 09:18:35,758 INFO [train.py:1198] (1/4) Epoch 26, batch 3150, loss[loss=0.1951, ctc_loss=0.1288, cr_loss=0.3315, over 17291.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1336, cr_loss=0.3496, over 3354736.98 frames. ], batch size: 49, lr: 4.57e-03, grad_scale: 32.0
2024-09-24 09:18:45,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=469238.0, ans=0.0
2024-09-24 09:19:06,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=469331.3333333333, ans=0.1
2024-09-24 09:19:14,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=469331.3333333333, ans=0.2
2024-09-24 09:19:37,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=469378.0, ans=0.1
2024-09-24 09:19:47,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=469424.6666666667, ans=0.125
2024-09-24 09:19:56,681 INFO [train.py:1198] (1/4) Epoch 26, batch 3200, loss[loss=0.2064, ctc_loss=0.1383, cr_loss=0.3407, over 17005.00 frames. ], tot_loss[loss=0.2041, ctc_loss=0.134, cr_loss=0.3503, over 3357565.44 frames. ], batch size: 51, lr: 4.57e-03, grad_scale: 32.0
2024-09-24 09:20:09,363 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.09 vs. limit=15.0
2024-09-24 09:20:35,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=469564.6666666667, ans=0.125
2024-09-24 09:20:44,065 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.250e+02 1.395e+02 1.507e+02 2.152e+02, threshold=2.790e+02, percent-clipped=0.0
2024-09-24 09:21:12,571 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.42 vs. limit=15.0
2024-09-24 09:21:15,230 INFO [train.py:1198] (1/4) Epoch 26, batch 3250, loss[loss=0.1968, ctc_loss=0.1278, cr_loss=0.345, over 17008.00 frames. ], tot_loss[loss=0.2055, ctc_loss=0.135, cr_loss=0.3525, over 3356036.01 frames. ], batch size: 51, lr: 4.57e-03, grad_scale: 16.0
2024-09-24 09:21:15,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=14.60 vs. limit=15.0
2024-09-24 09:21:18,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.max_abs, batch_count=469704.6666666667, ans=10.0
2024-09-24 09:21:38,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=469751.3333333333, ans=0.125
2024-09-24 09:21:49,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=469798.0, ans=0.125
2024-09-24 09:22:07,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=469844.6666666667, ans=0.125
2024-09-24 09:22:32,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=469891.3333333333, ans=0.125
2024-09-24 09:22:35,465 INFO [train.py:1198] (1/4) Epoch 26, batch 3300, loss[loss=0.2111, ctc_loss=0.1393, cr_loss=0.3589, over 17240.00 frames. ], tot_loss[loss=0.207, ctc_loss=0.136, cr_loss=0.3547, over 3347783.80 frames. ], batch size: 50, lr: 4.57e-03, grad_scale: 16.0
2024-09-24 09:22:35,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=469938.0, ans=0.1
2024-09-24 09:22:42,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=469938.0, ans=0.1
2024-09-24 09:22:49,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=469984.6666666667, ans=0.1
2024-09-24 09:22:53,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=469984.6666666667, ans=0.0
2024-09-24 09:22:56,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=469984.6666666667, ans=0.125
2024-09-24 09:23:04,509 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.95 vs. limit=10.0
2024-09-24 09:23:06,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer_na.min_abs, batch_count=470031.3333333333, ans=0.02
2024-09-24 09:23:16,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.30 vs. limit=15.0
2024-09-24 09:23:24,579 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.271e+02 1.364e+02 1.523e+02 2.164e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-24 09:23:25,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.34 vs. limit=22.5
2024-09-24 09:23:46,935 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.04 vs. limit=15.0
2024-09-24 09:23:55,639 INFO [train.py:1198] (1/4) Epoch 26, batch 3350, loss[loss=0.2084, ctc_loss=0.1377, cr_loss=0.3532, over 17237.00 frames. ], tot_loss[loss=0.207, ctc_loss=0.1361, cr_loss=0.3544, over 3346479.12 frames. ], batch size: 50, lr: 4.57e-03, grad_scale: 16.0
2024-09-24 09:24:03,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=470171.3333333333, ans=0.1
2024-09-24 09:24:45,690 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.93 vs. limit=6.0
2024-09-24 09:24:49,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=470311.3333333333, ans=0.05
2024-09-24 09:25:14,414 INFO [train.py:1198] (1/4) Epoch 26, batch 3400, loss[loss=0.1909, ctc_loss=0.1217, cr_loss=0.3459, over 17333.00 frames. ], tot_loss[loss=0.2064, ctc_loss=0.1357, cr_loss=0.3531, over 3350751.65 frames. ], batch size: 46, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:25:19,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=470404.6666666667, ans=0.025
2024-09-24 09:25:25,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=470404.6666666667, ans=0.025
2024-09-24 09:25:25,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=470404.6666666667, ans=0.125
2024-09-24 09:25:53,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=470498.0, ans=0.125
2024-09-24 09:26:01,626 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.254e+02 1.315e+02 1.422e+02 2.277e+02, threshold=2.630e+02, percent-clipped=0.0
2024-09-24 09:26:17,931 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.31 vs. limit=15.0
2024-09-24 09:26:32,735 INFO [train.py:1198] (1/4) Epoch 26, batch 3450, loss[loss=0.1818, ctc_loss=0.1174, cr_loss=0.3218, over 16300.00 frames. ], tot_loss[loss=0.206, ctc_loss=0.1355, cr_loss=0.3526, over 3351588.21 frames. ], batch size: 36, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:26:37,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=470638.0, ans=0.0
2024-09-24 09:26:39,476 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 09:26:39,628 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.34 vs. limit=15.0
2024-09-24 09:26:42,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=470638.0, ans=0.0
2024-09-24 09:26:42,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=470638.0, ans=0.0
2024-09-24 09:27:21,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=470778.0, ans=0.0
2024-09-24 09:27:23,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=470778.0, ans=0.0
2024-09-24 09:27:31,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=470778.0, ans=0.1
2024-09-24 09:27:31,506 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.98 vs. limit=6.0
2024-09-24 09:27:50,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=470824.6666666667, ans=0.2
2024-09-24 09:27:52,185 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.43 vs. limit=22.5
2024-09-24 09:27:53,198 INFO [train.py:1198] (1/4) Epoch 26, batch 3500, loss[loss=0.1714, ctc_loss=0.1113, cr_loss=0.3007, over 17211.00 frames. ], tot_loss[loss=0.2053, ctc_loss=0.135, cr_loss=0.3516, over 3362642.52 frames. ], batch size: 47, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:28:38,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=471011.3333333333, ans=0.025
2024-09-24 09:28:39,878 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.277e+02 1.357e+02 1.514e+02 2.797e+02, threshold=2.715e+02, percent-clipped=1.0
2024-09-24 09:28:51,528 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.47 vs. limit=15.0
2024-09-24 09:28:57,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=471058.0, ans=0.125
2024-09-24 09:28:57,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=54.19 vs. limit=15.0
2024-09-24 09:29:11,281 INFO [train.py:1198] (1/4) Epoch 26, batch 3550, loss[loss=0.1773, ctc_loss=0.1103, cr_loss=0.3349, over 17037.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1349, cr_loss=0.3515, over 3352794.85 frames. ], batch size: 39, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:29:13,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=471104.6666666667, ans=0.1
2024-09-24 09:29:38,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=471151.3333333333, ans=0.0
2024-09-24 09:29:49,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=471198.0, ans=0.2
2024-09-24 09:30:02,871 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.76 vs. limit=15.0
2024-09-24 09:30:32,070 INFO [train.py:1198] (1/4) Epoch 26, batch 3600, loss[loss=0.2276, ctc_loss=0.1507, cr_loss=0.3844, over 17193.00 frames. ], tot_loss[loss=0.205, ctc_loss=0.1347, cr_loss=0.3519, over 3364250.41 frames. ], batch size: 55, lr: 4.56e-03, grad_scale: 32.0
2024-09-24 09:30:40,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=471338.0, ans=0.0
2024-09-24 09:30:41,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=471338.0, ans=0.125
2024-09-24 09:30:44,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=471338.0, ans=0.0
2024-09-24 09:30:50,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=471384.6666666667, ans=0.2
2024-09-24 09:31:15,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=471431.3333333333, ans=0.125
2024-09-24 09:31:15,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=471431.3333333333, ans=0.125
2024-09-24 09:31:20,017 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.297e+02 1.443e+02 1.634e+02 2.383e+02, threshold=2.886e+02, percent-clipped=0.0
2024-09-24 09:31:31,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=471478.0, ans=0.125
2024-09-24 09:31:48,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=471571.3333333333, ans=0.125
2024-09-24 09:31:49,546 INFO [train.py:1198] (1/4) Epoch 26, batch 3650, loss[loss=0.1916, ctc_loss=0.1214, cr_loss=0.3511, over 17093.00 frames. ], tot_loss[loss=0.2053, ctc_loss=0.1349, cr_loss=0.3521, over 3360616.98 frames. ], batch size: 43, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:31:55,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=471571.3333333333, ans=0.0
2024-09-24 09:32:30,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=471664.6666666667, ans=0.125
2024-09-24 09:32:43,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=471711.3333333333, ans=0.125
2024-09-24 09:32:46,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=471711.3333333333, ans=0.125
2024-09-24 09:33:00,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=471758.0, ans=0.025
2024-09-24 09:33:04,177 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.61 vs. limit=10.0
2024-09-24 09:33:08,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=471758.0, ans=0.0
2024-09-24 09:33:12,495 INFO [train.py:1198] (1/4) Epoch 26, batch 3700, loss[loss=0.241, ctc_loss=0.1578, cr_loss=0.4161, over 17303.00 frames. ], tot_loss[loss=0.2041, ctc_loss=0.134, cr_loss=0.3505, over 3361272.02 frames. ], batch size: 51, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:33:40,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=471851.3333333333, ans=0.0
2024-09-24 09:33:42,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=7.48 vs. limit=12.0
2024-09-24 09:34:01,194 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.253e+02 1.341e+02 1.434e+02 1.966e+02, threshold=2.682e+02, percent-clipped=0.0
2024-09-24 09:34:30,816 INFO [train.py:1198] (1/4) Epoch 26, batch 3750, loss[loss=0.2231, ctc_loss=0.1505, cr_loss=0.3632, over 16904.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1342, cr_loss=0.3517, over 3372444.31 frames. ], batch size: 58, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:34:34,491 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.74 vs. limit=22.5
2024-09-24 09:34:41,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=472038.0, ans=0.125
2024-09-24 09:35:47,118 INFO [train.py:1198] (1/4) Epoch 26, batch 3800, loss[loss=0.1927, ctc_loss=0.1248, cr_loss=0.3393, over 17013.00 frames. ], tot_loss[loss=0.2062, ctc_loss=0.1355, cr_loss=0.3536, over 3349206.88 frames. ], batch size: 51, lr: 4.56e-03, grad_scale: 16.0
2024-09-24 09:35:56,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=472271.3333333333, ans=0.125
2024-09-24 09:36:14,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=472318.0, ans=0.0
2024-09-24 09:36:20,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=472364.6666666667, ans=0.1
2024-09-24 09:36:22,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=472364.6666666667, ans=0.125
2024-09-24 09:36:34,124 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.296e+02 1.371e+02 1.525e+02 1.900e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-24 09:37:03,758 INFO [train.py:1198] (1/4) Epoch 26, batch 3850, loss[loss=0.1994, ctc_loss=0.1315, cr_loss=0.3397, over 17022.00 frames. ], tot_loss[loss=0.2098, ctc_loss=0.1384, cr_loss=0.3572, over 3300678.26 frames. ], batch size: 44, lr: 4.55e-03, grad_scale: 16.0
2024-09-24 09:37:20,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=472551.3333333333, ans=0.125
2024-09-24 09:37:26,054 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=8.67 vs. limit=15.0
2024-09-24 09:37:34,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=472598.0, ans=0.125
2024-09-24 09:37:36,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=472598.0, ans=0.125
2024-09-24 09:37:52,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=472644.6666666667, ans=0.0
2024-09-24 09:39:03,326 INFO [train.py:1198] (1/4) Epoch 27, batch 0, loss[loss=0.1795, ctc_loss=0.1148, cr_loss=0.3234, over 16281.00 frames. ], tot_loss[loss=0.1795, ctc_loss=0.1148, cr_loss=0.3234, over 16281.00 frames. ], batch size: 36, lr: 4.47e-03, grad_scale: 32.0
2024-09-24 09:39:03,327 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 09:39:13,747 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.0.layers.0.self_attn_weights, attn_weights_entropy = tensor([6.2665, 6.0006, 5.6277, 5.7344], device='cuda:1')
2024-09-24 09:39:20,228 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([3.8411, 3.2361, 3.5383, 3.4215], device='cuda:1')
2024-09-24 09:39:21,545 INFO [train.py:1230] (1/4) Epoch 27, validation: loss=0.03741, ctc_loss=0.03741, cr_loss=8.388e-15, over 944034.00 frames. 
2024-09-24 09:39:21,546 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 09:39:35,018 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.81 vs. limit=15.0
2024-09-24 09:39:53,457 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.77 vs. limit=15.0
2024-09-24 09:40:00,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=472812.6666666667, ans=0.125
2024-09-24 09:40:04,587 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.71 vs. limit=12.0
2024-09-24 09:40:06,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.83 vs. limit=6.0
2024-09-24 09:40:10,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=472859.3333333333, ans=0.0
2024-09-24 09:40:12,618 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.33 vs. limit=15.0
2024-09-24 09:40:13,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=472859.3333333333, ans=0.0
2024-09-24 09:40:18,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=472859.3333333333, ans=0.0
2024-09-24 09:40:21,652 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.362e+02 1.530e+02 1.663e+02 2.257e+02, threshold=3.060e+02, percent-clipped=0.0
2024-09-24 09:40:41,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=472906.0, ans=0.2
2024-09-24 09:40:45,608 INFO [train.py:1198] (1/4) Epoch 27, batch 50, loss[loss=0.1681, ctc_loss=0.1075, cr_loss=0.303, over 17167.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1325, cr_loss=0.3468, over 755993.21 frames. ], batch size: 41, lr: 4.47e-03, grad_scale: 32.0
2024-09-24 09:41:04,862 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=12.32 vs. limit=15.0
2024-09-24 09:41:13,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=472999.3333333333, ans=0.125
2024-09-24 09:41:24,159 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.53 vs. limit=15.0
2024-09-24 09:41:44,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=473092.6666666667, ans=0.0
2024-09-24 09:41:44,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=473092.6666666667, ans=0.1
2024-09-24 09:41:45,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=473092.6666666667, ans=0.0
2024-09-24 09:41:54,763 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.83 vs. limit=12.0
2024-09-24 09:41:55,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=473139.3333333333, ans=0.04949747468305833
2024-09-24 09:41:57,596 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.96 vs. limit=15.0
2024-09-24 09:42:04,848 INFO [train.py:1198] (1/4) Epoch 27, batch 100, loss[loss=0.2531, ctc_loss=0.1714, cr_loss=0.4086, over 14830.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1327, cr_loss=0.3495, over 1328717.62 frames. ], batch size: 89, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:42:45,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=473279.3333333333, ans=0.025
2024-09-24 09:43:03,942 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.054e+02 1.215e+02 1.307e+02 1.417e+02 1.891e+02, threshold=2.615e+02, percent-clipped=0.0
2024-09-24 09:43:09,023 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 09:43:18,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=473372.6666666667, ans=0.125
2024-09-24 09:43:28,090 INFO [train.py:1198] (1/4) Epoch 27, batch 150, loss[loss=0.1875, ctc_loss=0.1213, cr_loss=0.3308, over 17220.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.1342, cr_loss=0.3505, over 1770919.20 frames. ], batch size: 47, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:43:33,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=473419.3333333333, ans=0.07
2024-09-24 09:44:19,978 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=473559.3333333333, ans=22.5
2024-09-24 09:44:26,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=473559.3333333333, ans=0.125
2024-09-24 09:44:53,617 INFO [train.py:1198] (1/4) Epoch 27, batch 200, loss[loss=0.1668, ctc_loss=0.1077, cr_loss=0.2956, over 16999.00 frames. ], tot_loss[loss=0.2041, ctc_loss=0.1341, cr_loss=0.3503, over 2129367.78 frames. ], batch size: 39, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:45:20,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=473699.3333333333, ans=0.125
2024-09-24 09:45:22,437 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.24 vs. limit=12.0
2024-09-24 09:45:42,425 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.72 vs. limit=6.0
2024-09-24 09:45:43,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=473792.6666666667, ans=0.0
2024-09-24 09:45:44,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=473792.6666666667, ans=0.125
2024-09-24 09:45:47,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=473792.6666666667, ans=0.2
2024-09-24 09:45:47,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=473792.6666666667, ans=0.025
2024-09-24 09:45:52,247 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.251e+02 1.322e+02 1.422e+02 2.046e+02, threshold=2.645e+02, percent-clipped=0.0
2024-09-24 09:45:57,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=473792.6666666667, ans=0.125
2024-09-24 09:46:00,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=473839.3333333333, ans=0.1
2024-09-24 09:46:16,198 INFO [train.py:1198] (1/4) Epoch 27, batch 250, loss[loss=0.1804, ctc_loss=0.1224, cr_loss=0.29, over 17081.00 frames. ], tot_loss[loss=0.2037, ctc_loss=0.1337, cr_loss=0.35, over 2402199.32 frames. ], batch size: 43, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:46:40,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=473932.6666666667, ans=0.2
2024-09-24 09:46:45,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=473932.6666666667, ans=0.125
2024-09-24 09:46:49,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=473979.3333333333, ans=0.0
2024-09-24 09:47:02,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=474026.0, ans=0.1
2024-09-24 09:47:04,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=474026.0, ans=0.125
2024-09-24 09:47:18,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=474072.6666666667, ans=0.1
2024-09-24 09:47:20,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=474072.6666666667, ans=0.0
2024-09-24 09:47:25,429 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.95 vs. limit=22.5
2024-09-24 09:47:33,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=474072.6666666667, ans=0.1
2024-09-24 09:47:38,636 INFO [train.py:1198] (1/4) Epoch 27, batch 300, loss[loss=0.2104, ctc_loss=0.1399, cr_loss=0.3525, over 17230.00 frames. ], tot_loss[loss=0.2028, ctc_loss=0.133, cr_loss=0.3486, over 2616130.88 frames. ], batch size: 50, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:47:47,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=474119.3333333333, ans=0.1
2024-09-24 09:47:47,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=474119.3333333333, ans=0.1
2024-09-24 09:47:47,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=474119.3333333333, ans=0.125
2024-09-24 09:48:30,177 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.85 vs. limit=15.0
2024-09-24 09:48:35,476 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.322e+02 1.414e+02 1.594e+02 2.687e+02, threshold=2.828e+02, percent-clipped=1.0
2024-09-24 09:48:37,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=474259.3333333333, ans=0.2
2024-09-24 09:48:59,274 INFO [train.py:1198] (1/4) Epoch 27, batch 350, loss[loss=0.1811, ctc_loss=0.1171, cr_loss=0.32, over 16753.00 frames. ], tot_loss[loss=0.202, ctc_loss=0.1325, cr_loss=0.3476, over 2777808.94 frames. ], batch size: 61, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:49:01,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=474352.6666666667, ans=0.125
2024-09-24 09:49:49,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=474446.0, ans=0.1
2024-09-24 09:49:57,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=474492.6666666667, ans=0.0
2024-09-24 09:50:03,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=474492.6666666667, ans=0.04949747468305833
2024-09-24 09:50:27,285 INFO [train.py:1198] (1/4) Epoch 27, batch 400, loss[loss=0.2006, ctc_loss=0.1298, cr_loss=0.3536, over 17259.00 frames. ], tot_loss[loss=0.2018, ctc_loss=0.1323, cr_loss=0.3476, over 2917349.26 frames. ], batch size: 42, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:50:43,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=474632.6666666667, ans=0.2
2024-09-24 09:51:07,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=474679.3333333333, ans=0.0
2024-09-24 09:51:15,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=474726.0, ans=0.0
2024-09-24 09:51:21,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.min_positive, batch_count=474726.0, ans=0.025
2024-09-24 09:51:23,172 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.248e+02 1.347e+02 1.469e+02 2.188e+02, threshold=2.694e+02, percent-clipped=0.0
2024-09-24 09:51:39,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=474772.6666666667, ans=0.0
2024-09-24 09:51:47,497 INFO [train.py:1198] (1/4) Epoch 27, batch 450, loss[loss=0.19, ctc_loss=0.1211, cr_loss=0.3441, over 17206.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1323, cr_loss=0.3481, over 3016077.03 frames. ], batch size: 41, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:52:05,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=474866.0, ans=0.2
2024-09-24 09:52:09,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=474866.0, ans=0.2
2024-09-24 09:52:24,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer_na.min_abs, batch_count=474912.6666666667, ans=0.02
2024-09-24 09:52:24,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=474912.6666666667, ans=0.125
2024-09-24 09:52:54,833 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.09 vs. limit=15.0
2024-09-24 09:53:07,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=475006.0, ans=22.5
2024-09-24 09:53:09,777 INFO [train.py:1198] (1/4) Epoch 27, batch 500, loss[loss=0.1779, ctc_loss=0.1159, cr_loss=0.3101, over 17278.00 frames. ], tot_loss[loss=0.2031, ctc_loss=0.1332, cr_loss=0.3497, over 3085129.47 frames. ], batch size: 46, lr: 4.46e-03, grad_scale: 32.0
2024-09-24 09:53:31,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=475099.3333333333, ans=0.125
2024-09-24 09:54:06,519 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.284e+02 1.368e+02 1.499e+02 2.424e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-24 09:54:06,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=475192.6666666667, ans=0.025
2024-09-24 09:54:07,257 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.75 vs. limit=15.0
2024-09-24 09:54:33,117 INFO [train.py:1198] (1/4) Epoch 27, batch 550, loss[loss=0.1927, ctc_loss=0.1262, cr_loss=0.3323, over 16965.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1336, cr_loss=0.351, over 3147350.95 frames. ], batch size: 42, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 09:54:43,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=475286.0, ans=0.1
2024-09-24 09:54:48,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=475286.0, ans=0.125
2024-09-24 09:55:02,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=475332.6666666667, ans=0.0
2024-09-24 09:55:26,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=475426.0, ans=0.125
2024-09-24 09:55:57,941 INFO [train.py:1198] (1/4) Epoch 27, batch 600, loss[loss=0.2018, ctc_loss=0.1316, cr_loss=0.351, over 17299.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1327, cr_loss=0.3493, over 3200612.33 frames. ], batch size: 51, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 09:55:59,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.79 vs. limit=22.5
2024-09-24 09:56:17,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=475566.0, ans=0.09899494936611666
2024-09-24 09:56:21,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=475566.0, ans=0.1
2024-09-24 09:56:53,667 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.260e+02 1.326e+02 1.393e+02 1.864e+02, threshold=2.652e+02, percent-clipped=0.0
2024-09-24 09:57:06,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=475706.0, ans=0.025
2024-09-24 09:57:07,421 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.50 vs. limit=10.0
2024-09-24 09:57:07,571 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.84 vs. limit=15.0
2024-09-24 09:57:17,880 INFO [train.py:1198] (1/4) Epoch 27, batch 650, loss[loss=0.1779, ctc_loss=0.1136, cr_loss=0.3213, over 17012.00 frames. ], tot_loss[loss=0.2027, ctc_loss=0.1328, cr_loss=0.3494, over 3245309.84 frames. ], batch size: 51, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 09:57:48,305 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.67 vs. limit=6.0
2024-09-24 09:57:55,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=475846.0, ans=0.1
2024-09-24 09:58:11,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=475892.6666666667, ans=0.125
2024-09-24 09:58:17,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=475892.6666666667, ans=0.125
2024-09-24 09:58:20,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=475892.6666666667, ans=0.1
2024-09-24 09:58:39,736 INFO [train.py:1198] (1/4) Epoch 27, batch 700, loss[loss=0.2228, ctc_loss=0.1491, cr_loss=0.3686, over 17032.00 frames. ], tot_loss[loss=0.2028, ctc_loss=0.133, cr_loss=0.349, over 3257263.15 frames. ], batch size: 51, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 09:58:48,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=475986.0, ans=0.125
2024-09-24 09:58:51,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=475986.0, ans=0.07
2024-09-24 09:59:10,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=476079.3333333333, ans=0.025
2024-09-24 09:59:10,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=476079.3333333333, ans=0.125
2024-09-24 09:59:17,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=476079.3333333333, ans=0.125
2024-09-24 09:59:40,876 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.256e+02 1.371e+02 1.501e+02 2.687e+02, threshold=2.742e+02, percent-clipped=1.0
2024-09-24 09:59:43,137 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.44 vs. limit=22.5
2024-09-24 09:59:57,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=476172.6666666667, ans=0.125
2024-09-24 10:00:04,597 INFO [train.py:1198] (1/4) Epoch 27, batch 750, loss[loss=0.2445, ctc_loss=0.1662, cr_loss=0.3916, over 14941.00 frames. ], tot_loss[loss=0.2034, ctc_loss=0.1335, cr_loss=0.3495, over 3279461.94 frames. ], batch size: 89, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 10:00:04,999 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:00:36,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=476266.0, ans=0.1
2024-09-24 10:01:02,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=476359.3333333333, ans=0.5
2024-09-24 10:01:11,986 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.38 vs. limit=12.0
2024-09-24 10:01:27,180 INFO [train.py:1198] (1/4) Epoch 27, batch 800, loss[loss=0.16, ctc_loss=0.1024, cr_loss=0.288, over 17275.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.1335, cr_loss=0.3492, over 3304368.13 frames. ], batch size: 42, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 10:01:45,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=476499.3333333333, ans=0.1
2024-09-24 10:02:01,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=476546.0, ans=0.0
2024-09-24 10:02:23,975 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.235e+02 1.338e+02 1.405e+02 1.662e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-24 10:02:32,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=476639.3333333333, ans=0.0
2024-09-24 10:02:49,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=476686.0, ans=0.125
2024-09-24 10:02:50,860 INFO [train.py:1198] (1/4) Epoch 27, batch 850, loss[loss=0.158, ctc_loss=0.09999, cr_loss=0.2902, over 17188.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1339, cr_loss=0.3497, over 3309495.89 frames. ], batch size: 41, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 10:02:54,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys.whitening_limit, batch_count=476686.0, ans=6.0
2024-09-24 10:02:56,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=476686.0, ans=0.125
2024-09-24 10:02:56,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=476686.0, ans=0.07
2024-09-24 10:03:26,088 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:03:29,673 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.52 vs. limit=22.5
2024-09-24 10:03:47,374 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=10.54 vs. limit=12.0
2024-09-24 10:03:51,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=476826.0, ans=0.0
2024-09-24 10:03:56,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=476872.6666666667, ans=0.0
2024-09-24 10:04:10,718 INFO [train.py:1198] (1/4) Epoch 27, batch 900, loss[loss=0.2045, ctc_loss=0.1336, cr_loss=0.3544, over 17115.00 frames. ], tot_loss[loss=0.2051, ctc_loss=0.1349, cr_loss=0.3511, over 3300469.07 frames. ], batch size: 40, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 10:04:26,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=476919.3333333333, ans=0.1
2024-09-24 10:04:38,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=476966.0, ans=0.125
2024-09-24 10:05:14,403 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.238e+02 1.345e+02 1.503e+02 3.181e+02, threshold=2.691e+02, percent-clipped=1.0
2024-09-24 10:05:38,931 INFO [train.py:1198] (1/4) Epoch 27, batch 950, loss[loss=0.1886, ctc_loss=0.1203, cr_loss=0.3418, over 17187.00 frames. ], tot_loss[loss=0.203, ctc_loss=0.1333, cr_loss=0.3482, over 3304451.63 frames. ], batch size: 41, lr: 4.45e-03, grad_scale: 32.0
2024-09-24 10:05:39,926 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.11 vs. limit=12.0
2024-09-24 10:06:11,459 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.51 vs. limit=15.0
2024-09-24 10:06:33,034 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.86 vs. limit=15.0
2024-09-24 10:06:36,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=477292.6666666667, ans=0.125
2024-09-24 10:06:36,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=477292.6666666667, ans=0.2
2024-09-24 10:06:58,798 INFO [train.py:1198] (1/4) Epoch 27, batch 1000, loss[loss=0.1999, ctc_loss=0.1333, cr_loss=0.3334, over 17145.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1327, cr_loss=0.3474, over 3310670.75 frames. ], batch size: 48, lr: 4.44e-03, grad_scale: 8.0
2024-09-24 10:07:15,101 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:07:21,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=477432.6666666667, ans=0.125
2024-09-24 10:07:37,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=6.01 vs. limit=15.0
2024-09-24 10:07:51,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=477526.0, ans=0.1
2024-09-24 10:07:56,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=477526.0, ans=0.0
2024-09-24 10:07:59,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=477526.0, ans=0.125
2024-09-24 10:08:00,887 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.228e+02 1.308e+02 1.402e+02 1.814e+02, threshold=2.617e+02, percent-clipped=0.0
2024-09-24 10:08:01,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=477526.0, ans=0.0
2024-09-24 10:08:19,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=477572.6666666667, ans=0.025
2024-09-24 10:08:22,251 INFO [train.py:1198] (1/4) Epoch 27, batch 1050, loss[loss=0.2208, ctc_loss=0.146, cr_loss=0.374, over 17300.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1329, cr_loss=0.348, over 3327568.64 frames. ], batch size: 49, lr: 4.44e-03, grad_scale: 8.0
2024-09-24 10:08:35,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=477619.3333333333, ans=0.1
2024-09-24 10:08:35,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=477619.3333333333, ans=0.0
2024-09-24 10:08:35,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=477619.3333333333, ans=0.2
2024-09-24 10:08:51,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=477666.0, ans=0.1
2024-09-24 10:09:21,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=477759.3333333333, ans=0.125
2024-09-24 10:09:24,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=477759.3333333333, ans=0.0
2024-09-24 10:09:47,485 INFO [train.py:1198] (1/4) Epoch 27, batch 1100, loss[loss=0.1902, ctc_loss=0.1245, cr_loss=0.3287, over 17092.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1336, cr_loss=0.3498, over 3335977.21 frames. ], batch size: 40, lr: 4.44e-03, grad_scale: 8.0
2024-09-24 10:09:54,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=477852.6666666667, ans=0.1
2024-09-24 10:10:00,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=477852.6666666667, ans=0.125
2024-09-24 10:10:05,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=477899.3333333333, ans=0.125
2024-09-24 10:10:07,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=477899.3333333333, ans=0.125
2024-09-24 10:10:14,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=477899.3333333333, ans=0.125
2024-09-24 10:10:38,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=477992.6666666667, ans=0.125
2024-09-24 10:10:46,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=477992.6666666667, ans=10.0
2024-09-24 10:10:49,361 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.279e+02 1.377e+02 1.520e+02 1.966e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-24 10:11:10,470 INFO [train.py:1198] (1/4) Epoch 27, batch 1150, loss[loss=0.2013, ctc_loss=0.1326, cr_loss=0.3431, over 16702.00 frames. ], tot_loss[loss=0.2049, ctc_loss=0.1346, cr_loss=0.3517, over 3339550.41 frames. ], batch size: 61, lr: 4.44e-03, grad_scale: 8.0
2024-09-24 10:11:24,306 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.19 vs. limit=22.5
2024-09-24 10:11:35,787 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.51 vs. limit=15.0
2024-09-24 10:12:10,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=478226.0, ans=0.025
2024-09-24 10:12:27,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=478272.6666666667, ans=0.125
2024-09-24 10:12:33,337 INFO [train.py:1198] (1/4) Epoch 27, batch 1200, loss[loss=0.205, ctc_loss=0.1345, cr_loss=0.3524, over 17040.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1353, cr_loss=0.3537, over 3346733.12 frames. ], batch size: 56, lr: 4.44e-03, grad_scale: 16.0
2024-09-24 10:12:49,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=478366.0, ans=0.1
2024-09-24 10:13:18,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=478412.6666666667, ans=0.125
2024-09-24 10:13:32,437 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.237e+02 1.313e+02 1.395e+02 2.791e+02, threshold=2.627e+02, percent-clipped=1.0
2024-09-24 10:13:47,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=478506.0, ans=0.1
2024-09-24 10:13:53,064 INFO [train.py:1198] (1/4) Epoch 27, batch 1250, loss[loss=0.2293, ctc_loss=0.1542, cr_loss=0.3754, over 16997.00 frames. ], tot_loss[loss=0.2064, ctc_loss=0.1356, cr_loss=0.3543, over 3350875.43 frames. ], batch size: 53, lr: 4.44e-03, grad_scale: 16.0
2024-09-24 10:14:24,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=478599.3333333333, ans=0.0
2024-09-24 10:14:28,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=478646.0, ans=0.125
2024-09-24 10:14:34,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=478646.0, ans=0.0
2024-09-24 10:15:16,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=478739.3333333333, ans=0.125
2024-09-24 10:15:21,232 INFO [train.py:1198] (1/4) Epoch 27, batch 1300, loss[loss=0.2242, ctc_loss=0.1457, cr_loss=0.3925, over 17225.00 frames. ], tot_loss[loss=0.2061, ctc_loss=0.1352, cr_loss=0.3545, over 3358420.58 frames. ], batch size: 55, lr: 4.44e-03, grad_scale: 16.0
2024-09-24 10:15:29,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=478786.0, ans=0.2
2024-09-24 10:15:58,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=478879.3333333333, ans=0.0
2024-09-24 10:15:58,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=478879.3333333333, ans=0.0
2024-09-24 10:16:06,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=478879.3333333333, ans=0.1
2024-09-24 10:16:08,642 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=19.37 vs. limit=22.5
2024-09-24 10:16:11,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=478926.0, ans=0.2
2024-09-24 10:16:21,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=478926.0, ans=0.125
2024-09-24 10:16:22,436 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.264e+02 1.365e+02 1.488e+02 1.950e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-24 10:16:37,661 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.73 vs. limit=22.5
2024-09-24 10:16:41,760 INFO [train.py:1198] (1/4) Epoch 27, batch 1350, loss[loss=0.2014, ctc_loss=0.1327, cr_loss=0.3437, over 17041.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.134, cr_loss=0.3524, over 3368917.59 frames. ], batch size: 52, lr: 4.44e-03, grad_scale: 8.0
2024-09-24 10:16:54,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=479019.3333333333, ans=0.1
2024-09-24 10:17:07,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=479066.0, ans=0.125
2024-09-24 10:17:20,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=479112.6666666667, ans=0.125
2024-09-24 10:17:22,418 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=10.17 vs. limit=12.0
2024-09-24 10:17:47,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.93 vs. limit=15.0
2024-09-24 10:18:04,480 INFO [train.py:1198] (1/4) Epoch 27, batch 1400, loss[loss=0.2248, ctc_loss=0.1471, cr_loss=0.3887, over 16576.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1342, cr_loss=0.3517, over 3351206.53 frames. ], batch size: 66, lr: 4.44e-03, grad_scale: 8.0
2024-09-24 10:18:04,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=479252.6666666667, ans=0.04949747468305833
2024-09-24 10:18:25,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=479299.3333333333, ans=0.0
2024-09-24 10:18:53,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=479392.6666666667, ans=0.125
2024-09-24 10:18:53,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=15.04 vs. limit=15.0
2024-09-24 10:19:08,017 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.257e+02 1.359e+02 1.482e+02 2.377e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-24 10:19:09,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=479439.3333333333, ans=0.125
2024-09-24 10:19:27,140 INFO [train.py:1198] (1/4) Epoch 27, batch 1450, loss[loss=0.2152, ctc_loss=0.1425, cr_loss=0.3636, over 17024.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1336, cr_loss=0.3503, over 3347611.34 frames. ], batch size: 51, lr: 4.43e-03, grad_scale: 8.0
2024-09-24 10:19:30,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=479486.0, ans=0.1
2024-09-24 10:19:36,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=479486.0, ans=0.5
2024-09-24 10:19:41,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=479486.0, ans=0.2
2024-09-24 10:20:12,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=479579.3333333333, ans=0.0
2024-09-24 10:20:36,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=479672.6666666667, ans=0.125
2024-09-24 10:20:43,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=479672.6666666667, ans=0.1
2024-09-24 10:20:52,646 INFO [train.py:1198] (1/4) Epoch 27, batch 1500, loss[loss=0.1965, ctc_loss=0.126, cr_loss=0.3525, over 17209.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.134, cr_loss=0.3503, over 3335693.98 frames. ], batch size: 47, lr: 4.43e-03, grad_scale: 8.0
2024-09-24 10:20:59,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=479719.3333333333, ans=0.125
2024-09-24 10:21:01,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=479719.3333333333, ans=0.125
2024-09-24 10:21:22,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=479766.0, ans=0.2
2024-09-24 10:21:30,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=479812.6666666667, ans=0.125
2024-09-24 10:21:35,327 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:21:41,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=479859.3333333333, ans=0.2
2024-09-24 10:21:44,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=479859.3333333333, ans=0.0
2024-09-24 10:21:54,047 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.289e+02 1.371e+02 1.496e+02 2.046e+02, threshold=2.742e+02, percent-clipped=0.0
2024-09-24 10:21:54,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=479859.3333333333, ans=0.05
2024-09-24 10:22:09,504 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.99 vs. limit=10.0
2024-09-24 10:22:13,081 INFO [train.py:1198] (1/4) Epoch 27, batch 1550, loss[loss=0.24, ctc_loss=0.1682, cr_loss=0.3593, over 11507.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1343, cr_loss=0.3508, over 3341946.15 frames. ], batch size: 123, lr: 4.43e-03, grad_scale: 8.0
2024-09-24 10:22:15,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=479952.6666666667, ans=0.125
2024-09-24 10:22:21,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=479952.6666666667, ans=0.2
2024-09-24 10:22:23,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten.whitening_limit, batch_count=479952.6666666667, ans=22.5
2024-09-24 10:22:43,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=479999.3333333333, ans=0.0
2024-09-24 10:22:49,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=480046.0, ans=0.125
2024-09-24 10:22:51,757 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.44 vs. limit=15.0
2024-09-24 10:23:25,144 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:23:36,295 INFO [train.py:1198] (1/4) Epoch 27, batch 1600, loss[loss=0.2211, ctc_loss=0.1436, cr_loss=0.3875, over 16575.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.1342, cr_loss=0.3505, over 3336653.78 frames. ], batch size: 66, lr: 4.43e-03, grad_scale: 16.0
2024-09-24 10:23:54,438 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.54 vs. limit=15.0
2024-09-24 10:24:38,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=480326.0, ans=0.125
2024-09-24 10:24:41,898 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.022e+02 1.244e+02 1.330e+02 1.456e+02 2.026e+02, threshold=2.660e+02, percent-clipped=0.0
2024-09-24 10:24:42,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=480326.0, ans=0.125
2024-09-24 10:25:03,406 INFO [train.py:1198] (1/4) Epoch 27, batch 1650, loss[loss=0.2294, ctc_loss=0.1522, cr_loss=0.3862, over 16716.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1339, cr_loss=0.3503, over 3344306.25 frames. ], batch size: 61, lr: 4.43e-03, grad_scale: 16.0
2024-09-24 10:25:08,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=480419.3333333333, ans=0.125
2024-09-24 10:25:16,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=480419.3333333333, ans=0.2
2024-09-24 10:25:35,998 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.99 vs. limit=6.0
2024-09-24 10:25:43,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=480512.6666666667, ans=0.1
2024-09-24 10:25:55,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=480559.3333333333, ans=0.125
2024-09-24 10:26:23,698 INFO [train.py:1198] (1/4) Epoch 27, batch 1700, loss[loss=0.1647, ctc_loss=0.105, cr_loss=0.2981, over 17086.00 frames. ], tot_loss[loss=0.2034, ctc_loss=0.1335, cr_loss=0.3497, over 3342892.66 frames. ], batch size: 43, lr: 4.43e-03, grad_scale: 16.0
2024-09-24 10:26:36,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.50 vs. limit=6.0
2024-09-24 10:27:00,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=480746.0, ans=0.125
2024-09-24 10:27:04,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=480746.0, ans=0.125
2024-09-24 10:27:26,755 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.231e+02 1.318e+02 1.430e+02 1.905e+02, threshold=2.636e+02, percent-clipped=0.0
2024-09-24 10:27:30,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=480839.3333333333, ans=0.125
2024-09-24 10:27:41,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=480839.3333333333, ans=0.125
2024-09-24 10:27:45,875 INFO [train.py:1198] (1/4) Epoch 27, batch 1750, loss[loss=0.2481, ctc_loss=0.1665, cr_loss=0.4077, over 17219.00 frames. ], tot_loss[loss=0.2027, ctc_loss=0.1329, cr_loss=0.3492, over 3346728.48 frames. ], batch size: 55, lr: 4.43e-03, grad_scale: 16.0
2024-09-24 10:27:46,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=480886.0, ans=0.025
2024-09-24 10:27:49,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=480886.0, ans=0.125
2024-09-24 10:28:04,072 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.41 vs. limit=22.5
2024-09-24 10:28:22,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=480979.3333333333, ans=0.04949747468305833
2024-09-24 10:28:26,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=480979.3333333333, ans=0.125
2024-09-24 10:28:59,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=481072.6666666667, ans=0.125
2024-09-24 10:29:08,830 INFO [train.py:1198] (1/4) Epoch 27, batch 1800, loss[loss=0.2251, ctc_loss=0.1505, cr_loss=0.3731, over 17305.00 frames. ], tot_loss[loss=0.2042, ctc_loss=0.1339, cr_loss=0.3514, over 3348022.21 frames. ], batch size: 51, lr: 4.43e-03, grad_scale: 16.0
2024-09-24 10:29:23,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=481166.0, ans=0.0
2024-09-24 10:29:54,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=481212.6666666667, ans=0.125
2024-09-24 10:30:14,559 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.286e+02 1.379e+02 1.526e+02 2.061e+02, threshold=2.757e+02, percent-clipped=0.0
2024-09-24 10:30:24,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=481306.0, ans=0.0
2024-09-24 10:30:33,713 INFO [train.py:1198] (1/4) Epoch 27, batch 1850, loss[loss=0.2115, ctc_loss=0.1384, cr_loss=0.3651, over 17286.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.1341, cr_loss=0.3521, over 3354178.06 frames. ], batch size: 51, lr: 4.43e-03, grad_scale: 16.0
2024-09-24 10:30:35,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=481352.6666666667, ans=0.0
2024-09-24 10:31:12,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=481446.0, ans=0.0
2024-09-24 10:31:34,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=481492.6666666667, ans=0.2
2024-09-24 10:31:53,595 INFO [train.py:1198] (1/4) Epoch 27, batch 1900, loss[loss=0.2061, ctc_loss=0.1335, cr_loss=0.3628, over 17370.00 frames. ], tot_loss[loss=0.2049, ctc_loss=0.1344, cr_loss=0.3524, over 3357126.23 frames. ], batch size: 48, lr: 4.43e-03, grad_scale: 16.0
2024-09-24 10:32:32,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=481679.3333333333, ans=0.0
2024-09-24 10:32:38,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=481679.3333333333, ans=0.125
2024-09-24 10:32:48,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=481726.0, ans=0.0
2024-09-24 10:32:57,446 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.257e+02 1.350e+02 1.464e+02 2.291e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-24 10:33:02,615 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:33:10,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=481772.6666666667, ans=0.05
2024-09-24 10:33:16,470 INFO [train.py:1198] (1/4) Epoch 27, batch 1950, loss[loss=0.2532, ctc_loss=0.1716, cr_loss=0.408, over 14930.00 frames. ], tot_loss[loss=0.203, ctc_loss=0.133, cr_loss=0.35, over 3364025.53 frames. ], batch size: 89, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:33:36,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=481866.0, ans=0.0
2024-09-24 10:33:37,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=481866.0, ans=0.0
2024-09-24 10:33:41,120 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:33:57,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=481912.6666666667, ans=0.05
2024-09-24 10:34:05,075 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.67 vs. limit=22.5
2024-09-24 10:34:36,264 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 10:34:36,639 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.76 vs. limit=15.0
2024-09-24 10:34:42,090 INFO [train.py:1198] (1/4) Epoch 27, batch 2000, loss[loss=0.1913, ctc_loss=0.1247, cr_loss=0.3334, over 17193.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1332, cr_loss=0.3503, over 3373015.60 frames. ], batch size: 47, lr: 4.42e-03, grad_scale: 32.0
2024-09-24 10:34:56,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=482099.3333333333, ans=0.125
2024-09-24 10:34:59,123 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.55 vs. limit=15.0
2024-09-24 10:35:18,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=482146.0, ans=0.05
2024-09-24 10:35:38,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=482192.6666666667, ans=0.025
2024-09-24 10:35:39,392 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.02 vs. limit=12.0
2024-09-24 10:35:46,430 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.275e+02 1.345e+02 1.450e+02 1.969e+02, threshold=2.691e+02, percent-clipped=0.0
2024-09-24 10:36:01,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=482239.3333333333, ans=0.125
2024-09-24 10:36:04,144 INFO [train.py:1198] (1/4) Epoch 27, batch 2050, loss[loss=0.1808, ctc_loss=0.1158, cr_loss=0.3253, over 17304.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.1337, cr_loss=0.3513, over 3368868.98 frames. ], batch size: 49, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:36:15,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=482286.0, ans=0.125
2024-09-24 10:36:20,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=482332.6666666667, ans=0.0
2024-09-24 10:36:52,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=482426.0, ans=0.035
2024-09-24 10:37:03,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=482426.0, ans=0.0
2024-09-24 10:37:27,022 INFO [train.py:1198] (1/4) Epoch 27, batch 2100, loss[loss=0.2139, ctc_loss=0.1406, cr_loss=0.3663, over 16902.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1332, cr_loss=0.3498, over 3373635.40 frames. ], batch size: 58, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:37:39,354 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.84 vs. limit=10.0
2024-09-24 10:37:42,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=482566.0, ans=0.0
2024-09-24 10:38:07,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=482612.6666666667, ans=0.125
2024-09-24 10:38:15,823 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.54 vs. limit=15.0
2024-09-24 10:38:29,421 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.227e+02 1.308e+02 1.433e+02 1.787e+02, threshold=2.617e+02, percent-clipped=0.0
2024-09-24 10:38:36,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=482706.0, ans=0.125
2024-09-24 10:38:47,142 INFO [train.py:1198] (1/4) Epoch 27, batch 2150, loss[loss=0.1995, ctc_loss=0.1331, cr_loss=0.3321, over 17308.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1335, cr_loss=0.3503, over 3367493.56 frames. ], batch size: 51, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:38:50,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=482752.6666666667, ans=0.125
2024-09-24 10:38:56,082 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.18 vs. limit=15.0
2024-09-24 10:39:03,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=482752.6666666667, ans=0.0
2024-09-24 10:40:02,190 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.59 vs. limit=15.0
2024-09-24 10:40:09,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=482939.3333333333, ans=0.125
2024-09-24 10:40:11,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=482939.3333333333, ans=0.025
2024-09-24 10:40:14,291 INFO [train.py:1198] (1/4) Epoch 27, batch 2200, loss[loss=0.18, ctc_loss=0.1172, cr_loss=0.3136, over 17033.00 frames. ], tot_loss[loss=0.2042, ctc_loss=0.134, cr_loss=0.3512, over 3366655.75 frames. ], batch size: 39, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:40:19,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=482986.0, ans=0.0
2024-09-24 10:40:25,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=482986.0, ans=0.2
2024-09-24 10:41:16,366 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.255e+02 1.303e+02 1.378e+02 1.644e+02, threshold=2.606e+02, percent-clipped=0.0
2024-09-24 10:41:34,084 INFO [train.py:1198] (1/4) Epoch 27, batch 2250, loss[loss=0.2157, ctc_loss=0.1411, cr_loss=0.3728, over 16981.00 frames. ], tot_loss[loss=0.2035, ctc_loss=0.1334, cr_loss=0.3504, over 3352687.19 frames. ], batch size: 53, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:41:44,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=483219.3333333333, ans=10.0
2024-09-24 10:42:18,563 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.79 vs. limit=22.5
2024-09-24 10:42:45,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=483406.0, ans=0.0
2024-09-24 10:42:56,647 INFO [train.py:1198] (1/4) Epoch 27, batch 2300, loss[loss=0.2448, ctc_loss=0.1606, cr_loss=0.4208, over 17139.00 frames. ], tot_loss[loss=0.203, ctc_loss=0.133, cr_loss=0.3499, over 3368226.88 frames. ], batch size: 48, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:43:03,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=483452.6666666667, ans=0.125
2024-09-24 10:43:04,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=483452.6666666667, ans=0.1
2024-09-24 10:43:22,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=483499.3333333333, ans=0.0
2024-09-24 10:43:41,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=483546.0, ans=0.0
2024-09-24 10:43:41,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.11 vs. limit=15.0
2024-09-24 10:43:50,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=483592.6666666667, ans=0.125
2024-09-24 10:43:58,587 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.254e+02 1.343e+02 1.462e+02 2.563e+02, threshold=2.686e+02, percent-clipped=0.0
2024-09-24 10:43:58,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=483639.3333333333, ans=0.1
2024-09-24 10:44:18,919 INFO [train.py:1198] (1/4) Epoch 27, batch 2350, loss[loss=0.1927, ctc_loss=0.125, cr_loss=0.3388, over 17033.00 frames. ], tot_loss[loss=0.2029, ctc_loss=0.133, cr_loss=0.3495, over 3370049.09 frames. ], batch size: 44, lr: 4.42e-03, grad_scale: 16.0
2024-09-24 10:44:40,850 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=483732.6666666667, ans=0.0
2024-09-24 10:44:56,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=483779.3333333333, ans=0.125
2024-09-24 10:45:00,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=483779.3333333333, ans=0.0
2024-09-24 10:45:28,512 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.74 vs. limit=15.0
2024-09-24 10:45:43,652 INFO [train.py:1198] (1/4) Epoch 27, batch 2400, loss[loss=0.2153, ctc_loss=0.1467, cr_loss=0.343, over 15814.00 frames. ], tot_loss[loss=0.2018, ctc_loss=0.1321, cr_loss=0.3481, over 3368496.53 frames. ], batch size: 74, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:45:54,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=483919.3333333333, ans=0.0
2024-09-24 10:46:10,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=483966.0, ans=0.125
2024-09-24 10:46:18,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=484012.6666666667, ans=0.0
2024-09-24 10:46:30,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=484059.3333333333, ans=0.125
2024-09-24 10:46:45,619 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.266e+02 1.332e+02 1.424e+02 3.115e+02, threshold=2.664e+02, percent-clipped=1.0
2024-09-24 10:46:54,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=484106.0, ans=0.025
2024-09-24 10:47:03,206 INFO [train.py:1198] (1/4) Epoch 27, batch 2450, loss[loss=0.2322, ctc_loss=0.1516, cr_loss=0.4029, over 17033.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1328, cr_loss=0.3491, over 3373717.43 frames. ], batch size: 52, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:47:08,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=484152.6666666667, ans=0.2
2024-09-24 10:47:13,412 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.58 vs. limit=15.0
2024-09-24 10:47:38,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=484246.0, ans=0.0
2024-09-24 10:47:42,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=484246.0, ans=0.125
2024-09-24 10:47:44,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=484246.0, ans=0.0
2024-09-24 10:48:25,694 INFO [train.py:1198] (1/4) Epoch 27, batch 2500, loss[loss=0.2322, ctc_loss=0.1515, cr_loss=0.4036, over 17338.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.1334, cr_loss=0.3499, over 3376030.27 frames. ], batch size: 52, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:48:53,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=484432.6666666667, ans=0.125
2024-09-24 10:48:53,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=484432.6666666667, ans=0.025
2024-09-24 10:49:30,485 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.304e+02 1.383e+02 1.470e+02 1.966e+02, threshold=2.767e+02, percent-clipped=0.0
2024-09-24 10:49:41,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.93 vs. limit=22.5
2024-09-24 10:49:46,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=484572.6666666667, ans=0.0
2024-09-24 10:49:50,882 INFO [train.py:1198] (1/4) Epoch 27, batch 2550, loss[loss=0.221, ctc_loss=0.1471, cr_loss=0.3695, over 16022.00 frames. ], tot_loss[loss=0.2034, ctc_loss=0.1334, cr_loss=0.3496, over 3374460.75 frames. ], batch size: 74, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:50:13,171 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.78 vs. limit=15.0
2024-09-24 10:50:19,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=484666.0, ans=0.1
2024-09-24 10:50:32,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=484712.6666666667, ans=0.1
2024-09-24 10:50:41,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=484759.3333333333, ans=0.125
2024-09-24 10:51:13,299 INFO [train.py:1198] (1/4) Epoch 27, batch 2600, loss[loss=0.2326, ctc_loss=0.1534, cr_loss=0.396, over 17001.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.1333, cr_loss=0.3501, over 3378269.71 frames. ], batch size: 53, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:51:17,059 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.92 vs. limit=10.0
2024-09-24 10:51:19,022 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.75 vs. limit=15.0
2024-09-24 10:51:23,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=484852.6666666667, ans=0.125
2024-09-24 10:51:36,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=484899.3333333333, ans=0.125
2024-09-24 10:51:39,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=484899.3333333333, ans=0.125
2024-09-24 10:51:39,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=484899.3333333333, ans=10.0
2024-09-24 10:51:50,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=484946.0, ans=0.125
2024-09-24 10:51:56,204 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.76 vs. limit=10.0
2024-09-24 10:52:15,856 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.258e+02 1.335e+02 1.464e+02 4.634e+02, threshold=2.669e+02, percent-clipped=1.0
2024-09-24 10:52:16,875 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=11.35 vs. limit=15.0
2024-09-24 10:52:36,180 INFO [train.py:1198] (1/4) Epoch 27, batch 2650, loss[loss=0.204, ctc_loss=0.1335, cr_loss=0.3527, over 17106.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1332, cr_loss=0.3501, over 3372383.45 frames. ], batch size: 49, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:52:47,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=485086.0, ans=0.125
2024-09-24 10:52:51,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=485132.6666666667, ans=0.125
2024-09-24 10:52:57,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=485132.6666666667, ans=0.125
2024-09-24 10:53:08,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=485179.3333333333, ans=0.125
2024-09-24 10:53:25,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=485226.0, ans=0.125
2024-09-24 10:53:29,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=485226.0, ans=0.125
2024-09-24 10:53:37,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=485226.0, ans=0.2
2024-09-24 10:53:43,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=485272.6666666667, ans=0.125
2024-09-24 10:53:45,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=485272.6666666667, ans=0.1
2024-09-24 10:53:45,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.17 vs. limit=6.0
2024-09-24 10:53:55,010 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.94 vs. limit=6.0
2024-09-24 10:53:55,756 INFO [train.py:1198] (1/4) Epoch 27, batch 2700, loss[loss=0.1848, ctc_loss=0.1189, cr_loss=0.3296, over 17092.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1332, cr_loss=0.3502, over 3366221.48 frames. ], batch size: 40, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:54:28,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=485366.0, ans=0.0
2024-09-24 10:54:39,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=485412.6666666667, ans=0.0
2024-09-24 10:55:02,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=485459.3333333333, ans=0.2
2024-09-24 10:55:08,209 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.258e+02 1.322e+02 1.410e+02 2.487e+02, threshold=2.644e+02, percent-clipped=0.0
2024-09-24 10:55:08,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=485506.0, ans=0.125
2024-09-24 10:55:25,585 INFO [train.py:1198] (1/4) Epoch 27, batch 2750, loss[loss=0.1661, ctc_loss=0.1092, cr_loss=0.2846, over 16970.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1335, cr_loss=0.3511, over 3367698.20 frames. ], batch size: 42, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:55:37,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=485552.6666666667, ans=0.125
2024-09-24 10:55:40,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=485599.3333333333, ans=0.1
2024-09-24 10:55:56,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=485646.0, ans=0.2
2024-09-24 10:55:57,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=485646.0, ans=0.025
2024-09-24 10:56:13,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=485692.6666666667, ans=0.05
2024-09-24 10:56:16,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=485692.6666666667, ans=0.1
2024-09-24 10:56:29,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=485739.3333333333, ans=0.1
2024-09-24 10:56:45,243 INFO [train.py:1198] (1/4) Epoch 27, batch 2800, loss[loss=0.2041, ctc_loss=0.1323, cr_loss=0.3591, over 17299.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1336, cr_loss=0.3513, over 3373035.79 frames. ], batch size: 46, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:57:03,626 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.47 vs. limit=15.0
2024-09-24 10:57:20,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=485879.3333333333, ans=0.0
2024-09-24 10:57:27,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=485879.3333333333, ans=0.125
2024-09-24 10:57:28,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=485879.3333333333, ans=0.1
2024-09-24 10:57:41,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=485926.0, ans=0.5
2024-09-24 10:57:50,313 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.262e+02 1.376e+02 1.500e+02 2.364e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-24 10:58:07,804 INFO [train.py:1198] (1/4) Epoch 27, batch 2850, loss[loss=0.2381, ctc_loss=0.1624, cr_loss=0.3785, over 15166.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.1339, cr_loss=0.3518, over 3358547.84 frames. ], batch size: 90, lr: 4.41e-03, grad_scale: 32.0
2024-09-24 10:58:08,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=486019.3333333333, ans=0.05
2024-09-24 10:58:17,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=486019.3333333333, ans=0.025
2024-09-24 10:58:28,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=486066.0, ans=0.125
2024-09-24 10:58:41,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=486112.6666666667, ans=0.125
2024-09-24 10:58:51,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=486112.6666666667, ans=0.2
2024-09-24 10:58:54,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=486159.3333333333, ans=0.1
2024-09-24 10:59:06,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=486159.3333333333, ans=0.125
2024-09-24 10:59:30,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=486206.0, ans=0.125
2024-09-24 10:59:33,000 INFO [train.py:1198] (1/4) Epoch 27, batch 2900, loss[loss=0.2178, ctc_loss=0.1448, cr_loss=0.3651, over 17197.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1347, cr_loss=0.3528, over 3356607.46 frames. ], batch size: 47, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 10:59:36,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=486252.6666666667, ans=0.125
2024-09-24 11:00:04,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=486299.3333333333, ans=0.125
2024-09-24 11:00:11,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.32 vs. limit=6.0
2024-09-24 11:00:20,810 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=15.58 vs. limit=15.0
2024-09-24 11:00:23,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=486392.6666666667, ans=0.0
2024-09-24 11:00:37,810 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.250e+02 1.339e+02 1.437e+02 2.331e+02, threshold=2.678e+02, percent-clipped=0.0
2024-09-24 11:00:55,786 INFO [train.py:1198] (1/4) Epoch 27, batch 2950, loss[loss=0.1959, ctc_loss=0.1293, cr_loss=0.3327, over 17226.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.1342, cr_loss=0.3522, over 3352457.21 frames. ], batch size: 47, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:00:57,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=486486.0, ans=0.1
2024-09-24 11:00:57,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=486486.0, ans=0.0
2024-09-24 11:00:59,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=486486.0, ans=0.125
2024-09-24 11:01:50,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=486626.0, ans=0.1
2024-09-24 11:02:04,886 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.74 vs. limit=22.5
2024-09-24 11:02:13,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=486719.3333333333, ans=0.05
2024-09-24 11:02:14,940 INFO [train.py:1198] (1/4) Epoch 27, batch 3000, loss[loss=0.1758, ctc_loss=0.1126, cr_loss=0.3162, over 17161.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1336, cr_loss=0.3511, over 3355770.80 frames. ], batch size: 41, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:02:14,941 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 11:02:30,463 INFO [train.py:1230] (1/4) Epoch 27, validation: loss=0.03681, ctc_loss=0.03681, cr_loss=8.353e-15, over 944034.00 frames. 
2024-09-24 11:02:30,464 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 11:02:39,055 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.10 vs. limit=15.0
2024-09-24 11:02:54,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=486766.0, ans=0.0
2024-09-24 11:03:01,341 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.12 vs. limit=10.0
2024-09-24 11:03:02,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=486812.6666666667, ans=0.0
2024-09-24 11:03:05,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=486812.6666666667, ans=0.125
2024-09-24 11:03:14,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=486812.6666666667, ans=0.0
2024-09-24 11:03:25,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=486859.3333333333, ans=0.025
2024-09-24 11:03:30,149 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.58 vs. limit=5.0
2024-09-24 11:03:32,006 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.263e+02 1.339e+02 1.435e+02 2.051e+02, threshold=2.678e+02, percent-clipped=0.0
2024-09-24 11:03:41,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=486906.0, ans=0.1
2024-09-24 11:03:46,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=486906.0, ans=0.125
2024-09-24 11:03:49,183 INFO [train.py:1198] (1/4) Epoch 27, batch 3050, loss[loss=0.1729, ctc_loss=0.1098, cr_loss=0.3154, over 17275.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1336, cr_loss=0.3513, over 3356661.42 frames. ], batch size: 42, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:03:56,157 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.52 vs. limit=15.0
2024-09-24 11:04:03,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=486999.3333333333, ans=0.125
2024-09-24 11:04:04,218 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.25 vs. limit=15.0
2024-09-24 11:04:32,480 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.51 vs. limit=15.0
2024-09-24 11:04:41,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=487092.6666666667, ans=0.0
2024-09-24 11:04:44,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=487092.6666666667, ans=0.125
2024-09-24 11:05:07,555 INFO [train.py:1198] (1/4) Epoch 27, batch 3100, loss[loss=0.2501, ctc_loss=0.1675, cr_loss=0.4128, over 17039.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1335, cr_loss=0.3513, over 3366115.12 frames. ], batch size: 52, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:05:07,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=487186.0, ans=0.125
2024-09-24 11:05:10,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=487186.0, ans=0.125
2024-09-24 11:05:42,256 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 11:05:47,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=487279.3333333333, ans=0.125
2024-09-24 11:05:56,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=487326.0, ans=0.05
2024-09-24 11:06:07,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=487326.0, ans=0.025
2024-09-24 11:06:11,764 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.251e+02 1.335e+02 1.455e+02 2.261e+02, threshold=2.670e+02, percent-clipped=0.0
2024-09-24 11:06:22,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=487372.6666666667, ans=0.125
2024-09-24 11:06:28,978 INFO [train.py:1198] (1/4) Epoch 27, batch 3150, loss[loss=0.2006, ctc_loss=0.1336, cr_loss=0.335, over 17215.00 frames. ], tot_loss[loss=0.2035, ctc_loss=0.1333, cr_loss=0.3508, over 3370907.48 frames. ], batch size: 55, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:07:04,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=487512.6666666667, ans=0.125
2024-09-24 11:07:12,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=487512.6666666667, ans=0.0
2024-09-24 11:07:15,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=487512.6666666667, ans=0.125
2024-09-24 11:07:19,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=487559.3333333333, ans=0.0
2024-09-24 11:07:25,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=487559.3333333333, ans=0.2
2024-09-24 11:07:46,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=487606.0, ans=0.1
2024-09-24 11:07:50,459 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.76 vs. limit=15.0
2024-09-24 11:07:51,271 INFO [train.py:1198] (1/4) Epoch 27, batch 3200, loss[loss=0.1956, ctc_loss=0.1266, cr_loss=0.3452, over 17023.00 frames. ], tot_loss[loss=0.2041, ctc_loss=0.1338, cr_loss=0.3515, over 3358298.01 frames. ], batch size: 44, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:07:56,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=487652.6666666667, ans=0.125
2024-09-24 11:08:09,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=487699.3333333333, ans=0.2
2024-09-24 11:08:21,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=487746.0, ans=0.125
2024-09-24 11:08:32,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=487746.0, ans=0.0
2024-09-24 11:08:33,004 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.65 vs. limit=15.0
2024-09-24 11:08:36,511 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=6.71 vs. limit=10.0
2024-09-24 11:08:52,485 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.262e+02 1.380e+02 1.502e+02 1.892e+02, threshold=2.760e+02, percent-clipped=0.0
2024-09-24 11:09:02,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=487839.3333333333, ans=0.125
2024-09-24 11:09:09,782 INFO [train.py:1198] (1/4) Epoch 27, batch 3250, loss[loss=0.2198, ctc_loss=0.146, cr_loss=0.3689, over 15032.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1335, cr_loss=0.3505, over 3354086.26 frames. ], batch size: 89, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:09:14,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=487886.0, ans=0.2
2024-09-24 11:09:19,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=487886.0, ans=0.07
2024-09-24 11:09:28,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=487932.6666666667, ans=0.2
2024-09-24 11:09:35,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=487932.6666666667, ans=0.1
2024-09-24 11:09:43,505 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.66 vs. limit=22.5
2024-09-24 11:09:52,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=487979.3333333333, ans=0.125
2024-09-24 11:09:52,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=487979.3333333333, ans=0.125
2024-09-24 11:09:52,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=487979.3333333333, ans=0.0
2024-09-24 11:10:13,402 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.76 vs. limit=6.0
2024-09-24 11:10:23,852 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.29 vs. limit=15.0
2024-09-24 11:10:27,842 INFO [train.py:1198] (1/4) Epoch 27, batch 3300, loss[loss=0.1688, ctc_loss=0.1109, cr_loss=0.2895, over 17287.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1335, cr_loss=0.3506, over 3355869.60 frames. ], batch size: 46, lr: 4.40e-03, grad_scale: 32.0
2024-09-24 11:10:51,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=488166.0, ans=0.07
2024-09-24 11:10:51,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten.whitening_limit, batch_count=488166.0, ans=15.0
2024-09-24 11:11:05,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=488212.6666666667, ans=0.125
2024-09-24 11:11:05,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=488212.6666666667, ans=0.125
2024-09-24 11:11:07,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=488212.6666666667, ans=0.125
2024-09-24 11:11:10,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=488212.6666666667, ans=0.1
2024-09-24 11:11:28,980 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.251e+02 1.338e+02 1.445e+02 2.480e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-24 11:11:46,441 INFO [train.py:1198] (1/4) Epoch 27, batch 3350, loss[loss=0.1944, ctc_loss=0.1253, cr_loss=0.3457, over 17105.00 frames. ], tot_loss[loss=0.2031, ctc_loss=0.1332, cr_loss=0.3499, over 3350322.50 frames. ], batch size: 43, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:12:07,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=488399.3333333333, ans=0.2
2024-09-24 11:12:13,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=488399.3333333333, ans=0.1
2024-09-24 11:12:14,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=488399.3333333333, ans=0.0
2024-09-24 11:12:25,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=488446.0, ans=0.0
2024-09-24 11:12:31,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=488446.0, ans=0.2
2024-09-24 11:12:32,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=488446.0, ans=0.125
2024-09-24 11:12:33,084 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.99 vs. limit=15.0
2024-09-24 11:12:51,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=488539.3333333333, ans=0.1
2024-09-24 11:13:02,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=488539.3333333333, ans=0.2
2024-09-24 11:13:06,075 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.18 vs. limit=12.0
2024-09-24 11:13:07,096 INFO [train.py:1198] (1/4) Epoch 27, batch 3400, loss[loss=0.2193, ctc_loss=0.148, cr_loss=0.3568, over 16936.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1337, cr_loss=0.3512, over 3354446.45 frames. ], batch size: 58, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:13:07,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=488586.0, ans=0.025
2024-09-24 11:13:23,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=488632.6666666667, ans=0.07
2024-09-24 11:13:24,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=488632.6666666667, ans=0.0
2024-09-24 11:13:32,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=488632.6666666667, ans=0.0
2024-09-24 11:13:51,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer_ff2.min_abs, batch_count=488679.3333333333, ans=0.1
2024-09-24 11:13:54,846 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.61 vs. limit=15.0
2024-09-24 11:13:57,731 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 11:14:08,004 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.290e+02 1.371e+02 1.502e+02 1.850e+02, threshold=2.743e+02, percent-clipped=0.0
2024-09-24 11:14:11,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=488772.6666666667, ans=0.09899494936611666
2024-09-24 11:14:12,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=488772.6666666667, ans=0.0
2024-09-24 11:14:25,084 INFO [train.py:1198] (1/4) Epoch 27, batch 3450, loss[loss=0.1757, ctc_loss=0.1136, cr_loss=0.3105, over 17085.00 frames. ], tot_loss[loss=0.2052, ctc_loss=0.1346, cr_loss=0.3532, over 3355604.93 frames. ], batch size: 43, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:14:39,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=488866.0, ans=0.125
2024-09-24 11:15:13,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=488959.3333333333, ans=0.125
2024-09-24 11:15:27,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=489006.0, ans=0.1
2024-09-24 11:15:44,983 INFO [train.py:1198] (1/4) Epoch 27, batch 3500, loss[loss=0.2248, ctc_loss=0.147, cr_loss=0.3894, over 17018.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1341, cr_loss=0.3519, over 3353646.38 frames. ], batch size: 56, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:16:01,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=489099.3333333333, ans=0.1
2024-09-24 11:16:01,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=489099.3333333333, ans=0.125
2024-09-24 11:16:04,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=489099.3333333333, ans=0.125
2024-09-24 11:16:04,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=489099.3333333333, ans=15.0
2024-09-24 11:16:08,249 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=17.20 vs. limit=22.5
2024-09-24 11:16:22,038 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 11:16:23,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=489146.0, ans=0.125
2024-09-24 11:16:44,358 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=11.25 vs. limit=15.0
2024-09-24 11:16:46,472 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.271e+02 1.363e+02 1.500e+02 3.531e+02, threshold=2.727e+02, percent-clipped=1.0
2024-09-24 11:16:58,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=489239.3333333333, ans=0.025
2024-09-24 11:17:02,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=489239.3333333333, ans=0.125
2024-09-24 11:17:04,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=489286.0, ans=0.125
2024-09-24 11:17:05,735 INFO [train.py:1198] (1/4) Epoch 27, batch 3550, loss[loss=0.2036, ctc_loss=0.1347, cr_loss=0.3447, over 17297.00 frames. ], tot_loss[loss=0.2044, ctc_loss=0.134, cr_loss=0.3517, over 3354243.15 frames. ], batch size: 46, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:17:40,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=489379.3333333333, ans=0.1
2024-09-24 11:17:46,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=489379.3333333333, ans=0.0
2024-09-24 11:17:48,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=489379.3333333333, ans=0.125
2024-09-24 11:17:57,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=489426.0, ans=0.05
2024-09-24 11:18:25,765 INFO [train.py:1198] (1/4) Epoch 27, batch 3600, loss[loss=0.1979, ctc_loss=0.1313, cr_loss=0.3331, over 17214.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.1337, cr_loss=0.3513, over 3353955.04 frames. ], batch size: 50, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:18:44,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=489566.0, ans=0.125
2024-09-24 11:19:11,969 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.88 vs. limit=10.0
2024-09-24 11:19:26,811 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.270e+02 1.354e+02 1.435e+02 1.974e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-24 11:19:31,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=489706.0, ans=0.1
2024-09-24 11:19:44,335 INFO [train.py:1198] (1/4) Epoch 27, batch 3650, loss[loss=0.181, ctc_loss=0.1177, cr_loss=0.3164, over 17090.00 frames. ], tot_loss[loss=0.2047, ctc_loss=0.1343, cr_loss=0.3519, over 3352253.46 frames. ], batch size: 43, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:19:52,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=489752.6666666667, ans=0.0
2024-09-24 11:19:52,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=489752.6666666667, ans=0.04949747468305833
2024-09-24 11:20:00,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=489799.3333333333, ans=0.2
2024-09-24 11:20:03,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=489799.3333333333, ans=0.05
2024-09-24 11:20:06,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=489799.3333333333, ans=0.2
2024-09-24 11:20:16,968 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.53 vs. limit=15.0
2024-09-24 11:20:30,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=489892.6666666667, ans=0.0
2024-09-24 11:20:37,070 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.05 vs. limit=10.0
2024-09-24 11:20:56,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=489939.3333333333, ans=0.125
2024-09-24 11:21:03,534 INFO [train.py:1198] (1/4) Epoch 27, batch 3700, loss[loss=0.2151, ctc_loss=0.1394, cr_loss=0.3784, over 17153.00 frames. ], tot_loss[loss=0.2048, ctc_loss=0.1343, cr_loss=0.3523, over 3355094.88 frames. ], batch size: 48, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:21:30,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.28 vs. limit=22.5
2024-09-24 11:21:47,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=490079.3333333333, ans=0.125
2024-09-24 11:22:03,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=490126.0, ans=0.125
2024-09-24 11:22:05,030 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.302e+02 1.450e+02 1.561e+02 3.629e+02, threshold=2.900e+02, percent-clipped=2.0
2024-09-24 11:22:05,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=490172.6666666667, ans=0.0
2024-09-24 11:22:17,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=490172.6666666667, ans=0.125
2024-09-24 11:22:21,896 INFO [train.py:1198] (1/4) Epoch 27, batch 3750, loss[loss=0.1823, ctc_loss=0.1168, cr_loss=0.3276, over 16196.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1342, cr_loss=0.3513, over 3333106.77 frames. ], batch size: 36, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:22:51,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=490312.6666666667, ans=0.2
2024-09-24 11:23:18,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=490359.3333333333, ans=0.125
2024-09-24 11:23:39,567 INFO [train.py:1198] (1/4) Epoch 27, batch 3800, loss[loss=0.249, ctc_loss=0.1747, cr_loss=0.3717, over 11531.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.1338, cr_loss=0.3508, over 3322750.68 frames. ], batch size: 123, lr: 4.39e-03, grad_scale: 32.0
2024-09-24 11:24:16,799 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.83 vs. limit=15.0
2024-09-24 11:24:41,923 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.297e+02 1.423e+02 1.568e+02 2.554e+02, threshold=2.846e+02, percent-clipped=0.0
2024-09-24 11:24:59,148 INFO [train.py:1198] (1/4) Epoch 27, batch 3850, loss[loss=0.2193, ctc_loss=0.1435, cr_loss=0.379, over 16847.00 frames. ], tot_loss[loss=0.2057, ctc_loss=0.1353, cr_loss=0.3524, over 3293592.13 frames. ], batch size: 58, lr: 4.38e-03, grad_scale: 32.0
2024-09-24 11:24:59,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer_ff2.min_abs, batch_count=490686.0, ans=0.1
2024-09-24 11:25:24,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.49 vs. limit=15.0
2024-09-24 11:25:36,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=490779.3333333333, ans=0.125
2024-09-24 11:25:38,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=490779.3333333333, ans=0.125
2024-09-24 11:27:00,314 INFO [train.py:1198] (1/4) Epoch 28, batch 0, loss[loss=0.219, ctc_loss=0.145, cr_loss=0.37, over 17135.00 frames. ], tot_loss[loss=0.219, ctc_loss=0.145, cr_loss=0.37, over 17135.00 frames. ], batch size: 48, lr: 4.30e-03, grad_scale: 32.0
2024-09-24 11:27:00,314 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 11:27:15,946 INFO [train.py:1230] (1/4) Epoch 28, validation: loss=0.03666, ctc_loss=0.03666, cr_loss=9.126e-15, over 944034.00 frames. 
2024-09-24 11:27:15,947 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 11:27:51,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.whiten.whitening_limit, batch_count=490994.0, ans=12.0
2024-09-24 11:28:14,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.05 vs. limit=15.0
2024-09-24 11:28:27,521 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.316e+02 1.494e+02 1.657e+02 3.455e+02, threshold=2.987e+02, percent-clipped=1.0
2024-09-24 11:28:32,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=491087.3333333333, ans=0.125
2024-09-24 11:28:38,881 INFO [train.py:1198] (1/4) Epoch 28, batch 50, loss[loss=0.1771, ctc_loss=0.1157, cr_loss=0.3072, over 17301.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.1303, cr_loss=0.346, over 764901.46 frames. ], batch size: 46, lr: 4.30e-03, grad_scale: 32.0
2024-09-24 11:29:05,180 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.94 vs. limit=10.0
2024-09-24 11:29:36,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=491274.0, ans=0.125
2024-09-24 11:29:58,609 INFO [train.py:1198] (1/4) Epoch 28, batch 100, loss[loss=0.2099, ctc_loss=0.1395, cr_loss=0.3517, over 17122.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1316, cr_loss=0.3473, over 1339312.68 frames. ], batch size: 49, lr: 4.30e-03, grad_scale: 32.0
2024-09-24 11:30:23,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten.whitening_limit, batch_count=491414.0, ans=15.0
2024-09-24 11:30:30,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=491414.0, ans=0.04949747468305833
2024-09-24 11:30:30,868 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=8.15 vs. limit=15.0
2024-09-24 11:30:32,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=491460.6666666667, ans=0.125
2024-09-24 11:30:52,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=491507.3333333333, ans=0.1
2024-09-24 11:31:05,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=491554.0, ans=0.125
2024-09-24 11:31:11,239 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.221e+02 1.303e+02 1.406e+02 2.036e+02, threshold=2.607e+02, percent-clipped=0.0
2024-09-24 11:31:20,724 INFO [train.py:1198] (1/4) Epoch 28, batch 150, loss[loss=0.2384, ctc_loss=0.1626, cr_loss=0.3793, over 14594.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1322, cr_loss=0.3482, over 1779073.39 frames. ], batch size: 88, lr: 4.30e-03, grad_scale: 32.0
2024-09-24 11:31:32,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=491600.6666666667, ans=0.1
2024-09-24 11:31:50,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=491647.3333333333, ans=15.0
2024-09-24 11:31:52,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=491694.0, ans=0.125
2024-09-24 11:31:53,040 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.11 vs. limit=22.5
2024-09-24 11:32:41,057 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=8.08 vs. limit=12.0
2024-09-24 11:32:44,506 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.74 vs. limit=22.5
2024-09-24 11:32:48,350 INFO [train.py:1198] (1/4) Epoch 28, batch 200, loss[loss=0.2102, ctc_loss=0.1368, cr_loss=0.3671, over 17213.00 frames. ], tot_loss[loss=0.2016, ctc_loss=0.1319, cr_loss=0.3482, over 2129257.78 frames. ], batch size: 55, lr: 4.30e-03, grad_scale: 16.0
2024-09-24 11:32:50,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=491834.0, ans=0.04949747468305833
2024-09-24 11:32:57,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=491834.0, ans=0.125
2024-09-24 11:33:17,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=491880.6666666667, ans=0.125
2024-09-24 11:33:18,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=491927.3333333333, ans=0.0
2024-09-24 11:33:25,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=491927.3333333333, ans=0.1
2024-09-24 11:33:34,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=491974.0, ans=0.125
2024-09-24 11:33:38,267 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=11.96 vs. limit=22.5
2024-09-24 11:33:58,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=492020.6666666667, ans=0.0
2024-09-24 11:34:00,088 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.241e+02 1.331e+02 1.443e+02 1.741e+02, threshold=2.662e+02, percent-clipped=0.0
2024-09-24 11:34:08,075 INFO [train.py:1198] (1/4) Epoch 28, batch 250, loss[loss=0.2306, ctc_loss=0.1544, cr_loss=0.3813, over 16635.00 frames. ], tot_loss[loss=0.2012, ctc_loss=0.1316, cr_loss=0.3482, over 2406711.35 frames. ], batch size: 66, lr: 4.30e-03, grad_scale: 16.0
2024-09-24 11:34:16,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=492067.3333333333, ans=0.125
2024-09-24 11:34:37,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=492114.0, ans=0.125
2024-09-24 11:34:45,404 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.38 vs. limit=15.0
2024-09-24 11:34:51,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=492160.6666666667, ans=0.1
2024-09-24 11:35:26,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=492254.0, ans=0.125
2024-09-24 11:35:30,770 INFO [train.py:1198] (1/4) Epoch 28, batch 300, loss[loss=0.1841, ctc_loss=0.1179, cr_loss=0.3312, over 16953.00 frames. ], tot_loss[loss=0.2015, ctc_loss=0.1319, cr_loss=0.3479, over 2619468.93 frames. ], batch size: 42, lr: 4.30e-03, grad_scale: 16.0
2024-09-24 11:35:31,428 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.46 vs. limit=22.5
2024-09-24 11:35:51,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=492347.3333333333, ans=0.0
2024-09-24 11:36:04,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=492394.0, ans=0.0
2024-09-24 11:36:22,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=492440.6666666667, ans=0.0
2024-09-24 11:36:26,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.94 vs. limit=10.0
2024-09-24 11:36:33,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=492487.3333333333, ans=0.1
2024-09-24 11:36:36,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=492487.3333333333, ans=0.125
2024-09-24 11:36:42,613 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.243e+02 1.313e+02 1.459e+02 2.703e+02, threshold=2.626e+02, percent-clipped=1.0
2024-09-24 11:36:50,619 INFO [train.py:1198] (1/4) Epoch 28, batch 350, loss[loss=0.1935, ctc_loss=0.1253, cr_loss=0.3409, over 17240.00 frames. ], tot_loss[loss=0.2003, ctc_loss=0.131, cr_loss=0.3462, over 2784804.90 frames. ], batch size: 42, lr: 4.30e-03, grad_scale: 16.0
2024-09-24 11:36:54,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=492534.0, ans=0.0
2024-09-24 11:37:57,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=492674.0, ans=0.025
2024-09-24 11:38:01,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=492720.6666666667, ans=0.2
2024-09-24 11:38:18,683 INFO [train.py:1198] (1/4) Epoch 28, batch 400, loss[loss=0.1983, ctc_loss=0.1272, cr_loss=0.3553, over 17282.00 frames. ], tot_loss[loss=0.1999, ctc_loss=0.1307, cr_loss=0.3462, over 2917165.98 frames. ], batch size: 42, lr: 4.29e-03, grad_scale: 32.0
2024-09-24 11:38:32,536 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.88 vs. limit=22.5
2024-09-24 11:38:36,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=492814.0, ans=0.125
2024-09-24 11:38:49,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=492860.6666666667, ans=0.0
2024-09-24 11:39:05,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=492907.3333333333, ans=0.125
2024-09-24 11:39:19,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=492907.3333333333, ans=0.025
2024-09-24 11:39:30,616 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.218e+02 1.284e+02 1.430e+02 3.216e+02, threshold=2.568e+02, percent-clipped=1.0
2024-09-24 11:39:38,502 INFO [train.py:1198] (1/4) Epoch 28, batch 450, loss[loss=0.1806, ctc_loss=0.1165, cr_loss=0.3203, over 17190.00 frames. ], tot_loss[loss=0.1997, ctc_loss=0.1305, cr_loss=0.346, over 3022681.44 frames. ], batch size: 41, lr: 4.29e-03, grad_scale: 32.0
2024-09-24 11:40:16,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=493094.0, ans=0.1
2024-09-24 11:40:33,386 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.03 vs. limit=6.0
2024-09-24 11:41:00,830 INFO [train.py:1198] (1/4) Epoch 28, batch 500, loss[loss=0.2103, ctc_loss=0.1362, cr_loss=0.3709, over 17365.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1312, cr_loss=0.3476, over 3090423.19 frames. ], batch size: 48, lr: 4.29e-03, grad_scale: 16.0
2024-09-24 11:41:09,442 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=7.33 vs. limit=12.0
2024-09-24 11:41:29,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=493280.6666666667, ans=0.0
2024-09-24 11:41:29,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=493280.6666666667, ans=0.125
2024-09-24 11:41:50,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=493374.0, ans=0.025
2024-09-24 11:42:12,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=493420.6666666667, ans=0.2
2024-09-24 11:42:21,660 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.063e+02 1.292e+02 1.373e+02 1.533e+02 1.928e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-24 11:42:27,936 INFO [train.py:1198] (1/4) Epoch 28, batch 550, loss[loss=0.2076, ctc_loss=0.1332, cr_loss=0.3722, over 17207.00 frames. ], tot_loss[loss=0.2012, ctc_loss=0.1315, cr_loss=0.3485, over 3151030.48 frames. ], batch size: 47, lr: 4.29e-03, grad_scale: 16.0
2024-09-24 11:42:29,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=493467.3333333333, ans=0.0
2024-09-24 11:42:38,715 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=6.25 vs. limit=15.0
2024-09-24 11:42:56,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=493514.0, ans=0.0
2024-09-24 11:43:17,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=493607.3333333333, ans=0.05
2024-09-24 11:43:32,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=493654.0, ans=0.0
2024-09-24 11:43:33,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=493654.0, ans=0.2
2024-09-24 11:43:47,658 INFO [train.py:1198] (1/4) Epoch 28, batch 600, loss[loss=0.2368, ctc_loss=0.1561, cr_loss=0.4036, over 17005.00 frames. ], tot_loss[loss=0.2016, ctc_loss=0.1317, cr_loss=0.3493, over 3200029.90 frames. ], batch size: 53, lr: 4.29e-03, grad_scale: 16.0
2024-09-24 11:43:56,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=493700.6666666667, ans=0.0
2024-09-24 11:44:02,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=493747.3333333333, ans=0.0
2024-09-24 11:44:17,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=493747.3333333333, ans=0.2
2024-09-24 11:44:28,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=493794.0, ans=0.125
2024-09-24 11:44:54,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=493887.3333333333, ans=0.0
2024-09-24 11:44:57,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=493887.3333333333, ans=0.0
2024-09-24 11:45:03,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=493887.3333333333, ans=0.125
2024-09-24 11:45:04,823 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.264e+02 1.373e+02 1.490e+02 2.458e+02, threshold=2.746e+02, percent-clipped=0.0
2024-09-24 11:45:11,168 INFO [train.py:1198] (1/4) Epoch 28, batch 650, loss[loss=0.2411, ctc_loss=0.1708, cr_loss=0.3514, over 11897.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1325, cr_loss=0.3501, over 3233965.63 frames. ], batch size: 124, lr: 4.29e-03, grad_scale: 16.0
2024-09-24 11:45:14,969 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=8.68 vs. limit=15.0
2024-09-24 11:45:29,635 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.25 vs. limit=22.5
2024-09-24 11:45:32,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=493980.6666666667, ans=0.1
2024-09-24 11:45:46,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=494027.3333333333, ans=0.0
2024-09-24 11:46:31,266 INFO [train.py:1198] (1/4) Epoch 28, batch 700, loss[loss=0.172, ctc_loss=0.1076, cr_loss=0.3222, over 17009.00 frames. ], tot_loss[loss=0.2035, ctc_loss=0.1332, cr_loss=0.3513, over 3255571.37 frames. ], batch size: 44, lr: 4.29e-03, grad_scale: 16.0
2024-09-24 11:46:37,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=494167.3333333333, ans=0.0
2024-09-24 11:46:39,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=494167.3333333333, ans=0.125
2024-09-24 11:47:48,447 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.03 vs. limit=22.5
2024-09-24 11:47:51,748 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.40 vs. limit=22.5
2024-09-24 11:47:52,209 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.257e+02 1.354e+02 1.480e+02 2.179e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-24 11:47:55,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=494354.0, ans=0.125
2024-09-24 11:47:58,822 INFO [train.py:1198] (1/4) Epoch 28, batch 750, loss[loss=0.2058, ctc_loss=0.1303, cr_loss=0.3776, over 17108.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1321, cr_loss=0.349, over 3283273.68 frames. ], batch size: 43, lr: 4.29e-03, grad_scale: 16.0
2024-09-24 11:48:24,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=494447.3333333333, ans=0.1
2024-09-24 11:49:18,698 INFO [train.py:1198] (1/4) Epoch 28, batch 800, loss[loss=0.222, ctc_loss=0.1462, cr_loss=0.3792, over 15941.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1327, cr_loss=0.3496, over 3294686.64 frames. ], batch size: 74, lr: 4.29e-03, grad_scale: 32.0
2024-09-24 11:49:36,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=494680.6666666667, ans=0.025
2024-09-24 11:49:54,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=494727.3333333333, ans=0.0
2024-09-24 11:49:57,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=494727.3333333333, ans=0.1
2024-09-24 11:50:20,369 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.13 vs. limit=6.0
2024-09-24 11:50:30,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=494820.6666666667, ans=0.05
2024-09-24 11:50:33,711 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.239e+02 1.301e+02 1.424e+02 2.595e+02, threshold=2.602e+02, percent-clipped=0.0
2024-09-24 11:50:36,624 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.99 vs. limit=10.0
2024-09-24 11:50:40,254 INFO [train.py:1198] (1/4) Epoch 28, batch 850, loss[loss=0.1959, ctc_loss=0.1263, cr_loss=0.3481, over 17188.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1325, cr_loss=0.3484, over 3301253.35 frames. ], batch size: 41, lr: 4.29e-03, grad_scale: 32.0
2024-09-24 11:50:43,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=494867.3333333333, ans=0.0
2024-09-24 11:50:43,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=494867.3333333333, ans=0.0
2024-09-24 11:51:45,490 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 11:51:49,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=495054.0, ans=0.0
2024-09-24 11:52:04,921 INFO [train.py:1198] (1/4) Epoch 28, batch 900, loss[loss=0.2129, ctc_loss=0.1398, cr_loss=0.3655, over 17257.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1327, cr_loss=0.3489, over 3310654.45 frames. ], batch size: 44, lr: 4.28e-03, grad_scale: 16.0
2024-09-24 11:52:11,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=495100.6666666667, ans=0.125
2024-09-24 11:52:16,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=495100.6666666667, ans=0.025
2024-09-24 11:52:16,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=495100.6666666667, ans=0.0
2024-09-24 11:52:27,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=495147.3333333333, ans=0.125
2024-09-24 11:52:29,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=495147.3333333333, ans=0.2
2024-09-24 11:52:51,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=495240.6666666667, ans=0.0
2024-09-24 11:53:06,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=495240.6666666667, ans=0.2
2024-09-24 11:53:06,285 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 11:53:10,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=495287.3333333333, ans=0.025
2024-09-24 11:53:20,248 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.272e+02 1.349e+02 1.497e+02 2.522e+02, threshold=2.699e+02, percent-clipped=0.0
2024-09-24 11:53:25,042 INFO [train.py:1198] (1/4) Epoch 28, batch 950, loss[loss=0.1999, ctc_loss=0.1302, cr_loss=0.3485, over 17284.00 frames. ], tot_loss[loss=0.2023, ctc_loss=0.1325, cr_loss=0.3489, over 3322361.02 frames. ], batch size: 51, lr: 4.28e-03, grad_scale: 16.0
2024-09-24 11:53:44,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=495380.6666666667, ans=0.0
2024-09-24 11:53:52,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=495380.6666666667, ans=0.0
2024-09-24 11:54:08,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=495427.3333333333, ans=0.125
2024-09-24 11:54:10,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=495427.3333333333, ans=0.0
2024-09-24 11:54:23,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=495474.0, ans=0.125
2024-09-24 11:54:47,629 INFO [train.py:1198] (1/4) Epoch 28, batch 1000, loss[loss=0.2415, ctc_loss=0.1606, cr_loss=0.4047, over 16581.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1323, cr_loss=0.3495, over 3335432.82 frames. ], batch size: 66, lr: 4.28e-03, grad_scale: 16.0
2024-09-24 11:55:09,421 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.30 vs. limit=6.0
2024-09-24 11:55:11,015 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.66 vs. limit=22.5
2024-09-24 11:55:23,741 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.62 vs. limit=15.0
2024-09-24 11:55:39,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=495707.3333333333, ans=0.0
2024-09-24 11:55:58,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=495754.0, ans=0.0
2024-09-24 11:56:02,811 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.044e+02 1.229e+02 1.298e+02 1.405e+02 1.761e+02, threshold=2.595e+02, percent-clipped=0.0
2024-09-24 11:56:07,659 INFO [train.py:1198] (1/4) Epoch 28, batch 1050, loss[loss=0.1753, ctc_loss=0.1121, cr_loss=0.316, over 16979.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1314, cr_loss=0.348, over 3348897.68 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 16.0
2024-09-24 11:56:22,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=495847.3333333333, ans=0.5
2024-09-24 11:57:04,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=495940.6666666667, ans=0.2
2024-09-24 11:57:35,041 INFO [train.py:1198] (1/4) Epoch 28, batch 1100, loss[loss=0.196, ctc_loss=0.1284, cr_loss=0.3383, over 17097.00 frames. ], tot_loss[loss=0.2014, ctc_loss=0.1317, cr_loss=0.3485, over 3352906.75 frames. ], batch size: 49, lr: 4.28e-03, grad_scale: 16.0
2024-09-24 11:58:10,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=496127.3333333333, ans=0.125
2024-09-24 11:58:18,930 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.76 vs. limit=15.0
2024-09-24 11:58:20,292 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=20.36 vs. limit=22.5
2024-09-24 11:58:39,761 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.19 vs. limit=15.0
2024-09-24 11:58:43,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=496220.6666666667, ans=0.0
2024-09-24 11:58:48,959 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 11:58:50,153 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.267e+02 1.365e+02 1.472e+02 3.756e+02, threshold=2.729e+02, percent-clipped=1.0
2024-09-24 11:58:54,888 INFO [train.py:1198] (1/4) Epoch 28, batch 1150, loss[loss=0.1886, ctc_loss=0.1235, cr_loss=0.3255, over 16960.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1314, cr_loss=0.3477, over 3354165.84 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 16.0
2024-09-24 11:59:08,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=496267.3333333333, ans=0.125
2024-09-24 11:59:29,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=496360.6666666667, ans=0.2
2024-09-24 11:59:45,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=496407.3333333333, ans=0.125
2024-09-24 11:59:50,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=496407.3333333333, ans=0.0
2024-09-24 11:59:58,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=496407.3333333333, ans=0.1
2024-09-24 12:00:17,103 INFO [train.py:1198] (1/4) Epoch 28, batch 1200, loss[loss=0.1626, ctc_loss=0.1036, cr_loss=0.2947, over 17029.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1327, cr_loss=0.3495, over 3333005.56 frames. ], batch size: 39, lr: 4.28e-03, grad_scale: 32.0
2024-09-24 12:00:23,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=496500.6666666667, ans=0.125
2024-09-24 12:00:30,997 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.56 vs. limit=15.0
2024-09-24 12:00:38,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=496547.3333333333, ans=0.125
2024-09-24 12:01:32,308 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.289e+02 1.355e+02 1.454e+02 3.122e+02, threshold=2.710e+02, percent-clipped=1.0
2024-09-24 12:01:35,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=496734.0, ans=0.125
2024-09-24 12:01:37,159 INFO [train.py:1198] (1/4) Epoch 28, batch 1250, loss[loss=0.1919, ctc_loss=0.1263, cr_loss=0.3281, over 17040.00 frames. ], tot_loss[loss=0.2024, ctc_loss=0.1326, cr_loss=0.3494, over 3340027.81 frames. ], batch size: 39, lr: 4.28e-03, grad_scale: 32.0
2024-09-24 12:01:41,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=496734.0, ans=0.125
2024-09-24 12:01:49,293 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.99 vs. limit=15.0
2024-09-24 12:02:27,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=496827.3333333333, ans=0.125
2024-09-24 12:03:03,730 INFO [train.py:1198] (1/4) Epoch 28, batch 1300, loss[loss=0.2075, ctc_loss=0.1367, cr_loss=0.3542, over 16693.00 frames. ], tot_loss[loss=0.2024, ctc_loss=0.1325, cr_loss=0.3495, over 3347901.64 frames. ], batch size: 61, lr: 4.28e-03, grad_scale: 32.0
2024-09-24 12:03:40,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=497060.6666666667, ans=0.09899494936611666
2024-09-24 12:04:18,679 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.268e+02 1.342e+02 1.439e+02 2.491e+02, threshold=2.683e+02, percent-clipped=0.0
2024-09-24 12:04:23,559 INFO [train.py:1198] (1/4) Epoch 28, batch 1350, loss[loss=0.1972, ctc_loss=0.1292, cr_loss=0.3399, over 17041.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1326, cr_loss=0.3494, over 3347444.16 frames. ], batch size: 52, lr: 4.28e-03, grad_scale: 32.0
2024-09-24 12:04:41,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=497247.3333333333, ans=0.125
2024-09-24 12:04:52,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=497247.3333333333, ans=0.125
2024-09-24 12:04:55,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=497247.3333333333, ans=0.125
2024-09-24 12:05:04,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=497294.0, ans=0.125
2024-09-24 12:05:17,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=497340.6666666667, ans=0.0
2024-09-24 12:05:45,900 INFO [train.py:1198] (1/4) Epoch 28, batch 1400, loss[loss=0.2098, ctc_loss=0.1408, cr_loss=0.3452, over 17013.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1323, cr_loss=0.3494, over 3359788.39 frames. ], batch size: 51, lr: 4.27e-03, grad_scale: 16.0
2024-09-24 12:05:54,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=497434.0, ans=0.0
2024-09-24 12:05:59,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=497434.0, ans=0.2
2024-09-24 12:06:28,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=497527.3333333333, ans=0.05
2024-09-24 12:07:00,665 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.60 vs. limit=22.5
2024-09-24 12:07:08,054 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.287e+02 1.409e+02 1.543e+02 2.036e+02, threshold=2.818e+02, percent-clipped=0.0
2024-09-24 12:07:09,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.87 vs. limit=10.0
2024-09-24 12:07:11,200 INFO [train.py:1198] (1/4) Epoch 28, batch 1450, loss[loss=0.1944, ctc_loss=0.1244, cr_loss=0.3501, over 17012.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1319, cr_loss=0.349, over 3368027.41 frames. ], batch size: 51, lr: 4.27e-03, grad_scale: 16.0
2024-09-24 12:07:17,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=497667.3333333333, ans=0.025
2024-09-24 12:07:22,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=497667.3333333333, ans=6.0
2024-09-24 12:07:30,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=497714.0, ans=0.0
2024-09-24 12:07:33,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=497714.0, ans=0.0
2024-09-24 12:07:41,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=497760.6666666667, ans=0.05
2024-09-24 12:08:16,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=497854.0, ans=0.125
2024-09-24 12:08:22,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=497854.0, ans=0.125
2024-09-24 12:08:23,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=497854.0, ans=0.1
2024-09-24 12:08:30,998 INFO [train.py:1198] (1/4) Epoch 28, batch 1500, loss[loss=0.1842, ctc_loss=0.1202, cr_loss=0.3203, over 17179.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1321, cr_loss=0.3491, over 3372589.06 frames. ], batch size: 45, lr: 4.27e-03, grad_scale: 16.0
2024-09-24 12:08:50,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=497947.3333333333, ans=0.0
2024-09-24 12:08:57,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=497947.3333333333, ans=0.125
2024-09-24 12:09:10,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=497994.0, ans=0.125
2024-09-24 12:09:50,698 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.264e+02 1.369e+02 1.478e+02 2.054e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-24 12:09:53,873 INFO [train.py:1198] (1/4) Epoch 28, batch 1550, loss[loss=0.1974, ctc_loss=0.1302, cr_loss=0.336, over 17247.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1323, cr_loss=0.3495, over 3366640.24 frames. ], batch size: 44, lr: 4.27e-03, grad_scale: 16.0
2024-09-24 12:10:02,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.61 vs. limit=15.0
2024-09-24 12:10:47,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=498274.0, ans=0.0
2024-09-24 12:10:47,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=498274.0, ans=0.125
2024-09-24 12:11:06,422 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.62 vs. limit=15.0
2024-09-24 12:11:11,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=498320.6666666667, ans=0.1
2024-09-24 12:11:13,886 INFO [train.py:1198] (1/4) Epoch 28, batch 1600, loss[loss=0.199, ctc_loss=0.1287, cr_loss=0.3514, over 17363.00 frames. ], tot_loss[loss=0.2028, ctc_loss=0.1328, cr_loss=0.3501, over 3352217.54 frames. ], batch size: 48, lr: 4.27e-03, grad_scale: 32.0
2024-09-24 12:11:30,561 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.29 vs. limit=15.0
2024-09-24 12:11:35,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=498414.0, ans=0.125
2024-09-24 12:11:55,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=498460.6666666667, ans=0.07
2024-09-24 12:11:58,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=498460.6666666667, ans=0.1
2024-09-24 12:12:08,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.97 vs. limit=15.0
2024-09-24 12:12:17,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=498507.3333333333, ans=0.2
2024-09-24 12:12:24,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=498554.0, ans=0.125
2024-09-24 12:12:35,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=498554.0, ans=0.0
2024-09-24 12:12:38,198 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.239e+02 1.329e+02 1.427e+02 2.041e+02, threshold=2.658e+02, percent-clipped=0.0
2024-09-24 12:12:41,511 INFO [train.py:1198] (1/4) Epoch 28, batch 1650, loss[loss=0.2242, ctc_loss=0.1467, cr_loss=0.3874, over 17209.00 frames. ], tot_loss[loss=0.2029, ctc_loss=0.1329, cr_loss=0.35, over 3339064.00 frames. ], batch size: 55, lr: 4.27e-03, grad_scale: 32.0
2024-09-24 12:12:41,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=498600.6666666667, ans=0.125
2024-09-24 12:12:46,914 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.86 vs. limit=10.0
2024-09-24 12:12:54,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=5.14 vs. limit=12.0
2024-09-24 12:13:02,914 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.49 vs. limit=15.0
2024-09-24 12:14:00,723 INFO [train.py:1198] (1/4) Epoch 28, batch 1700, loss[loss=0.2174, ctc_loss=0.1441, cr_loss=0.3662, over 15923.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1331, cr_loss=0.3502, over 3339398.63 frames. ], batch size: 74, lr: 4.27e-03, grad_scale: 32.0
2024-09-24 12:14:00,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=498834.0, ans=0.0
2024-09-24 12:14:15,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=498880.6666666667, ans=0.125
2024-09-24 12:14:17,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=498880.6666666667, ans=0.1
2024-09-24 12:14:46,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=498927.3333333333, ans=0.0
2024-09-24 12:15:19,451 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.240e+02 1.314e+02 1.418e+02 1.778e+02, threshold=2.628e+02, percent-clipped=0.0
2024-09-24 12:15:22,616 INFO [train.py:1198] (1/4) Epoch 28, batch 1750, loss[loss=0.2175, ctc_loss=0.1443, cr_loss=0.3663, over 17236.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1325, cr_loss=0.3486, over 3341821.67 frames. ], batch size: 50, lr: 4.27e-03, grad_scale: 32.0
2024-09-24 12:15:24,569 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=499067.3333333333, ans=0.125
2024-09-24 12:15:45,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=499114.0, ans=0.125
2024-09-24 12:16:04,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=499160.6666666667, ans=0.0
2024-09-24 12:16:10,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=499207.3333333333, ans=0.2
2024-09-24 12:16:13,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=499207.3333333333, ans=0.5
2024-09-24 12:16:24,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=499254.0, ans=0.07
2024-09-24 12:16:47,612 INFO [train.py:1198] (1/4) Epoch 28, batch 1800, loss[loss=0.1709, ctc_loss=0.1085, cr_loss=0.3122, over 17103.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1321, cr_loss=0.3488, over 3339721.24 frames. ], batch size: 40, lr: 4.27e-03, grad_scale: 32.0
2024-09-24 12:17:00,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=499300.6666666667, ans=0.95
2024-09-24 12:17:04,458 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.50 vs. limit=10.0
2024-09-24 12:17:26,042 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 12:17:26,714 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.06 vs. limit=15.0
2024-09-24 12:17:41,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=499440.6666666667, ans=0.0
2024-09-24 12:17:46,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=499440.6666666667, ans=0.125
2024-09-24 12:17:49,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=499487.3333333333, ans=0.0
2024-09-24 12:17:54,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=499487.3333333333, ans=0.1
2024-09-24 12:17:59,514 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=9.86 vs. limit=22.5
2024-09-24 12:18:00,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=499487.3333333333, ans=0.0
2024-09-24 12:18:03,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.272e+02 1.347e+02 1.424e+02 2.124e+02, threshold=2.693e+02, percent-clipped=0.0
2024-09-24 12:18:06,858 INFO [train.py:1198] (1/4) Epoch 28, batch 1850, loss[loss=0.2119, ctc_loss=0.1374, cr_loss=0.3722, over 17220.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.132, cr_loss=0.3485, over 3334706.93 frames. ], batch size: 55, lr: 4.27e-03, grad_scale: 32.0
2024-09-24 12:18:48,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=499627.3333333333, ans=0.025
2024-09-24 12:19:29,609 INFO [train.py:1198] (1/4) Epoch 28, batch 1900, loss[loss=0.2032, ctc_loss=0.1321, cr_loss=0.3554, over 17343.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.132, cr_loss=0.3485, over 3340747.96 frames. ], batch size: 52, lr: 4.26e-03, grad_scale: 32.0
2024-09-24 12:19:29,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=499767.3333333333, ans=0.0
2024-09-24 12:20:01,228 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.87 vs. limit=15.0
2024-09-24 12:20:02,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=499860.6666666667, ans=0.1
2024-09-24 12:20:08,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=499860.6666666667, ans=0.2
2024-09-24 12:20:14,777 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 12:20:17,215 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.40 vs. limit=6.0
2024-09-24 12:20:38,884 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 12:20:46,566 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.256e+02 1.349e+02 1.450e+02 2.234e+02, threshold=2.698e+02, percent-clipped=0.0
2024-09-24 12:20:49,741 INFO [train.py:1198] (1/4) Epoch 28, batch 1950, loss[loss=0.2079, ctc_loss=0.1347, cr_loss=0.3662, over 17035.00 frames. ], tot_loss[loss=0.2021, ctc_loss=0.1322, cr_loss=0.3492, over 3350297.95 frames. ], batch size: 44, lr: 4.26e-03, grad_scale: 32.0
2024-09-24 12:21:17,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=500047.3333333333, ans=0.125
2024-09-24 12:21:37,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=500094.0, ans=0.1
2024-09-24 12:21:39,464 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.61 vs. limit=22.5
2024-09-24 12:22:06,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=500187.3333333333, ans=0.1
2024-09-24 12:22:17,929 INFO [train.py:1198] (1/4) Epoch 28, batch 2000, loss[loss=0.2048, ctc_loss=0.1327, cr_loss=0.3605, over 17070.00 frames. ], tot_loss[loss=0.2028, ctc_loss=0.1327, cr_loss=0.3503, over 3350145.90 frames. ], batch size: 46, lr: 4.26e-03, grad_scale: 32.0
2024-09-24 12:22:23,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=500234.0, ans=0.125
2024-09-24 12:22:28,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.06 vs. limit=15.0
2024-09-24 12:22:47,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=500280.6666666667, ans=0.07
2024-09-24 12:23:32,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=500420.6666666667, ans=0.0
2024-09-24 12:23:34,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.66 vs. limit=15.0
2024-09-24 12:23:35,003 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.274e+02 1.350e+02 1.459e+02 2.226e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-24 12:23:38,213 INFO [train.py:1198] (1/4) Epoch 28, batch 2050, loss[loss=0.2294, ctc_loss=0.1498, cr_loss=0.3982, over 16997.00 frames. ], tot_loss[loss=0.2035, ctc_loss=0.1332, cr_loss=0.3515, over 3355550.63 frames. ], batch size: 53, lr: 4.26e-03, grad_scale: 32.0
2024-09-24 12:23:49,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=500467.3333333333, ans=0.125
2024-09-24 12:24:19,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.whiten.whitening_limit, batch_count=500560.6666666667, ans=12.0
2024-09-24 12:24:50,371 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.95 vs. limit=15.0
2024-09-24 12:24:54,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=500654.0, ans=0.125
2024-09-24 12:24:57,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=500654.0, ans=0.0
2024-09-24 12:25:00,539 INFO [train.py:1198] (1/4) Epoch 28, batch 2100, loss[loss=0.2012, ctc_loss=0.1311, cr_loss=0.3505, over 17053.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1334, cr_loss=0.3517, over 3352412.24 frames. ], batch size: 52, lr: 4.26e-03, grad_scale: 32.0
2024-09-24 12:25:07,747 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=6.28 vs. limit=15.0
2024-09-24 12:25:23,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=500747.3333333333, ans=0.035
2024-09-24 12:25:24,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=500747.3333333333, ans=0.125
2024-09-24 12:25:38,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=500794.0, ans=0.0
2024-09-24 12:25:53,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=500840.6666666667, ans=0.0
2024-09-24 12:26:02,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=500887.3333333333, ans=0.1
2024-09-24 12:26:16,783 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.293e+02 1.395e+02 1.538e+02 2.200e+02, threshold=2.790e+02, percent-clipped=0.0
2024-09-24 12:26:25,165 INFO [train.py:1198] (1/4) Epoch 28, batch 2150, loss[loss=0.1724, ctc_loss=0.1102, cr_loss=0.3109, over 17125.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1336, cr_loss=0.3511, over 3347724.17 frames. ], batch size: 40, lr: 4.26e-03, grad_scale: 32.0
2024-09-24 12:26:51,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.max_positive, batch_count=500980.6666666667, ans=0.95
2024-09-24 12:26:59,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=501027.3333333333, ans=0.1
2024-09-24 12:27:04,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=501027.3333333333, ans=0.125
2024-09-24 12:27:11,589 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.99 vs. limit=6.0
2024-09-24 12:27:47,640 INFO [train.py:1198] (1/4) Epoch 28, batch 2200, loss[loss=0.2232, ctc_loss=0.1506, cr_loss=0.3633, over 15928.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1334, cr_loss=0.3508, over 3349175.71 frames. ], batch size: 74, lr: 4.26e-03, grad_scale: 16.0
2024-09-24 12:28:05,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=501214.0, ans=0.035
2024-09-24 12:28:08,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=501214.0, ans=0.1
2024-09-24 12:28:16,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=501214.0, ans=0.125
2024-09-24 12:28:28,514 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.89 vs. limit=22.5
2024-09-24 12:28:28,655 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.76 vs. limit=15.0
2024-09-24 12:28:34,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.18 vs. limit=6.0
2024-09-24 12:29:06,263 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.288e+02 1.410e+02 1.537e+02 2.040e+02, threshold=2.819e+02, percent-clipped=0.0
2024-09-24 12:29:10,548 INFO [train.py:1198] (1/4) Epoch 28, batch 2250, loss[loss=0.1902, ctc_loss=0.1243, cr_loss=0.3295, over 17294.00 frames. ], tot_loss[loss=0.2049, ctc_loss=0.1344, cr_loss=0.3526, over 3347625.33 frames. ], batch size: 49, lr: 4.26e-03, grad_scale: 16.0
2024-09-24 12:29:37,475 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.94 vs. limit=15.0
2024-09-24 12:29:40,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.43 vs. limit=15.0
2024-09-24 12:30:26,787 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.08 vs. limit=10.0
2024-09-24 12:30:30,752 INFO [train.py:1198] (1/4) Epoch 28, batch 2300, loss[loss=0.2168, ctc_loss=0.1428, cr_loss=0.37, over 17301.00 frames. ], tot_loss[loss=0.2054, ctc_loss=0.1347, cr_loss=0.3531, over 3342849.87 frames. ], batch size: 51, lr: 4.26e-03, grad_scale: 16.0
2024-09-24 12:30:42,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=501634.0, ans=0.04949747468305833
2024-09-24 12:31:06,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=501727.3333333333, ans=0.2
2024-09-24 12:31:22,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=501774.0, ans=0.125
2024-09-24 12:31:25,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=501774.0, ans=0.2
2024-09-24 12:31:56,843 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.285e+02 1.365e+02 1.487e+02 3.397e+02, threshold=2.730e+02, percent-clipped=1.0
2024-09-24 12:31:58,415 INFO [train.py:1198] (1/4) Epoch 28, batch 2350, loss[loss=0.227, ctc_loss=0.152, cr_loss=0.3749, over 16990.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.134, cr_loss=0.3519, over 3344127.00 frames. ], batch size: 56, lr: 4.26e-03, grad_scale: 16.0
2024-09-24 12:32:01,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=501867.3333333333, ans=0.0
2024-09-24 12:32:08,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=501867.3333333333, ans=0.125
2024-09-24 12:32:45,125 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.58 vs. limit=15.0
2024-09-24 12:33:00,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=502054.0, ans=0.1
2024-09-24 12:33:02,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=502054.0, ans=0.0
2024-09-24 12:33:05,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=502054.0, ans=0.1
2024-09-24 12:33:10,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=502054.0, ans=0.0
2024-09-24 12:33:17,990 INFO [train.py:1198] (1/4) Epoch 28, batch 2400, loss[loss=0.1767, ctc_loss=0.1141, cr_loss=0.3133, over 17063.00 frames. ], tot_loss[loss=0.2034, ctc_loss=0.1333, cr_loss=0.3508, over 3345047.14 frames. ], batch size: 39, lr: 4.25e-03, grad_scale: 32.0
2024-09-24 12:33:34,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=502147.3333333333, ans=0.1
2024-09-24 12:34:27,172 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 12:34:39,437 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.025e+02 1.258e+02 1.381e+02 1.475e+02 2.172e+02, threshold=2.761e+02, percent-clipped=0.0
2024-09-24 12:34:41,086 INFO [train.py:1198] (1/4) Epoch 28, batch 2450, loss[loss=0.2001, ctc_loss=0.1313, cr_loss=0.3444, over 16751.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.133, cr_loss=0.351, over 3354544.22 frames. ], batch size: 61, lr: 4.25e-03, grad_scale: 32.0
2024-09-24 12:34:41,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=502334.0, ans=0.2
2024-09-24 12:35:33,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=502474.0, ans=0.125
2024-09-24 12:35:33,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=502474.0, ans=0.125
2024-09-24 12:35:59,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=502567.3333333333, ans=0.125
2024-09-24 12:36:00,606 INFO [train.py:1198] (1/4) Epoch 28, batch 2500, loss[loss=0.2204, ctc_loss=0.1459, cr_loss=0.3728, over 17029.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.1331, cr_loss=0.3506, over 3344359.34 frames. ], batch size: 52, lr: 4.25e-03, grad_scale: 16.0
2024-09-24 12:36:02,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=502567.3333333333, ans=0.125
2024-09-24 12:36:27,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.05 vs. limit=15.0
2024-09-24 12:36:58,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=502707.3333333333, ans=0.1
2024-09-24 12:37:18,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=5.01 vs. limit=15.0
2024-09-24 12:37:20,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=502754.0, ans=0.07
2024-09-24 12:37:28,345 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.269e+02 1.387e+02 1.478e+02 2.068e+02, threshold=2.774e+02, percent-clipped=0.0
2024-09-24 12:37:28,369 INFO [train.py:1198] (1/4) Epoch 28, batch 2550, loss[loss=0.2295, ctc_loss=0.1512, cr_loss=0.3915, over 16496.00 frames. ], tot_loss[loss=0.2042, ctc_loss=0.1338, cr_loss=0.3521, over 3345459.26 frames. ], batch size: 66, lr: 4.25e-03, grad_scale: 16.0
2024-09-24 12:37:41,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=502800.6666666667, ans=0.015
2024-09-24 12:38:10,550 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.52 vs. limit=12.0
2024-09-24 12:38:24,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=502940.6666666667, ans=0.125
2024-09-24 12:38:24,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=10.24 vs. limit=15.0
2024-09-24 12:38:30,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=502987.3333333333, ans=0.125
2024-09-24 12:38:38,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=502987.3333333333, ans=0.0
2024-09-24 12:38:48,319 INFO [train.py:1198] (1/4) Epoch 28, batch 2600, loss[loss=0.1981, ctc_loss=0.1297, cr_loss=0.3415, over 17301.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.1341, cr_loss=0.3527, over 3350712.24 frames. ], batch size: 51, lr: 4.25e-03, grad_scale: 16.0
2024-09-24 12:38:49,376 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.79 vs. limit=22.5
2024-09-24 12:38:50,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=503034.0, ans=0.1
2024-09-24 12:39:07,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=503080.6666666667, ans=10.0
2024-09-24 12:39:18,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=503080.6666666667, ans=0.2
2024-09-24 12:39:18,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=503080.6666666667, ans=0.1
2024-09-24 12:39:26,447 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 12:39:29,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=503127.3333333333, ans=0.125
2024-09-24 12:39:33,394 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.92 vs. limit=15.0
2024-09-24 12:39:52,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=503174.0, ans=0.0
2024-09-24 12:39:58,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=503220.6666666667, ans=0.125
2024-09-24 12:40:10,912 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.254e+02 1.337e+02 1.417e+02 2.017e+02, threshold=2.675e+02, percent-clipped=0.0
2024-09-24 12:40:10,936 INFO [train.py:1198] (1/4) Epoch 28, batch 2650, loss[loss=0.1998, ctc_loss=0.13, cr_loss=0.3488, over 17070.00 frames. ], tot_loss[loss=0.2044, ctc_loss=0.1339, cr_loss=0.3524, over 3356785.64 frames. ], batch size: 46, lr: 4.25e-03, grad_scale: 16.0
2024-09-24 12:40:12,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=503267.3333333333, ans=0.125
2024-09-24 12:40:47,371 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.49 vs. limit=22.5
2024-09-24 12:40:56,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=503360.6666666667, ans=0.125
2024-09-24 12:40:57,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=503407.3333333333, ans=0.0
2024-09-24 12:41:05,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=503407.3333333333, ans=0.125
2024-09-24 12:41:23,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=503454.0, ans=0.125
2024-09-24 12:41:27,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=503454.0, ans=0.0
2024-09-24 12:41:37,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=503500.6666666667, ans=0.125
2024-09-24 12:41:38,425 INFO [train.py:1198] (1/4) Epoch 28, batch 2700, loss[loss=0.2177, ctc_loss=0.1574, cr_loss=0.3016, over 11589.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1335, cr_loss=0.3513, over 3357421.62 frames. ], batch size: 123, lr: 4.25e-03, grad_scale: 16.0
2024-09-24 12:41:45,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=503500.6666666667, ans=0.2
2024-09-24 12:42:06,736 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.84 vs. limit=12.0
2024-09-24 12:42:21,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=503594.0, ans=0.125
2024-09-24 12:42:25,279 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.70 vs. limit=15.0
2024-09-24 12:42:47,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=503687.3333333333, ans=0.0
2024-09-24 12:42:49,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=503687.3333333333, ans=0.0
2024-09-24 12:42:53,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.63 vs. limit=22.5
2024-09-24 12:42:58,696 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.278e+02 1.354e+02 1.486e+02 3.287e+02, threshold=2.708e+02, percent-clipped=2.0
2024-09-24 12:42:58,720 INFO [train.py:1198] (1/4) Epoch 28, batch 2750, loss[loss=0.1803, ctc_loss=0.1188, cr_loss=0.3076, over 17067.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1334, cr_loss=0.3522, over 3359543.75 frames. ], batch size: 46, lr: 4.25e-03, grad_scale: 16.0
2024-09-24 12:43:40,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=503827.3333333333, ans=0.2
2024-09-24 12:43:58,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=503874.0, ans=0.125
2024-09-24 12:44:01,209 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.07 vs. limit=12.0
2024-09-24 12:44:01,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=503874.0, ans=0.125
2024-09-24 12:44:09,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=503920.6666666667, ans=0.125
2024-09-24 12:44:14,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=503920.6666666667, ans=0.0
2024-09-24 12:44:20,597 INFO [train.py:1198] (1/4) Epoch 28, batch 2800, loss[loss=0.2263, ctc_loss=0.1446, cr_loss=0.4087, over 17022.00 frames. ], tot_loss[loss=0.2043, ctc_loss=0.1338, cr_loss=0.3528, over 3351015.97 frames. ], batch size: 53, lr: 4.25e-03, grad_scale: 32.0
2024-09-24 12:44:48,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=504014.0, ans=0.1
2024-09-24 12:45:04,902 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 12:45:14,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=504107.3333333333, ans=0.125
2024-09-24 12:45:27,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=504154.0, ans=0.1
2024-09-24 12:45:38,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=504154.0, ans=0.125
2024-09-24 12:45:43,157 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.267e+02 1.390e+02 1.555e+02 2.170e+02, threshold=2.781e+02, percent-clipped=0.0
2024-09-24 12:45:43,182 INFO [train.py:1198] (1/4) Epoch 28, batch 2850, loss[loss=0.1831, ctc_loss=0.1183, cr_loss=0.3237, over 17096.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.134, cr_loss=0.3527, over 3344516.31 frames. ], batch size: 40, lr: 4.25e-03, grad_scale: 32.0
2024-09-24 12:45:46,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=504200.6666666667, ans=0.125
2024-09-24 12:45:59,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=504247.3333333333, ans=0.0
2024-09-24 12:46:39,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=504340.6666666667, ans=0.125
2024-09-24 12:46:42,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=504340.6666666667, ans=0.0
2024-09-24 12:46:48,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=504340.6666666667, ans=0.125
2024-09-24 12:46:56,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=504387.3333333333, ans=0.2
2024-09-24 12:47:10,934 INFO [train.py:1198] (1/4) Epoch 28, batch 2900, loss[loss=0.2115, ctc_loss=0.1375, cr_loss=0.3698, over 17371.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.1331, cr_loss=0.3509, over 3348387.28 frames. ], batch size: 48, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:47:19,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=504434.0, ans=0.125
2024-09-24 12:47:30,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=504480.6666666667, ans=0.1
2024-09-24 12:47:41,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=504527.3333333333, ans=0.125
2024-09-24 12:48:31,272 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.289e+02 1.394e+02 1.570e+02 3.133e+02, threshold=2.787e+02, percent-clipped=1.0
2024-09-24 12:48:31,296 INFO [train.py:1198] (1/4) Epoch 28, batch 2950, loss[loss=0.22, ctc_loss=0.1419, cr_loss=0.3904, over 16911.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1333, cr_loss=0.3512, over 3348586.77 frames. ], batch size: 58, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:48:31,742 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 12:48:42,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=504667.3333333333, ans=0.125
2024-09-24 12:49:00,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=504714.0, ans=0.035
2024-09-24 12:49:04,918 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.91 vs. limit=12.0
2024-09-24 12:49:07,668 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.95 vs. limit=22.5
2024-09-24 12:49:09,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=504760.6666666667, ans=0.0
2024-09-24 12:49:13,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=504760.6666666667, ans=0.125
2024-09-24 12:49:51,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=504900.6666666667, ans=0.2
2024-09-24 12:49:53,070 INFO [train.py:1198] (1/4) Epoch 28, batch 3000, loss[loss=0.2002, ctc_loss=0.1327, cr_loss=0.3372, over 17210.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.1341, cr_loss=0.3524, over 3340704.98 frames. ], batch size: 47, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:49:53,071 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 12:50:08,098 INFO [train.py:1230] (1/4) Epoch 28, validation: loss=0.03718, ctc_loss=0.03718, cr_loss=8.452e-15, over 944034.00 frames. 
2024-09-24 12:50:08,098 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 12:50:35,827 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.32 vs. limit=22.5
2024-09-24 12:50:47,239 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=10.91 vs. limit=22.5
2024-09-24 12:51:10,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.53 vs. limit=6.0
2024-09-24 12:51:22,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=505087.3333333333, ans=0.0
2024-09-24 12:51:26,594 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.294e+02 1.362e+02 1.471e+02 2.139e+02, threshold=2.723e+02, percent-clipped=0.0
2024-09-24 12:51:26,619 INFO [train.py:1198] (1/4) Epoch 28, batch 3050, loss[loss=0.2062, ctc_loss=0.135, cr_loss=0.356, over 17214.00 frames. ], tot_loss[loss=0.2035, ctc_loss=0.1332, cr_loss=0.3513, over 3353174.80 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:51:36,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=505134.0, ans=0.1
2024-09-24 12:51:54,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=505180.6666666667, ans=0.125
2024-09-24 12:51:56,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=505227.3333333333, ans=0.025
2024-09-24 12:51:59,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=505227.3333333333, ans=0.125
2024-09-24 12:51:59,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=505227.3333333333, ans=0.2
2024-09-24 12:52:04,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=505227.3333333333, ans=0.125
2024-09-24 12:52:18,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten.whitening_limit, batch_count=505274.0, ans=15.0
2024-09-24 12:52:25,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=505274.0, ans=0.0
2024-09-24 12:52:48,943 INFO [train.py:1198] (1/4) Epoch 28, batch 3100, loss[loss=0.1578, ctc_loss=0.0989, cr_loss=0.2947, over 17076.00 frames. ], tot_loss[loss=0.2031, ctc_loss=0.1329, cr_loss=0.3511, over 3354397.53 frames. ], batch size: 39, lr: 4.24e-03, grad_scale: 16.0
2024-09-24 12:52:49,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=505367.3333333333, ans=0.125
2024-09-24 12:52:52,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=505367.3333333333, ans=0.025
2024-09-24 12:52:52,539 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.84 vs. limit=15.0
2024-09-24 12:52:53,209 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.82 vs. limit=6.0
2024-09-24 12:52:56,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=505367.3333333333, ans=0.125
2024-09-24 12:53:32,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=505460.6666666667, ans=0.0
2024-09-24 12:53:38,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=505507.3333333333, ans=0.1
2024-09-24 12:53:43,778 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.83 vs. limit=15.0
2024-09-24 12:53:53,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=505554.0, ans=0.1
2024-09-24 12:54:00,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=505554.0, ans=0.0
2024-09-24 12:54:09,360 INFO [train.py:1198] (1/4) Epoch 28, batch 3150, loss[loss=0.1888, ctc_loss=0.1202, cr_loss=0.3428, over 17189.00 frames. ], tot_loss[loss=0.2033, ctc_loss=0.133, cr_loss=0.3514, over 3356924.77 frames. ], batch size: 45, lr: 4.24e-03, grad_scale: 16.0
2024-09-24 12:54:10,884 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.260e+02 1.340e+02 1.441e+02 3.228e+02, threshold=2.680e+02, percent-clipped=2.0
2024-09-24 12:54:12,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=505600.6666666667, ans=0.125
2024-09-24 12:54:34,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=505647.3333333333, ans=0.125
2024-09-24 12:54:37,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=505647.3333333333, ans=0.2
2024-09-24 12:54:38,388 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=14.17 vs. limit=15.0
2024-09-24 12:54:47,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=505694.0, ans=0.125
2024-09-24 12:54:53,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=505694.0, ans=0.0
2024-09-24 12:55:10,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=505787.3333333333, ans=0.125
2024-09-24 12:55:15,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=505787.3333333333, ans=0.125
2024-09-24 12:55:15,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=505787.3333333333, ans=0.125
2024-09-24 12:55:16,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=505787.3333333333, ans=0.125
2024-09-24 12:55:27,146 INFO [train.py:1198] (1/4) Epoch 28, batch 3200, loss[loss=0.2117, ctc_loss=0.1385, cr_loss=0.3661, over 16869.00 frames. ], tot_loss[loss=0.2029, ctc_loss=0.1328, cr_loss=0.3505, over 3355498.37 frames. ], batch size: 58, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:55:27,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=505834.0, ans=0.1
2024-09-24 12:55:50,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=505880.6666666667, ans=0.1
2024-09-24 12:55:59,143 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.03 vs. limit=10.0
2024-09-24 12:56:03,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=505927.3333333333, ans=0.125
2024-09-24 12:56:11,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=505927.3333333333, ans=0.125
2024-09-24 12:56:20,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=505974.0, ans=0.0
2024-09-24 12:56:25,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=505974.0, ans=0.125
2024-09-24 12:56:40,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=506020.6666666667, ans=0.0
2024-09-24 12:56:45,209 INFO [train.py:1198] (1/4) Epoch 28, batch 3250, loss[loss=0.1891, ctc_loss=0.1211, cr_loss=0.3397, over 17254.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1325, cr_loss=0.3499, over 3358691.86 frames. ], batch size: 44, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:56:46,875 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.036e+02 1.261e+02 1.336e+02 1.430e+02 2.422e+02, threshold=2.672e+02, percent-clipped=0.0
2024-09-24 12:57:11,223 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.31 vs. limit=15.0
2024-09-24 12:57:11,461 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.60 vs. limit=15.0
2024-09-24 12:57:13,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=506114.0, ans=0.2
2024-09-24 12:57:33,045 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.18 vs. limit=10.0
2024-09-24 12:57:45,592 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.13 vs. limit=22.5
2024-09-24 12:58:04,252 INFO [train.py:1198] (1/4) Epoch 28, batch 3300, loss[loss=0.236, ctc_loss=0.1576, cr_loss=0.3916, over 17329.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1333, cr_loss=0.3515, over 3360971.54 frames. ], batch size: 48, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:58:15,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=506300.6666666667, ans=0.125
2024-09-24 12:58:20,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=506347.3333333333, ans=0.125
2024-09-24 12:58:26,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=506347.3333333333, ans=0.1
2024-09-24 12:58:32,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=506347.3333333333, ans=0.125
2024-09-24 12:58:38,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=506394.0, ans=10.0
2024-09-24 12:58:42,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=506394.0, ans=0.1
2024-09-24 12:58:56,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.min_positive, batch_count=506440.6666666667, ans=0.025
2024-09-24 12:59:06,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.98 vs. limit=15.0
2024-09-24 12:59:07,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=506487.3333333333, ans=0.1
2024-09-24 12:59:12,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=506487.3333333333, ans=0.2
2024-09-24 12:59:24,426 INFO [train.py:1198] (1/4) Epoch 28, batch 3350, loss[loss=0.1863, ctc_loss=0.1228, cr_loss=0.3173, over 17309.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1331, cr_loss=0.3506, over 3351109.11 frames. ], batch size: 49, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 12:59:25,944 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.246e+02 1.321e+02 1.387e+02 1.674e+02, threshold=2.642e+02, percent-clipped=0.0
2024-09-24 12:59:49,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=506580.6666666667, ans=0.125
2024-09-24 13:00:10,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=506674.0, ans=0.1
2024-09-24 13:00:25,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=506720.6666666667, ans=0.125
2024-09-24 13:00:34,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=506720.6666666667, ans=0.125
2024-09-24 13:00:42,380 INFO [train.py:1198] (1/4) Epoch 28, batch 3400, loss[loss=0.2157, ctc_loss=0.1437, cr_loss=0.3599, over 17151.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1325, cr_loss=0.3497, over 3348280.32 frames. ], batch size: 48, lr: 4.24e-03, grad_scale: 32.0
2024-09-24 13:00:55,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=506767.3333333333, ans=0.0
2024-09-24 13:00:57,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=506814.0, ans=0.125
2024-09-24 13:00:58,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=506814.0, ans=0.0
2024-09-24 13:01:13,010 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=5.54 vs. limit=15.0
2024-09-24 13:01:40,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=506907.3333333333, ans=0.2
2024-09-24 13:01:43,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=506954.0, ans=0.2
2024-09-24 13:01:51,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=506954.0, ans=0.1
2024-09-24 13:02:00,540 INFO [train.py:1198] (1/4) Epoch 28, batch 3450, loss[loss=0.2117, ctc_loss=0.135, cr_loss=0.3832, over 17160.00 frames. ], tot_loss[loss=0.203, ctc_loss=0.133, cr_loss=0.3501, over 3348702.73 frames. ], batch size: 45, lr: 4.23e-03, grad_scale: 32.0
2024-09-24 13:02:00,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=507000.6666666667, ans=0.125
2024-09-24 13:02:02,018 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.326e+02 1.438e+02 1.586e+02 2.934e+02, threshold=2.877e+02, percent-clipped=1.0
2024-09-24 13:02:06,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=507000.6666666667, ans=0.1
2024-09-24 13:02:08,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=507000.6666666667, ans=0.0
2024-09-24 13:02:33,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=507094.0, ans=0.125
2024-09-24 13:02:40,830 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.47 vs. limit=22.5
2024-09-24 13:02:42,438 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.95 vs. limit=15.0
2024-09-24 13:02:56,628 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.74 vs. limit=15.0
2024-09-24 13:03:01,102 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 13:03:19,717 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.85 vs. limit=10.0
2024-09-24 13:03:25,066 INFO [train.py:1198] (1/4) Epoch 28, batch 3500, loss[loss=0.2154, ctc_loss=0.1433, cr_loss=0.3609, over 16407.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1334, cr_loss=0.3509, over 3342913.77 frames. ], batch size: 66, lr: 4.23e-03, grad_scale: 32.0
2024-09-24 13:03:31,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=507234.0, ans=0.1
2024-09-24 13:03:36,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=507234.0, ans=0.2
2024-09-24 13:03:41,232 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.52 vs. limit=6.0
2024-09-24 13:03:56,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=507327.3333333333, ans=0.2
2024-09-24 13:03:56,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=507327.3333333333, ans=0.125
2024-09-24 13:04:13,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=507374.0, ans=0.0
2024-09-24 13:04:36,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=507420.6666666667, ans=0.05
2024-09-24 13:04:42,941 INFO [train.py:1198] (1/4) Epoch 28, batch 3550, loss[loss=0.1491, ctc_loss=0.09332, cr_loss=0.2787, over 16309.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1336, cr_loss=0.3516, over 3344202.47 frames. ], batch size: 36, lr: 4.23e-03, grad_scale: 32.0
2024-09-24 13:04:43,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=507467.3333333333, ans=0.2
2024-09-24 13:04:44,466 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.292e+02 1.411e+02 1.555e+02 1.879e+02, threshold=2.822e+02, percent-clipped=0.0
2024-09-24 13:04:49,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=507467.3333333333, ans=0.0
2024-09-24 13:04:55,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=507467.3333333333, ans=0.0
2024-09-24 13:05:02,476 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.36 vs. limit=15.0
2024-09-24 13:05:04,109 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.49 vs. limit=15.0
2024-09-24 13:05:13,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=507560.6666666667, ans=0.2
2024-09-24 13:05:28,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=507607.3333333333, ans=0.125
2024-09-24 13:05:44,569 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.85 vs. limit=10.0
2024-09-24 13:06:00,780 INFO [train.py:1198] (1/4) Epoch 28, batch 3600, loss[loss=0.1912, ctc_loss=0.1252, cr_loss=0.3304, over 17004.00 frames. ], tot_loss[loss=0.2035, ctc_loss=0.1333, cr_loss=0.3512, over 3346294.27 frames. ], batch size: 44, lr: 4.23e-03, grad_scale: 32.0
2024-09-24 13:06:08,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=507700.6666666667, ans=0.125
2024-09-24 13:07:04,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=507887.3333333333, ans=0.125
2024-09-24 13:07:10,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=507887.3333333333, ans=0.0
2024-09-24 13:07:10,731 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.50 vs. limit=22.5
2024-09-24 13:07:13,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=507887.3333333333, ans=0.0
2024-09-24 13:07:18,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=507934.0, ans=10.0
2024-09-24 13:07:19,235 INFO [train.py:1198] (1/4) Epoch 28, batch 3650, loss[loss=0.205, ctc_loss=0.1337, cr_loss=0.3561, over 17090.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.1336, cr_loss=0.3518, over 3336419.43 frames. ], batch size: 49, lr: 4.23e-03, grad_scale: 32.0
2024-09-24 13:07:20,720 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.240e+02 1.329e+02 1.473e+02 2.274e+02, threshold=2.658e+02, percent-clipped=0.0
2024-09-24 13:07:30,483 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 13:07:36,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=507980.6666666667, ans=0.125
2024-09-24 13:07:39,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=507980.6666666667, ans=0.2
2024-09-24 13:07:44,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=507980.6666666667, ans=0.0
2024-09-24 13:08:14,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=508074.0, ans=0.125
2024-09-24 13:08:40,141 INFO [train.py:1198] (1/4) Epoch 28, batch 3700, loss[loss=0.2302, ctc_loss=0.1519, cr_loss=0.3912, over 16465.00 frames. ], tot_loss[loss=0.2029, ctc_loss=0.1328, cr_loss=0.3505, over 3352750.81 frames. ], batch size: 66, lr: 4.23e-03, grad_scale: 32.0
2024-09-24 13:09:42,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=508354.0, ans=0.125
2024-09-24 13:09:43,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=508354.0, ans=0.125
2024-09-24 13:09:45,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=508354.0, ans=0.125
2024-09-24 13:09:58,945 INFO [train.py:1198] (1/4) Epoch 28, batch 3750, loss[loss=0.2359, ctc_loss=0.1588, cr_loss=0.3855, over 14988.00 frames. ], tot_loss[loss=0.2038, ctc_loss=0.1335, cr_loss=0.3515, over 3336982.93 frames. ], batch size: 89, lr: 4.23e-03, grad_scale: 32.0
2024-09-24 13:10:00,437 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.248e+02 1.336e+02 1.422e+02 1.841e+02, threshold=2.673e+02, percent-clipped=0.0
2024-09-24 13:10:00,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=508400.6666666667, ans=0.025
2024-09-24 13:10:03,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=508400.6666666667, ans=0.2
2024-09-24 13:10:04,220 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=10.55 vs. limit=12.0
2024-09-24 13:10:33,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=508494.0, ans=0.125
2024-09-24 13:10:40,412 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.61 vs. limit=15.0
2024-09-24 13:10:49,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=508540.6666666667, ans=0.0
2024-09-24 13:11:01,099 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=5.99 vs. limit=15.0
2024-09-24 13:11:17,814 INFO [train.py:1198] (1/4) Epoch 28, batch 3800, loss[loss=0.2234, ctc_loss=0.1454, cr_loss=0.3897, over 16862.00 frames. ], tot_loss[loss=0.2036, ctc_loss=0.1334, cr_loss=0.3509, over 3314176.85 frames. ], batch size: 58, lr: 4.23e-03, grad_scale: 16.0
2024-09-24 13:11:18,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=508634.0, ans=0.1
2024-09-24 13:11:40,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=508680.6666666667, ans=0.025
2024-09-24 13:11:48,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=508727.3333333333, ans=10.0
2024-09-24 13:11:51,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=508727.3333333333, ans=0.0
2024-09-24 13:12:25,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=508820.6666666667, ans=0.2
2024-09-24 13:12:32,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=508820.6666666667, ans=0.1
2024-09-24 13:12:34,422 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=9.21 vs. limit=15.0
2024-09-24 13:12:38,536 INFO [train.py:1198] (1/4) Epoch 28, batch 3850, loss[loss=0.2435, ctc_loss=0.1672, cr_loss=0.3814, over 11828.00 frames. ], tot_loss[loss=0.2041, ctc_loss=0.134, cr_loss=0.3503, over 3281767.85 frames. ], batch size: 123, lr: 4.23e-03, grad_scale: 16.0
2024-09-24 13:12:42,159 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.266e+02 1.354e+02 1.491e+02 2.044e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-24 13:12:51,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=508867.3333333333, ans=0.0
2024-09-24 13:13:23,367 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.15 vs. limit=15.0
2024-09-24 13:14:43,486 INFO [train.py:1198] (1/4) Epoch 29, batch 0, loss[loss=0.1871, ctc_loss=0.1205, cr_loss=0.3331, over 17070.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1205, cr_loss=0.3331, over 17070.00 frames. ], batch size: 43, lr: 4.15e-03, grad_scale: 32.0
2024-09-24 13:14:43,487 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 13:14:57,637 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.4367, 3.9776, 3.8685, 4.0379], device='cuda:1')
2024-09-24 13:14:58,955 INFO [train.py:1230] (1/4) Epoch 29, validation: loss=0.03615, ctc_loss=0.03615, cr_loss=9.405e-15, over 944034.00 frames. 
2024-09-24 13:14:58,955 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 13:15:01,379 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.28 vs. limit=12.0
2024-09-24 13:16:13,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.max_positive, batch_count=509268.6666666667, ans=0.95
2024-09-24 13:16:21,076 INFO [train.py:1198] (1/4) Epoch 29, batch 50, loss[loss=0.2096, ctc_loss=0.1399, cr_loss=0.3488, over 15781.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1333, cr_loss=0.353, over 753474.69 frames. ], batch size: 74, lr: 4.15e-03, grad_scale: 32.0
2024-09-24 13:16:25,440 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.55 vs. limit=15.0
2024-09-24 13:16:26,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=509315.3333333333, ans=0.1
2024-09-24 13:16:30,903 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.316e+02 1.466e+02 1.618e+02 2.901e+02, threshold=2.933e+02, percent-clipped=1.0
2024-09-24 13:16:40,140 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.25 vs. limit=22.5
2024-09-24 13:16:40,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=509362.0, ans=0.125
2024-09-24 13:16:56,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=509408.6666666667, ans=0.0
2024-09-24 13:17:09,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=509455.3333333333, ans=0.0
2024-09-24 13:17:44,485 INFO [train.py:1198] (1/4) Epoch 29, batch 100, loss[loss=0.1953, ctc_loss=0.1252, cr_loss=0.3509, over 17213.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.1335, cr_loss=0.3521, over 1327064.89 frames. ], batch size: 50, lr: 4.15e-03, grad_scale: 32.0
2024-09-24 13:17:54,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=509548.6666666667, ans=0.125
2024-09-24 13:18:01,319 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.47 vs. limit=15.0
2024-09-24 13:18:09,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=509595.3333333333, ans=0.1
2024-09-24 13:18:17,711 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.14 vs. limit=15.0
2024-09-24 13:18:28,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=509642.0, ans=0.0
2024-09-24 13:18:47,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=509688.6666666667, ans=0.125
2024-09-24 13:18:52,801 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.63 vs. limit=12.0
2024-09-24 13:19:09,196 INFO [train.py:1198] (1/4) Epoch 29, batch 150, loss[loss=0.1746, ctc_loss=0.111, cr_loss=0.3175, over 17273.00 frames. ], tot_loss[loss=0.2044, ctc_loss=0.1338, cr_loss=0.3529, over 1761682.92 frames. ], batch size: 42, lr: 4.15e-03, grad_scale: 32.0
2024-09-24 13:19:15,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=509782.0, ans=0.125
2024-09-24 13:19:18,661 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.009e+02 1.258e+02 1.349e+02 1.479e+02 2.103e+02, threshold=2.697e+02, percent-clipped=0.0
2024-09-24 13:19:19,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=509782.0, ans=0.1
2024-09-24 13:19:41,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=509875.3333333333, ans=0.2
2024-09-24 13:20:04,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=509922.0, ans=0.125
2024-09-24 13:20:31,475 INFO [train.py:1198] (1/4) Epoch 29, batch 200, loss[loss=0.1994, ctc_loss=0.1322, cr_loss=0.3361, over 16745.00 frames. ], tot_loss[loss=0.2054, ctc_loss=0.1346, cr_loss=0.3541, over 2118813.46 frames. ], batch size: 61, lr: 4.15e-03, grad_scale: 32.0
2024-09-24 13:20:49,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=510062.0, ans=0.125
2024-09-24 13:20:51,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=510062.0, ans=0.025
2024-09-24 13:21:00,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=510062.0, ans=0.04949747468305833
2024-09-24 13:21:00,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=510062.0, ans=0.125
2024-09-24 13:21:03,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=510108.6666666667, ans=0.125
2024-09-24 13:21:09,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=510108.6666666667, ans=0.125
2024-09-24 13:21:51,165 INFO [train.py:1198] (1/4) Epoch 29, batch 250, loss[loss=0.1898, ctc_loss=0.1219, cr_loss=0.3399, over 17325.00 frames. ], tot_loss[loss=0.2045, ctc_loss=0.1338, cr_loss=0.3535, over 2399927.57 frames. ], batch size: 46, lr: 4.15e-03, grad_scale: 32.0
2024-09-24 13:22:00,734 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.248e+02 1.376e+02 1.477e+02 2.189e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-24 13:22:04,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=510248.6666666667, ans=0.025
2024-09-24 13:22:05,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=510295.3333333333, ans=0.125
2024-09-24 13:22:24,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=510342.0, ans=0.125
2024-09-24 13:23:07,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=510435.3333333333, ans=0.035
2024-09-24 13:23:07,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=510435.3333333333, ans=0.125
2024-09-24 13:23:16,526 INFO [train.py:1198] (1/4) Epoch 29, batch 300, loss[loss=0.2323, ctc_loss=0.1581, cr_loss=0.371, over 11965.00 frames. ], tot_loss[loss=0.2028, ctc_loss=0.1326, cr_loss=0.3511, over 2609390.74 frames. ], batch size: 123, lr: 4.14e-03, grad_scale: 16.0
2024-09-24 13:23:19,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=510482.0, ans=0.1
2024-09-24 13:23:23,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=510482.0, ans=0.05
2024-09-24 13:23:23,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=510482.0, ans=0.125
2024-09-24 13:23:24,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=510482.0, ans=0.04949747468305833
2024-09-24 13:23:42,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=510528.6666666667, ans=0.0
2024-09-24 13:24:03,205 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 13:24:09,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=510622.0, ans=0.125
2024-09-24 13:24:18,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=510622.0, ans=0.0
2024-09-24 13:24:20,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=510622.0, ans=0.2
2024-09-24 13:24:24,455 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.68 vs. limit=12.0
2024-09-24 13:24:25,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=510668.6666666667, ans=0.0
2024-09-24 13:24:31,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=510668.6666666667, ans=0.1
2024-09-24 13:24:39,066 INFO [train.py:1198] (1/4) Epoch 29, batch 350, loss[loss=0.22, ctc_loss=0.1441, cr_loss=0.3794, over 17205.00 frames. ], tot_loss[loss=0.202, ctc_loss=0.132, cr_loss=0.3497, over 2777133.15 frames. ], batch size: 47, lr: 4.14e-03, grad_scale: 16.0
2024-09-24 13:24:50,233 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.267e+02 1.359e+02 1.490e+02 2.133e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-24 13:25:03,735 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.03 vs. limit=15.0
2024-09-24 13:25:20,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=510808.6666666667, ans=0.0
2024-09-24 13:25:37,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=510855.3333333333, ans=0.0
2024-09-24 13:25:45,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=510902.0, ans=0.0
2024-09-24 13:26:01,404 INFO [train.py:1198] (1/4) Epoch 29, batch 400, loss[loss=0.1669, ctc_loss=0.1073, cr_loss=0.2977, over 17091.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1322, cr_loss=0.3497, over 2907169.58 frames. ], batch size: 43, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:26:22,751 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.41 vs. limit=15.0
2024-09-24 13:26:35,759 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.93 vs. limit=15.0
2024-09-24 13:26:36,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=511042.0, ans=0.0
2024-09-24 13:26:38,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=511042.0, ans=0.125
2024-09-24 13:27:21,139 INFO [train.py:1198] (1/4) Epoch 29, batch 450, loss[loss=0.1979, ctc_loss=0.1284, cr_loss=0.3474, over 17228.00 frames. ], tot_loss[loss=0.2014, ctc_loss=0.1317, cr_loss=0.3482, over 2999936.48 frames. ], batch size: 47, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:27:21,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=511182.0, ans=0.0
2024-09-24 13:27:35,217 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.269e+02 1.351e+02 1.464e+02 1.902e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-24 13:27:48,826 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.11 vs. limit=15.0
2024-09-24 13:27:52,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=511228.6666666667, ans=0.0
2024-09-24 13:28:27,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=511322.0, ans=0.125
2024-09-24 13:28:46,768 INFO [train.py:1198] (1/4) Epoch 29, batch 500, loss[loss=0.1618, ctc_loss=0.1019, cr_loss=0.2995, over 17062.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1319, cr_loss=0.349, over 3088188.49 frames. ], batch size: 39, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:28:54,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=511415.3333333333, ans=0.0
2024-09-24 13:28:55,275 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.57 vs. limit=15.0
2024-09-24 13:28:57,381 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.53 vs. limit=15.0
2024-09-24 13:29:15,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=511462.0, ans=0.125
2024-09-24 13:29:26,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=511508.6666666667, ans=0.0
2024-09-24 13:29:58,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=511602.0, ans=0.07
2024-09-24 13:30:09,200 INFO [train.py:1198] (1/4) Epoch 29, batch 550, loss[loss=0.1825, ctc_loss=0.117, cr_loss=0.3273, over 17288.00 frames. ], tot_loss[loss=0.202, ctc_loss=0.1322, cr_loss=0.3494, over 3139651.08 frames. ], batch size: 46, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:30:20,288 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.249e+02 1.312e+02 1.438e+02 1.848e+02, threshold=2.623e+02, percent-clipped=0.0
2024-09-24 13:30:20,630 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=511648.6666666667, ans=0.2
2024-09-24 13:30:28,532 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.91 vs. limit=22.5
2024-09-24 13:30:30,282 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=17.73 vs. limit=22.5
2024-09-24 13:31:30,726 INFO [train.py:1198] (1/4) Epoch 29, batch 600, loss[loss=0.1612, ctc_loss=0.101, cr_loss=0.3013, over 17189.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1326, cr_loss=0.3499, over 3180901.24 frames. ], batch size: 41, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:31:44,745 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.00 vs. limit=10.0
2024-09-24 13:32:07,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=511975.3333333333, ans=0.125
2024-09-24 13:32:12,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=511975.3333333333, ans=0.2
2024-09-24 13:32:36,011 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 13:32:36,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=512068.6666666667, ans=0.125
2024-09-24 13:32:43,104 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.97 vs. limit=12.0
2024-09-24 13:32:53,242 INFO [train.py:1198] (1/4) Epoch 29, batch 650, loss[loss=0.1834, ctc_loss=0.1147, cr_loss=0.3434, over 17051.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1325, cr_loss=0.3486, over 3193448.42 frames. ], batch size: 39, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:32:57,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.90 vs. limit=22.5
2024-09-24 13:33:04,448 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.266e+02 1.354e+02 1.448e+02 2.374e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-24 13:33:04,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=512115.3333333333, ans=0.2
2024-09-24 13:33:09,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=512162.0, ans=0.0
2024-09-24 13:33:16,040 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.55 vs. limit=15.0
2024-09-24 13:33:54,888 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.87 vs. limit=15.0
2024-09-24 13:34:19,040 INFO [train.py:1198] (1/4) Epoch 29, batch 700, loss[loss=0.2182, ctc_loss=0.1455, cr_loss=0.3634, over 16519.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1316, cr_loss=0.3472, over 3236062.67 frames. ], batch size: 66, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:34:22,594 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 13:34:40,868 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.97 vs. limit=22.5
2024-09-24 13:35:41,601 INFO [train.py:1198] (1/4) Epoch 29, batch 750, loss[loss=0.1918, ctc_loss=0.1254, cr_loss=0.3318, over 17281.00 frames. ], tot_loss[loss=0.2008, ctc_loss=0.1314, cr_loss=0.3468, over 3251249.27 frames. ], batch size: 42, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:35:52,784 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.251e+02 1.333e+02 1.428e+02 1.733e+02, threshold=2.666e+02, percent-clipped=0.0
2024-09-24 13:36:00,156 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.81 vs. limit=6.0
2024-09-24 13:36:04,523 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.81 vs. limit=15.0
2024-09-24 13:36:21,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=512675.3333333333, ans=0.125
2024-09-24 13:36:31,784 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.71 vs. limit=15.0
2024-09-24 13:37:01,404 INFO [train.py:1198] (1/4) Epoch 29, batch 800, loss[loss=0.2339, ctc_loss=0.1574, cr_loss=0.3827, over 15144.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.132, cr_loss=0.3487, over 3276060.75 frames. ], batch size: 89, lr: 4.14e-03, grad_scale: 32.0
2024-09-24 13:37:01,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=512815.3333333333, ans=0.0
2024-09-24 13:37:03,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=512815.3333333333, ans=0.125
2024-09-24 13:37:27,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=512862.0, ans=0.2
2024-09-24 13:37:48,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=512908.6666666667, ans=0.0
2024-09-24 13:37:54,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=512955.3333333333, ans=0.125
2024-09-24 13:37:54,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=512955.3333333333, ans=0.0
2024-09-24 13:38:10,512 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=6.85 vs. limit=12.0
2024-09-24 13:38:15,028 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.73 vs. limit=22.5
2024-09-24 13:38:19,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=513002.0, ans=0.125
2024-09-24 13:38:24,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=513002.0, ans=0.0
2024-09-24 13:38:27,133 INFO [train.py:1198] (1/4) Epoch 29, batch 850, loss[loss=0.2115, ctc_loss=0.1413, cr_loss=0.3513, over 17125.00 frames. ], tot_loss[loss=0.2023, ctc_loss=0.1325, cr_loss=0.3489, over 3281287.57 frames. ], batch size: 48, lr: 4.13e-03, grad_scale: 32.0
2024-09-24 13:38:38,336 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.286e+02 1.364e+02 1.497e+02 3.898e+02, threshold=2.729e+02, percent-clipped=1.0
2024-09-24 13:38:45,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=513095.3333333333, ans=0.125
2024-09-24 13:38:58,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=513095.3333333333, ans=0.125
2024-09-24 13:39:00,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=513142.0, ans=0.0
2024-09-24 13:39:03,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=513142.0, ans=0.2
2024-09-24 13:39:05,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=513142.0, ans=0.0
2024-09-24 13:39:06,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=513142.0, ans=0.0
2024-09-24 13:39:17,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=513188.6666666667, ans=0.025
2024-09-24 13:39:49,251 INFO [train.py:1198] (1/4) Epoch 29, batch 900, loss[loss=0.1936, ctc_loss=0.1231, cr_loss=0.353, over 17192.00 frames. ], tot_loss[loss=0.2023, ctc_loss=0.1325, cr_loss=0.349, over 3290402.37 frames. ], batch size: 41, lr: 4.13e-03, grad_scale: 32.0
2024-09-24 13:39:49,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=513282.0, ans=0.0
2024-09-24 13:39:51,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=513282.0, ans=0.0
2024-09-24 13:40:31,351 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.78 vs. limit=22.5
2024-09-24 13:40:32,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=513375.3333333333, ans=0.125
2024-09-24 13:40:33,046 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.73 vs. limit=15.0
2024-09-24 13:40:39,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=513422.0, ans=0.07
2024-09-24 13:41:11,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=513515.3333333333, ans=0.125
2024-09-24 13:41:12,250 INFO [train.py:1198] (1/4) Epoch 29, batch 950, loss[loss=0.2457, ctc_loss=0.1676, cr_loss=0.3908, over 11562.00 frames. ], tot_loss[loss=0.2028, ctc_loss=0.1328, cr_loss=0.3502, over 3300793.95 frames. ], batch size: 124, lr: 4.13e-03, grad_scale: 32.0
2024-09-24 13:41:23,505 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.274e+02 1.388e+02 1.487e+02 2.628e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-24 13:41:41,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=513562.0, ans=0.0
2024-09-24 13:41:46,542 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.02 vs. limit=15.0
2024-09-24 13:42:35,065 INFO [train.py:1198] (1/4) Epoch 29, batch 1000, loss[loss=0.2086, ctc_loss=0.1365, cr_loss=0.3604, over 17367.00 frames. ], tot_loss[loss=0.203, ctc_loss=0.1329, cr_loss=0.3505, over 3314861.14 frames. ], batch size: 48, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:42:41,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=513748.6666666667, ans=0.05
2024-09-24 13:43:02,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=513795.3333333333, ans=0.125
2024-09-24 13:43:14,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=513842.0, ans=0.1
2024-09-24 13:43:35,556 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.61 vs. limit=15.0
2024-09-24 13:43:59,529 INFO [train.py:1198] (1/4) Epoch 29, batch 1050, loss[loss=0.1682, ctc_loss=0.1077, cr_loss=0.3024, over 16997.00 frames. ], tot_loss[loss=0.2031, ctc_loss=0.133, cr_loss=0.3505, over 3319585.87 frames. ], batch size: 39, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:44:12,020 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.285e+02 1.349e+02 1.436e+02 3.120e+02, threshold=2.698e+02, percent-clipped=1.0
2024-09-24 13:44:36,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.90 vs. limit=15.0
2024-09-24 13:44:55,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=514122.0, ans=0.0
2024-09-24 13:45:12,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=514168.6666666667, ans=0.0
2024-09-24 13:45:15,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=514168.6666666667, ans=0.125
2024-09-24 13:45:21,711 INFO [train.py:1198] (1/4) Epoch 29, batch 1100, loss[loss=0.1695, ctc_loss=0.1086, cr_loss=0.3044, over 16639.00 frames. ], tot_loss[loss=0.2029, ctc_loss=0.1328, cr_loss=0.3507, over 3333837.74 frames. ], batch size: 37, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:45:34,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=514215.3333333333, ans=0.125
2024-09-24 13:45:36,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=514262.0, ans=0.1
2024-09-24 13:45:43,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=514262.0, ans=0.0
2024-09-24 13:46:05,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=514308.6666666667, ans=0.125
2024-09-24 13:46:18,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=514355.3333333333, ans=0.125
2024-09-24 13:46:18,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=514355.3333333333, ans=0.0
2024-09-24 13:46:21,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=514355.3333333333, ans=0.0
2024-09-24 13:46:28,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=514402.0, ans=0.125
2024-09-24 13:46:42,099 INFO [train.py:1198] (1/4) Epoch 29, batch 1150, loss[loss=0.2029, ctc_loss=0.1329, cr_loss=0.3504, over 16958.00 frames. ], tot_loss[loss=0.2027, ctc_loss=0.1325, cr_loss=0.3506, over 3339382.92 frames. ], batch size: 58, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:46:49,240 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.72 vs. limit=12.0
2024-09-24 13:46:50,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=514448.6666666667, ans=0.125
2024-09-24 13:46:54,890 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.257e+02 1.322e+02 1.401e+02 2.069e+02, threshold=2.644e+02, percent-clipped=0.0
2024-09-24 13:46:55,323 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 13:47:04,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=514495.3333333333, ans=0.125
2024-09-24 13:47:04,941 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.75 vs. limit=15.0
2024-09-24 13:47:35,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=514588.6666666667, ans=0.125
2024-09-24 13:47:46,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=514635.3333333333, ans=0.125
2024-09-24 13:47:54,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=514635.3333333333, ans=0.1
2024-09-24 13:48:04,033 INFO [train.py:1198] (1/4) Epoch 29, batch 1200, loss[loss=0.1916, ctc_loss=0.1255, cr_loss=0.3304, over 17288.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1323, cr_loss=0.3498, over 3341787.19 frames. ], batch size: 49, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:48:30,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=514728.6666666667, ans=0.0
2024-09-24 13:49:13,066 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.58 vs. limit=15.0
2024-09-24 13:49:16,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=514868.6666666667, ans=0.0
2024-09-24 13:49:27,326 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 13:49:28,540 INFO [train.py:1198] (1/4) Epoch 29, batch 1250, loss[loss=0.2179, ctc_loss=0.1451, cr_loss=0.3643, over 17154.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1313, cr_loss=0.3484, over 3352988.85 frames. ], batch size: 48, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:49:42,755 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.263e+02 1.323e+02 1.423e+02 2.408e+02, threshold=2.646e+02, percent-clipped=0.0
2024-09-24 13:49:54,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=514962.0, ans=0.1
2024-09-24 13:50:00,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=515008.6666666667, ans=0.0
2024-09-24 13:50:04,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=515008.6666666667, ans=0.1
2024-09-24 13:50:05,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=515008.6666666667, ans=0.1
2024-09-24 13:50:20,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=515055.3333333333, ans=0.1
2024-09-24 13:50:29,335 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.38 vs. limit=15.0
2024-09-24 13:50:50,682 INFO [train.py:1198] (1/4) Epoch 29, batch 1300, loss[loss=0.1914, ctc_loss=0.1209, cr_loss=0.3523, over 17080.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1306, cr_loss=0.3475, over 3356997.75 frames. ], batch size: 43, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:50:57,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer_ff3.min_abs, batch_count=515148.6666666667, ans=0.2
2024-09-24 13:51:10,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=515195.3333333333, ans=0.0
2024-09-24 13:51:13,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=515195.3333333333, ans=0.0
2024-09-24 13:51:20,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=515195.3333333333, ans=0.125
2024-09-24 13:51:21,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=515242.0, ans=0.05
2024-09-24 13:52:10,986 INFO [train.py:1198] (1/4) Epoch 29, batch 1350, loss[loss=0.1844, ctc_loss=0.1211, cr_loss=0.3165, over 17044.00 frames. ], tot_loss[loss=0.1998, ctc_loss=0.1304, cr_loss=0.347, over 3364872.89 frames. ], batch size: 52, lr: 4.13e-03, grad_scale: 16.0
2024-09-24 13:52:14,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=515382.0, ans=0.05
2024-09-24 13:52:25,431 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.241e+02 1.323e+02 1.445e+02 2.039e+02, threshold=2.645e+02, percent-clipped=0.0
2024-09-24 13:53:05,722 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.76 vs. limit=12.0
2024-09-24 13:53:15,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=515522.0, ans=0.0
2024-09-24 13:53:35,818 INFO [train.py:1198] (1/4) Epoch 29, batch 1400, loss[loss=0.2216, ctc_loss=0.1464, cr_loss=0.3759, over 17303.00 frames. ], tot_loss[loss=0.2011, ctc_loss=0.1314, cr_loss=0.3483, over 3361687.51 frames. ], batch size: 49, lr: 4.12e-03, grad_scale: 16.0
2024-09-24 13:53:47,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=515615.3333333333, ans=0.1
2024-09-24 13:54:38,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=515755.3333333333, ans=0.1
2024-09-24 13:54:42,578 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.93 vs. limit=15.0
2024-09-24 13:54:57,872 INFO [train.py:1198] (1/4) Epoch 29, batch 1450, loss[loss=0.184, ctc_loss=0.1198, cr_loss=0.321, over 16779.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1319, cr_loss=0.3487, over 3354870.03 frames. ], batch size: 37, lr: 4.12e-03, grad_scale: 8.0
2024-09-24 13:55:16,336 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.288e+02 1.365e+02 1.470e+02 2.158e+02, threshold=2.730e+02, percent-clipped=0.0
2024-09-24 13:55:23,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=515895.3333333333, ans=0.0
2024-09-24 13:55:42,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=515942.0, ans=0.125
2024-09-24 13:55:43,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=515942.0, ans=0.2
2024-09-24 13:55:51,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=515988.6666666667, ans=0.0
2024-09-24 13:55:56,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=515988.6666666667, ans=0.0
2024-09-24 13:55:59,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=515988.6666666667, ans=0.2
2024-09-24 13:56:19,905 INFO [train.py:1198] (1/4) Epoch 29, batch 1500, loss[loss=0.2118, ctc_loss=0.1386, cr_loss=0.366, over 17308.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1311, cr_loss=0.3479, over 3367548.89 frames. ], batch size: 51, lr: 4.12e-03, grad_scale: 8.0
2024-09-24 13:56:34,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=516128.6666666667, ans=0.125
2024-09-24 13:56:56,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=516175.3333333333, ans=0.125
2024-09-24 13:57:06,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=516222.0, ans=0.125
2024-09-24 13:57:16,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=516222.0, ans=0.125
2024-09-24 13:57:21,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=516222.0, ans=0.1
2024-09-24 13:57:42,716 INFO [train.py:1198] (1/4) Epoch 29, batch 1550, loss[loss=0.2004, ctc_loss=0.1311, cr_loss=0.3461, over 16882.00 frames. ], tot_loss[loss=0.2016, ctc_loss=0.1317, cr_loss=0.3495, over 3363888.12 frames. ], batch size: 58, lr: 4.12e-03, grad_scale: 8.0
2024-09-24 13:57:54,807 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.53 vs. limit=10.0
2024-09-24 13:57:58,708 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.272e+02 1.340e+02 1.430e+02 4.940e+02, threshold=2.681e+02, percent-clipped=1.0
2024-09-24 13:58:03,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=516362.0, ans=0.1
2024-09-24 13:58:11,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=516362.0, ans=0.125
2024-09-24 13:58:41,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=516455.3333333333, ans=0.125
2024-09-24 13:59:07,816 INFO [train.py:1198] (1/4) Epoch 29, batch 1600, loss[loss=0.2046, ctc_loss=0.1359, cr_loss=0.3435, over 17063.00 frames. ], tot_loss[loss=0.2027, ctc_loss=0.1326, cr_loss=0.3505, over 3362406.54 frames. ], batch size: 46, lr: 4.12e-03, grad_scale: 16.0
2024-09-24 13:59:52,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=516642.0, ans=0.125
2024-09-24 13:59:58,056 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.61 vs. limit=10.0
2024-09-24 14:00:00,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=516688.6666666667, ans=0.1
2024-09-24 14:00:30,027 INFO [train.py:1198] (1/4) Epoch 29, batch 1650, loss[loss=0.1932, ctc_loss=0.126, cr_loss=0.3358, over 17012.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1324, cr_loss=0.3509, over 3365102.06 frames. ], batch size: 44, lr: 4.12e-03, grad_scale: 16.0
2024-09-24 14:00:45,987 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.269e+02 1.357e+02 1.497e+02 2.178e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-24 14:01:03,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=516875.3333333333, ans=0.0
2024-09-24 14:01:19,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=516922.0, ans=0.125
2024-09-24 14:01:19,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=516922.0, ans=0.2
2024-09-24 14:01:26,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=516922.0, ans=0.125
2024-09-24 14:01:35,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=516968.6666666667, ans=0.125
2024-09-24 14:01:49,853 INFO [train.py:1198] (1/4) Epoch 29, batch 1700, loss[loss=0.233, ctc_loss=0.155, cr_loss=0.3903, over 16874.00 frames. ], tot_loss[loss=0.2039, ctc_loss=0.1334, cr_loss=0.3527, over 3356394.78 frames. ], batch size: 58, lr: 4.12e-03, grad_scale: 16.0
2024-09-24 14:01:58,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=517015.3333333333, ans=0.1
2024-09-24 14:02:20,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=517108.6666666667, ans=0.125
2024-09-24 14:02:21,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=517108.6666666667, ans=0.0
2024-09-24 14:02:41,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=517155.3333333333, ans=0.0
2024-09-24 14:02:48,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=517155.3333333333, ans=0.1
2024-09-24 14:03:14,198 INFO [train.py:1198] (1/4) Epoch 29, batch 1750, loss[loss=0.1973, ctc_loss=0.1279, cr_loss=0.3472, over 17337.00 frames. ], tot_loss[loss=0.204, ctc_loss=0.1335, cr_loss=0.3523, over 3354870.21 frames. ], batch size: 52, lr: 4.12e-03, grad_scale: 16.0
2024-09-24 14:03:16,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=517248.6666666667, ans=0.125
2024-09-24 14:03:30,182 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.263e+02 1.353e+02 1.492e+02 2.426e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-24 14:04:19,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=517435.3333333333, ans=0.125
2024-09-24 14:04:22,296 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=517435.3333333333, ans=0.125
2024-09-24 14:04:34,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=517482.0, ans=0.125
2024-09-24 14:04:36,183 INFO [train.py:1198] (1/4) Epoch 29, batch 1800, loss[loss=0.2327, ctc_loss=0.1564, cr_loss=0.3813, over 15937.00 frames. ], tot_loss[loss=0.2029, ctc_loss=0.1328, cr_loss=0.3508, over 3348722.77 frames. ], batch size: 74, lr: 4.12e-03, grad_scale: 16.0
2024-09-24 14:04:42,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=517482.0, ans=0.125
2024-09-24 14:05:54,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=517668.6666666667, ans=0.2
2024-09-24 14:05:55,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=517668.6666666667, ans=0.125
2024-09-24 14:05:58,951 INFO [train.py:1198] (1/4) Epoch 29, batch 1850, loss[loss=0.2091, ctc_loss=0.1362, cr_loss=0.3645, over 17350.00 frames. ], tot_loss[loss=0.2014, ctc_loss=0.1316, cr_loss=0.3488, over 3363417.53 frames. ], batch size: 48, lr: 4.12e-03, grad_scale: 8.0
2024-09-24 14:06:16,452 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.270e+02 1.382e+02 1.482e+02 2.420e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-24 14:06:23,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.40 vs. limit=22.5
2024-09-24 14:06:48,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=517855.3333333333, ans=0.125
2024-09-24 14:06:50,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=517855.3333333333, ans=0.125
2024-09-24 14:06:51,086 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.91 vs. limit=15.0
2024-09-24 14:06:57,682 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.56 vs. limit=15.0
2024-09-24 14:07:00,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=517855.3333333333, ans=0.125
2024-09-24 14:07:00,517 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.51 vs. limit=15.0
2024-09-24 14:07:00,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.50 vs. limit=15.0
2024-09-24 14:07:06,740 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.86 vs. limit=15.0
2024-09-24 14:07:21,463 INFO [train.py:1198] (1/4) Epoch 29, batch 1900, loss[loss=0.1873, ctc_loss=0.1212, cr_loss=0.3304, over 17087.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1319, cr_loss=0.349, over 3356247.65 frames. ], batch size: 43, lr: 4.12e-03, grad_scale: 8.0
2024-09-24 14:07:34,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=517948.6666666667, ans=0.125
2024-09-24 14:07:46,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=517995.3333333333, ans=0.125
2024-09-24 14:08:05,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=518042.0, ans=0.125
2024-09-24 14:08:21,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=518088.6666666667, ans=0.1
2024-09-24 14:08:35,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=518135.3333333333, ans=0.1
2024-09-24 14:08:43,830 INFO [train.py:1198] (1/4) Epoch 29, batch 1950, loss[loss=0.2238, ctc_loss=0.1505, cr_loss=0.3665, over 16910.00 frames. ], tot_loss[loss=0.2027, ctc_loss=0.1327, cr_loss=0.3501, over 3352096.94 frames. ], batch size: 58, lr: 4.11e-03, grad_scale: 8.0
2024-09-24 14:09:03,911 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.275e+02 1.365e+02 1.439e+02 2.080e+02, threshold=2.731e+02, percent-clipped=0.0
2024-09-24 14:09:09,542 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.85 vs. limit=6.0
2024-09-24 14:09:13,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=518228.6666666667, ans=0.125
2024-09-24 14:09:21,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=518275.3333333333, ans=0.025
2024-09-24 14:10:08,969 INFO [train.py:1198] (1/4) Epoch 29, batch 2000, loss[loss=0.2105, ctc_loss=0.1355, cr_loss=0.3747, over 16321.00 frames. ], tot_loss[loss=0.2029, ctc_loss=0.1328, cr_loss=0.3504, over 3359565.67 frames. ], batch size: 36, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:10:12,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=518415.3333333333, ans=0.125
2024-09-24 14:10:48,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=8.66 vs. limit=15.0
2024-09-24 14:11:15,431 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.27 vs. limit=22.5
2024-09-24 14:11:26,417 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=14.51 vs. limit=15.0
2024-09-24 14:11:28,975 INFO [train.py:1198] (1/4) Epoch 29, batch 2050, loss[loss=0.2263, ctc_loss=0.1525, cr_loss=0.3691, over 17312.00 frames. ], tot_loss[loss=0.2023, ctc_loss=0.1324, cr_loss=0.3495, over 3358910.43 frames. ], batch size: 49, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:11:29,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.03 vs. limit=22.5
2024-09-24 14:11:46,520 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.263e+02 1.339e+02 1.463e+02 3.835e+02, threshold=2.678e+02, percent-clipped=1.0
2024-09-24 14:11:59,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=518742.0, ans=0.125
2024-09-24 14:12:08,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=518742.0, ans=0.0
2024-09-24 14:12:36,083 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.51 vs. limit=15.0
2024-09-24 14:12:54,001 INFO [train.py:1198] (1/4) Epoch 29, batch 2100, loss[loss=0.1682, ctc_loss=0.1084, cr_loss=0.2991, over 15860.00 frames. ], tot_loss[loss=0.2016, ctc_loss=0.1319, cr_loss=0.3485, over 3359002.17 frames. ], batch size: 35, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:13:35,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=518975.3333333333, ans=0.1
2024-09-24 14:13:40,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=519022.0, ans=0.125
2024-09-24 14:13:54,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.21 vs. limit=15.0
2024-09-24 14:14:00,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=519068.6666666667, ans=0.125
2024-09-24 14:14:15,827 INFO [train.py:1198] (1/4) Epoch 29, batch 2150, loss[loss=0.2302, ctc_loss=0.1549, cr_loss=0.3763, over 17175.00 frames. ], tot_loss[loss=0.2018, ctc_loss=0.1321, cr_loss=0.3488, over 3359707.10 frames. ], batch size: 55, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:14:33,680 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.281e+02 1.376e+02 1.508e+02 1.841e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-24 14:14:35,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=519162.0, ans=0.0
2024-09-24 14:14:43,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=519162.0, ans=0.2
2024-09-24 14:15:19,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=519255.3333333333, ans=0.125
2024-09-24 14:15:29,603 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 14:15:38,898 INFO [train.py:1198] (1/4) Epoch 29, batch 2200, loss[loss=0.2111, ctc_loss=0.1394, cr_loss=0.3585, over 16889.00 frames. ], tot_loss[loss=0.2016, ctc_loss=0.1318, cr_loss=0.3487, over 3364739.59 frames. ], batch size: 58, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:15:45,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=519348.6666666667, ans=0.1
2024-09-24 14:15:52,376 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.81 vs. limit=22.5
2024-09-24 14:16:03,611 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.49 vs. limit=15.0
2024-09-24 14:16:15,973 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 14:16:29,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=519488.6666666667, ans=0.125
2024-09-24 14:16:31,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn2.whiten.whitening_limit, batch_count=519488.6666666667, ans=22.5
2024-09-24 14:16:54,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.max_abs, batch_count=519535.3333333333, ans=10.0
2024-09-24 14:16:59,445 INFO [train.py:1198] (1/4) Epoch 29, batch 2250, loss[loss=0.1932, ctc_loss=0.1247, cr_loss=0.3423, over 17189.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1307, cr_loss=0.3466, over 3368091.40 frames. ], batch size: 45, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:17:19,626 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.297e+02 1.378e+02 1.443e+02 1.753e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-24 14:17:27,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=519628.6666666667, ans=0.5
2024-09-24 14:17:42,627 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.70 vs. limit=15.0
2024-09-24 14:17:55,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=519722.0, ans=0.0
2024-09-24 14:18:08,714 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.13 vs. limit=6.0
2024-09-24 14:18:09,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=519768.6666666667, ans=0.125
2024-09-24 14:18:23,697 INFO [train.py:1198] (1/4) Epoch 29, batch 2300, loss[loss=0.1894, ctc_loss=0.1231, cr_loss=0.3317, over 17288.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1314, cr_loss=0.3478, over 3367350.25 frames. ], batch size: 42, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:19:09,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=519908.6666666667, ans=0.125
2024-09-24 14:19:27,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=519955.3333333333, ans=0.0
2024-09-24 14:19:36,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=520002.0, ans=0.0
2024-09-24 14:19:38,817 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.46 vs. limit=10.0
2024-09-24 14:19:46,204 INFO [train.py:1198] (1/4) Epoch 29, batch 2350, loss[loss=0.1707, ctc_loss=0.1113, cr_loss=0.297, over 16243.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.1303, cr_loss=0.3457, over 3372674.13 frames. ], batch size: 36, lr: 4.11e-03, grad_scale: 16.0
2024-09-24 14:20:01,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=520048.6666666667, ans=0.125
2024-09-24 14:20:06,252 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.272e+02 1.348e+02 1.492e+02 2.219e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-24 14:20:14,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=520095.3333333333, ans=0.125
2024-09-24 14:20:18,211 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.45 vs. limit=22.5
2024-09-24 14:20:28,017 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.01 vs. limit=15.0
2024-09-24 14:20:46,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=520188.6666666667, ans=0.125
2024-09-24 14:20:51,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=520235.3333333333, ans=0.0
2024-09-24 14:21:08,865 INFO [train.py:1198] (1/4) Epoch 29, batch 2400, loss[loss=0.1627, ctc_loss=0.1041, cr_loss=0.2932, over 17113.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.13, cr_loss=0.3455, over 3368555.51 frames. ], batch size: 40, lr: 4.11e-03, grad_scale: 32.0
2024-09-24 14:21:23,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=520328.6666666667, ans=0.0
2024-09-24 14:21:29,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=520328.6666666667, ans=0.0
2024-09-24 14:21:34,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=520328.6666666667, ans=0.0
2024-09-24 14:21:38,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=520328.6666666667, ans=0.0
2024-09-24 14:22:08,918 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.76 vs. limit=15.0
2024-09-24 14:22:19,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=7.21 vs. limit=15.0
2024-09-24 14:22:27,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=520468.6666666667, ans=0.125
2024-09-24 14:22:31,467 INFO [train.py:1198] (1/4) Epoch 29, batch 2450, loss[loss=0.234, ctc_loss=0.1548, cr_loss=0.3959, over 17019.00 frames. ], tot_loss[loss=0.1986, ctc_loss=0.1295, cr_loss=0.3451, over 3371215.45 frames. ], batch size: 53, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:22:53,218 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.297e+02 1.387e+02 1.500e+02 2.305e+02, threshold=2.773e+02, percent-clipped=0.0
2024-09-24 14:22:58,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=520562.0, ans=0.125
2024-09-24 14:23:10,392 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.66 vs. limit=22.5
2024-09-24 14:23:15,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=520608.6666666667, ans=0.125
2024-09-24 14:23:56,403 INFO [train.py:1198] (1/4) Epoch 29, batch 2500, loss[loss=0.1831, ctc_loss=0.1225, cr_loss=0.3032, over 17304.00 frames. ], tot_loss[loss=0.1985, ctc_loss=0.1296, cr_loss=0.3446, over 3367690.22 frames. ], batch size: 46, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:23:59,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=520748.6666666667, ans=0.2
2024-09-24 14:24:03,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=520748.6666666667, ans=0.025
2024-09-24 14:24:53,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=520888.6666666667, ans=0.125
2024-09-24 14:24:58,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=520888.6666666667, ans=0.125
2024-09-24 14:25:01,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=520935.3333333333, ans=0.125
2024-09-24 14:25:18,997 INFO [train.py:1198] (1/4) Epoch 29, batch 2550, loss[loss=0.2251, ctc_loss=0.1482, cr_loss=0.3846, over 17102.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1298, cr_loss=0.3446, over 3354313.76 frames. ], batch size: 49, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:25:19,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=520982.0, ans=0.0
2024-09-24 14:25:33,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=521028.6666666667, ans=0.0
2024-09-24 14:25:38,169 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.280e+02 1.353e+02 1.438e+02 1.912e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-24 14:25:38,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=521028.6666666667, ans=0.125
2024-09-24 14:25:54,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=521075.3333333333, ans=0.1
2024-09-24 14:26:07,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.11 vs. limit=15.0
2024-09-24 14:26:12,333 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.54 vs. limit=15.0
2024-09-24 14:26:38,263 INFO [train.py:1198] (1/4) Epoch 29, batch 2600, loss[loss=0.224, ctc_loss=0.1506, cr_loss=0.3669, over 15104.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.1303, cr_loss=0.3458, over 3354436.71 frames. ], batch size: 89, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:26:38,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=521215.3333333333, ans=0.125
2024-09-24 14:26:43,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=521215.3333333333, ans=0.125
2024-09-24 14:27:06,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=521262.0, ans=0.1
2024-09-24 14:28:00,788 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.81 vs. limit=10.0
2024-09-24 14:28:03,407 INFO [train.py:1198] (1/4) Epoch 29, batch 2650, loss[loss=0.2036, ctc_loss=0.1311, cr_loss=0.3628, over 17303.00 frames. ], tot_loss[loss=0.1998, ctc_loss=0.1306, cr_loss=0.3459, over 3336741.71 frames. ], batch size: 49, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:28:05,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=521448.6666666667, ans=0.125
2024-09-24 14:28:13,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=521448.6666666667, ans=0.125
2024-09-24 14:28:13,784 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=15.99 vs. limit=22.5
2024-09-24 14:28:17,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.min_positive, batch_count=521495.3333333333, ans=0.05
2024-09-24 14:28:22,577 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.267e+02 1.389e+02 1.480e+02 2.068e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-24 14:28:35,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=521542.0, ans=0.1
2024-09-24 14:29:06,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=521588.6666666667, ans=0.125
2024-09-24 14:29:22,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=521635.3333333333, ans=0.0
2024-09-24 14:29:25,644 INFO [train.py:1198] (1/4) Epoch 29, batch 2700, loss[loss=0.2155, ctc_loss=0.1373, cr_loss=0.3906, over 17288.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.1312, cr_loss=0.347, over 3343581.40 frames. ], batch size: 51, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:30:20,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=521822.0, ans=0.125
2024-09-24 14:30:23,896 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.10 vs. limit=15.0
2024-09-24 14:30:23,917 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.36 vs. limit=15.0
2024-09-24 14:30:25,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=521822.0, ans=0.125
2024-09-24 14:30:34,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=521868.6666666667, ans=0.025
2024-09-24 14:30:46,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=521868.6666666667, ans=0.035
2024-09-24 14:30:46,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=521868.6666666667, ans=0.125
2024-09-24 14:30:49,227 INFO [train.py:1198] (1/4) Epoch 29, batch 2750, loss[loss=0.1855, ctc_loss=0.1188, cr_loss=0.3338, over 17108.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.1311, cr_loss=0.3473, over 3344599.82 frames. ], batch size: 49, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:30:54,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=521915.3333333333, ans=0.2
2024-09-24 14:31:08,250 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.228e+02 1.306e+02 1.411e+02 3.014e+02, threshold=2.612e+02, percent-clipped=1.0
2024-09-24 14:31:17,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=521962.0, ans=0.04949747468305833
2024-09-24 14:31:42,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=522055.3333333333, ans=0.125
2024-09-24 14:31:45,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=522055.3333333333, ans=0.1
2024-09-24 14:32:11,737 INFO [train.py:1198] (1/4) Epoch 29, batch 2800, loss[loss=0.1729, ctc_loss=0.1088, cr_loss=0.3205, over 17201.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1315, cr_loss=0.3473, over 3339617.70 frames. ], batch size: 41, lr: 4.10e-03, grad_scale: 32.0
2024-09-24 14:32:18,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=522148.6666666667, ans=0.025
2024-09-24 14:32:24,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=522148.6666666667, ans=0.0
2024-09-24 14:32:29,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=522195.3333333333, ans=0.0
2024-09-24 14:32:31,526 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.40 vs. limit=15.0
2024-09-24 14:32:41,184 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.04 vs. limit=10.0
2024-09-24 14:33:07,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=522288.6666666667, ans=0.0
2024-09-24 14:33:08,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=522288.6666666667, ans=0.125
2024-09-24 14:33:18,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=522335.3333333333, ans=0.0
2024-09-24 14:33:27,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.29 vs. limit=15.0
2024-09-24 14:33:28,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=522335.3333333333, ans=0.0
2024-09-24 14:33:34,159 INFO [train.py:1198] (1/4) Epoch 29, batch 2850, loss[loss=0.1627, ctc_loss=0.1026, cr_loss=0.3008, over 17296.00 frames. ], tot_loss[loss=0.2003, ctc_loss=0.1309, cr_loss=0.3469, over 3344619.05 frames. ], batch size: 49, lr: 4.10e-03, grad_scale: 32.0
2024-09-24 14:33:46,978 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=522382.0, ans=0.125
2024-09-24 14:33:56,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=522428.6666666667, ans=0.125
2024-09-24 14:33:57,876 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.283e+02 1.399e+02 1.566e+02 1.810e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-24 14:33:58,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=522428.6666666667, ans=0.0
2024-09-24 14:34:03,510 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.94 vs. limit=15.0
2024-09-24 14:35:00,062 INFO [train.py:1198] (1/4) Epoch 29, batch 2900, loss[loss=0.1882, ctc_loss=0.1243, cr_loss=0.3194, over 17001.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.131, cr_loss=0.348, over 3350644.51 frames. ], batch size: 44, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:35:14,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=522662.0, ans=0.125
2024-09-24 14:35:25,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=522662.0, ans=0.0
2024-09-24 14:35:43,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=522708.6666666667, ans=0.125
2024-09-24 14:36:22,783 INFO [train.py:1198] (1/4) Epoch 29, batch 2950, loss[loss=0.221, ctc_loss=0.1439, cr_loss=0.3856, over 17337.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.131, cr_loss=0.3479, over 3350584.71 frames. ], batch size: 48, lr: 4.10e-03, grad_scale: 16.0
2024-09-24 14:36:42,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=522895.3333333333, ans=0.125
2024-09-24 14:36:42,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=522895.3333333333, ans=0.0
2024-09-24 14:36:43,418 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.270e+02 1.367e+02 1.488e+02 1.985e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-24 14:37:37,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=523035.3333333333, ans=0.125
2024-09-24 14:37:44,729 INFO [train.py:1198] (1/4) Epoch 29, batch 3000, loss[loss=0.2092, ctc_loss=0.1358, cr_loss=0.3671, over 17298.00 frames. ], tot_loss[loss=0.1996, ctc_loss=0.1304, cr_loss=0.3463, over 3357538.86 frames. ], batch size: 46, lr: 4.09e-03, grad_scale: 16.0
2024-09-24 14:37:44,729 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 14:38:00,324 INFO [train.py:1230] (1/4) Epoch 29, validation: loss=0.03658, ctc_loss=0.03658, cr_loss=8.731e-15, over 944034.00 frames. 
2024-09-24 14:38:00,325 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 14:38:03,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=523082.0, ans=0.025
2024-09-24 14:38:46,855 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.32 vs. limit=12.0
2024-09-24 14:38:48,607 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.75 vs. limit=6.0
2024-09-24 14:38:52,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=523222.0, ans=0.125
2024-09-24 14:39:19,059 INFO [train.py:1198] (1/4) Epoch 29, batch 3050, loss[loss=0.2332, ctc_loss=0.1616, cr_loss=0.3579, over 12393.00 frames. ], tot_loss[loss=0.2013, ctc_loss=0.1316, cr_loss=0.3486, over 3362928.37 frames. ], batch size: 123, lr: 4.09e-03, grad_scale: 16.0
2024-09-24 14:39:42,288 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.250e+02 1.367e+02 1.506e+02 1.984e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-24 14:39:52,470 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.28 vs. limit=6.0
2024-09-24 14:40:01,907 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.99 vs. limit=22.5
2024-09-24 14:40:02,008 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.26 vs. limit=15.0
2024-09-24 14:40:12,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=523455.3333333333, ans=0.015
2024-09-24 14:40:17,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=523455.3333333333, ans=0.125
2024-09-24 14:40:17,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=523455.3333333333, ans=0.0
2024-09-24 14:40:28,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=523502.0, ans=0.0
2024-09-24 14:40:29,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=523502.0, ans=0.0
2024-09-24 14:40:30,069 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.55 vs. limit=15.0
2024-09-24 14:40:31,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=523502.0, ans=0.125
2024-09-24 14:40:40,602 INFO [train.py:1198] (1/4) Epoch 29, batch 3100, loss[loss=0.1849, ctc_loss=0.1214, cr_loss=0.3174, over 17247.00 frames. ], tot_loss[loss=0.2004, ctc_loss=0.131, cr_loss=0.3468, over 3350178.75 frames. ], batch size: 44, lr: 4.09e-03, grad_scale: 16.0
2024-09-24 14:41:11,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=523642.0, ans=0.2
2024-09-24 14:41:20,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=523642.0, ans=0.1
2024-09-24 14:41:37,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=523688.6666666667, ans=0.125
2024-09-24 14:41:50,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=523735.3333333333, ans=0.0
2024-09-24 14:41:57,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=523735.3333333333, ans=0.0
2024-09-24 14:42:01,583 INFO [train.py:1198] (1/4) Epoch 29, batch 3150, loss[loss=0.1908, ctc_loss=0.1233, cr_loss=0.3374, over 16876.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1299, cr_loss=0.3453, over 3359677.69 frames. ], batch size: 58, lr: 4.09e-03, grad_scale: 16.0
2024-09-24 14:42:21,628 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.246e+02 1.357e+02 1.527e+02 1.959e+02, threshold=2.714e+02, percent-clipped=0.0
2024-09-24 14:42:22,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.93 vs. limit=10.0
2024-09-24 14:42:40,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=523875.3333333333, ans=0.0
2024-09-24 14:42:45,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=523875.3333333333, ans=0.0
2024-09-24 14:42:57,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=523922.0, ans=0.1
2024-09-24 14:43:14,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=523968.6666666667, ans=0.1
2024-09-24 14:43:14,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.min_positive, batch_count=523968.6666666667, ans=0.025
2024-09-24 14:43:19,447 INFO [train.py:1198] (1/4) Epoch 29, batch 3200, loss[loss=0.1688, ctc_loss=0.1074, cr_loss=0.3067, over 16936.00 frames. ], tot_loss[loss=0.1981, ctc_loss=0.1294, cr_loss=0.3438, over 3347470.48 frames. ], batch size: 42, lr: 4.09e-03, grad_scale: 32.0
2024-09-24 14:43:52,771 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.87 vs. limit=10.0
2024-09-24 14:44:37,176 INFO [train.py:1198] (1/4) Epoch 29, batch 3250, loss[loss=0.2199, ctc_loss=0.1444, cr_loss=0.3778, over 16042.00 frames. ], tot_loss[loss=0.1986, ctc_loss=0.1298, cr_loss=0.3438, over 3340056.80 frames. ], batch size: 74, lr: 4.09e-03, grad_scale: 32.0
2024-09-24 14:44:51,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=524295.3333333334, ans=0.1
2024-09-24 14:44:56,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=524295.3333333334, ans=0.125
2024-09-24 14:44:57,321 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.269e+02 1.350e+02 1.434e+02 1.655e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-24 14:44:59,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=524295.3333333334, ans=0.025
2024-09-24 14:45:01,373 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.08 vs. limit=15.0
2024-09-24 14:45:02,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=524295.3333333334, ans=0.125
2024-09-24 14:45:03,904 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 14:45:35,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=524388.6666666666, ans=0.125
2024-09-24 14:45:47,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=524435.3333333334, ans=0.0
2024-09-24 14:45:51,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=524435.3333333334, ans=0.125
2024-09-24 14:45:52,994 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.58 vs. limit=22.5
2024-09-24 14:45:55,497 INFO [train.py:1198] (1/4) Epoch 29, batch 3300, loss[loss=0.1945, ctc_loss=0.124, cr_loss=0.3527, over 17295.00 frames. ], tot_loss[loss=0.2002, ctc_loss=0.131, cr_loss=0.3461, over 3332531.75 frames. ], batch size: 51, lr: 4.09e-03, grad_scale: 32.0
2024-09-24 14:46:09,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=524528.6666666666, ans=0.125
2024-09-24 14:46:32,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=524575.3333333334, ans=0.125
2024-09-24 14:46:37,528 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.71 vs. limit=6.0
2024-09-24 14:46:43,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=524622.0, ans=0.0
2024-09-24 14:46:51,333 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.64 vs. limit=15.0
2024-09-24 14:47:07,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=524668.6666666666, ans=0.1
2024-09-24 14:47:12,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=524668.6666666666, ans=0.1
2024-09-24 14:47:15,322 INFO [train.py:1198] (1/4) Epoch 29, batch 3350, loss[loss=0.2512, ctc_loss=0.169, cr_loss=0.4111, over 16055.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.1312, cr_loss=0.3471, over 3334188.27 frames. ], batch size: 74, lr: 4.09e-03, grad_scale: 32.0
2024-09-24 14:47:31,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=524762.0, ans=0.0
2024-09-24 14:47:35,614 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.319e+02 1.369e+02 1.450e+02 1.942e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-24 14:47:43,127 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.80 vs. limit=15.0
2024-09-24 14:48:10,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=524855.3333333334, ans=0.1
2024-09-24 14:48:35,633 INFO [train.py:1198] (1/4) Epoch 29, batch 3400, loss[loss=0.1683, ctc_loss=0.1077, cr_loss=0.3031, over 16958.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1299, cr_loss=0.3443, over 3334694.08 frames. ], batch size: 42, lr: 4.09e-03, grad_scale: 32.0
2024-09-24 14:48:46,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=524948.6666666666, ans=0.05
2024-09-24 14:48:50,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=524995.3333333334, ans=0.125
2024-09-24 14:48:54,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=524995.3333333334, ans=0.1
2024-09-24 14:48:56,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=524995.3333333334, ans=0.0
2024-09-24 14:48:56,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=524995.3333333334, ans=0.125
2024-09-24 14:48:59,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.18 vs. limit=10.0
2024-09-24 14:49:35,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=525088.6666666666, ans=0.0
2024-09-24 14:49:55,466 INFO [train.py:1198] (1/4) Epoch 29, batch 3450, loss[loss=0.2135, ctc_loss=0.143, cr_loss=0.3527, over 17211.00 frames. ], tot_loss[loss=0.1988, ctc_loss=0.1299, cr_loss=0.3447, over 3336093.86 frames. ], batch size: 55, lr: 4.09e-03, grad_scale: 32.0
2024-09-24 14:50:00,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=525182.0, ans=0.0
2024-09-24 14:50:10,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=525228.6666666666, ans=0.125
2024-09-24 14:50:15,930 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.258e+02 1.341e+02 1.460e+02 2.344e+02, threshold=2.681e+02, percent-clipped=0.0
2024-09-24 14:50:27,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=525275.3333333334, ans=0.09899494936611666
2024-09-24 14:50:44,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=525322.0, ans=0.0
2024-09-24 14:51:00,759 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.41 vs. limit=15.0
2024-09-24 14:51:05,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=525368.6666666666, ans=0.125
2024-09-24 14:51:09,950 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.76 vs. limit=15.0
2024-09-24 14:51:12,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=525415.3333333334, ans=0.125
2024-09-24 14:51:14,074 INFO [train.py:1198] (1/4) Epoch 29, batch 3500, loss[loss=0.1978, ctc_loss=0.1306, cr_loss=0.3359, over 17233.00 frames. ], tot_loss[loss=0.2008, ctc_loss=0.1314, cr_loss=0.3472, over 3330105.09 frames. ], batch size: 50, lr: 4.09e-03, grad_scale: 32.0
2024-09-24 14:51:58,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=525508.6666666666, ans=0.125
2024-09-24 14:52:03,092 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.83 vs. limit=12.0
2024-09-24 14:52:05,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=525555.3333333334, ans=0.1
2024-09-24 14:52:06,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=525555.3333333334, ans=0.125
2024-09-24 14:52:34,350 INFO [train.py:1198] (1/4) Epoch 29, batch 3550, loss[loss=0.1912, ctc_loss=0.1269, cr_loss=0.3212, over 16990.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.1312, cr_loss=0.3471, over 3340777.12 frames. ], batch size: 53, lr: 4.08e-03, grad_scale: 32.0
2024-09-24 14:52:34,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=525648.6666666666, ans=0.0
2024-09-24 14:52:50,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=525695.3333333334, ans=0.125
2024-09-24 14:52:54,489 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.276e+02 1.353e+02 1.468e+02 1.866e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-24 14:53:02,852 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.24 vs. limit=15.0
2024-09-24 14:53:52,065 INFO [train.py:1198] (1/4) Epoch 29, batch 3600, loss[loss=0.2139, ctc_loss=0.1391, cr_loss=0.3741, over 17021.00 frames. ], tot_loss[loss=0.2003, ctc_loss=0.1309, cr_loss=0.3468, over 3346415.07 frames. ], batch size: 56, lr: 4.08e-03, grad_scale: 32.0
2024-09-24 14:54:01,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=525882.0, ans=0.125
2024-09-24 14:54:06,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=525928.6666666666, ans=0.0
2024-09-24 14:54:09,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=525928.6666666666, ans=0.125
2024-09-24 14:54:34,863 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.25 vs. limit=15.0
2024-09-24 14:55:04,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=526068.6666666666, ans=0.125
2024-09-24 14:55:08,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=526115.3333333334, ans=0.1
2024-09-24 14:55:10,105 INFO [train.py:1198] (1/4) Epoch 29, batch 3650, loss[loss=0.2106, ctc_loss=0.137, cr_loss=0.3676, over 16719.00 frames. ], tot_loss[loss=0.2005, ctc_loss=0.131, cr_loss=0.3475, over 3355812.65 frames. ], batch size: 61, lr: 4.08e-03, grad_scale: 32.0
2024-09-24 14:55:24,783 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.50 vs. limit=10.0
2024-09-24 14:55:31,931 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.251e+02 1.330e+02 1.491e+02 2.044e+02, threshold=2.661e+02, percent-clipped=0.0
2024-09-24 14:55:43,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=526208.6666666666, ans=0.125
2024-09-24 14:56:31,269 INFO [train.py:1198] (1/4) Epoch 29, batch 3700, loss[loss=0.2072, ctc_loss=0.1353, cr_loss=0.3595, over 17298.00 frames. ], tot_loss[loss=0.2, ctc_loss=0.1306, cr_loss=0.3469, over 3356484.19 frames. ], batch size: 49, lr: 4.08e-03, grad_scale: 16.0
2024-09-24 14:56:31,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=526348.6666666666, ans=0.0
2024-09-24 14:56:45,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=526395.3333333334, ans=0.2
2024-09-24 14:57:10,778 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 14:57:21,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=526488.6666666666, ans=0.125
2024-09-24 14:57:23,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=526488.6666666666, ans=0.025
2024-09-24 14:57:37,945 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.96 vs. limit=10.0
2024-09-24 14:57:51,171 INFO [train.py:1198] (1/4) Epoch 29, batch 3750, loss[loss=0.2575, ctc_loss=0.1712, cr_loss=0.4317, over 15128.00 frames. ], tot_loss[loss=0.1999, ctc_loss=0.1306, cr_loss=0.3462, over 3358139.99 frames. ], batch size: 89, lr: 4.08e-03, grad_scale: 16.0
2024-09-24 14:58:13,097 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.272e+02 1.350e+02 1.451e+02 2.081e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-24 14:58:22,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=526675.3333333334, ans=0.0
2024-09-24 14:58:27,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=526675.3333333334, ans=0.2
2024-09-24 14:58:30,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=526675.3333333334, ans=0.2
2024-09-24 14:58:49,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=526722.0, ans=0.125
2024-09-24 14:58:55,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=526768.6666666666, ans=0.05
2024-09-24 14:59:04,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=526768.6666666666, ans=0.125
2024-09-24 14:59:07,828 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.39 vs. limit=15.0
2024-09-24 14:59:10,334 INFO [train.py:1198] (1/4) Epoch 29, batch 3800, loss[loss=0.1768, ctc_loss=0.1129, cr_loss=0.3196, over 16316.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1309, cr_loss=0.3462, over 3343125.15 frames. ], batch size: 36, lr: 4.08e-03, grad_scale: 16.0
2024-09-24 14:59:37,673 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.56 vs. limit=10.0
2024-09-24 14:59:49,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=526908.6666666666, ans=0.125
2024-09-24 15:00:17,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=527002.0, ans=0.0
2024-09-24 15:00:18,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=527002.0, ans=0.1
2024-09-24 15:00:27,867 INFO [train.py:1198] (1/4) Epoch 29, batch 3850, loss[loss=0.2424, ctc_loss=0.1639, cr_loss=0.3926, over 16486.00 frames. ], tot_loss[loss=0.2014, ctc_loss=0.132, cr_loss=0.3467, over 3312621.61 frames. ], batch size: 66, lr: 4.08e-03, grad_scale: 16.0
2024-09-24 15:00:29,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=527048.6666666666, ans=0.0
2024-09-24 15:00:49,021 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.344e+02 1.460e+02 1.620e+02 2.302e+02, threshold=2.919e+02, percent-clipped=0.0
2024-09-24 15:00:56,660 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=9.35 vs. limit=15.0
2024-09-24 15:01:24,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=527188.6666666666, ans=0.2
2024-09-24 15:02:28,429 INFO [train.py:1198] (1/4) Epoch 30, batch 0, loss[loss=0.2302, ctc_loss=0.1525, cr_loss=0.3885, over 15912.00 frames. ], tot_loss[loss=0.2302, ctc_loss=0.1525, cr_loss=0.3885, over 15912.00 frames. ], batch size: 74, lr: 4.01e-03, grad_scale: 32.0
2024-09-24 15:02:28,430 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 15:02:35,278 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.7276, 4.1345, 4.1226, 4.3398], device='cuda:1')
2024-09-24 15:02:43,742 INFO [train.py:1230] (1/4) Epoch 30, validation: loss=0.0352, ctc_loss=0.0352, cr_loss=9.262e-15, over 944034.00 frames. 
2024-09-24 15:02:43,743 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 15:02:50,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=527263.3333333334, ans=0.2
2024-09-24 15:03:14,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=527310.0, ans=0.125
2024-09-24 15:03:54,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=527450.0, ans=0.125
2024-09-24 15:04:07,010 INFO [train.py:1198] (1/4) Epoch 30, batch 50, loss[loss=0.2099, ctc_loss=0.1388, cr_loss=0.3555, over 16538.00 frames. ], tot_loss[loss=0.2015, ctc_loss=0.132, cr_loss=0.3476, over 748156.79 frames. ], batch size: 66, lr: 4.01e-03, grad_scale: 32.0
2024-09-24 15:04:23,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=527543.3333333334, ans=0.125
2024-09-24 15:04:32,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=527543.3333333334, ans=0.1
2024-09-24 15:04:35,822 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.261e+02 1.395e+02 1.583e+02 2.602e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-24 15:05:05,092 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.86 vs. limit=6.0
2024-09-24 15:05:09,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=527636.6666666666, ans=0.0
2024-09-24 15:05:14,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=527683.3333333334, ans=0.125
2024-09-24 15:05:30,395 INFO [train.py:1198] (1/4) Epoch 30, batch 100, loss[loss=0.2394, ctc_loss=0.1576, cr_loss=0.4093, over 17197.00 frames. ], tot_loss[loss=0.2008, ctc_loss=0.1312, cr_loss=0.3483, over 1333712.79 frames. ], batch size: 55, lr: 4.01e-03, grad_scale: 32.0
2024-09-24 15:06:04,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=527823.3333333334, ans=0.125
2024-09-24 15:06:51,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=527963.3333333334, ans=0.1
2024-09-24 15:06:55,899 INFO [train.py:1198] (1/4) Epoch 30, batch 150, loss[loss=0.2162, ctc_loss=0.1431, cr_loss=0.3658, over 17221.00 frames. ], tot_loss[loss=0.2012, ctc_loss=0.1315, cr_loss=0.3483, over 1777931.91 frames. ], batch size: 55, lr: 4.01e-03, grad_scale: 32.0
2024-09-24 15:07:00,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=527963.3333333334, ans=0.1
2024-09-24 15:07:07,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=527963.3333333334, ans=0.2
2024-09-24 15:07:12,365 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.09 vs. limit=15.0
2024-09-24 15:07:24,703 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.297e+02 1.402e+02 1.523e+02 2.631e+02, threshold=2.805e+02, percent-clipped=0.0
2024-09-24 15:07:29,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=528056.6666666666, ans=0.1
2024-09-24 15:07:39,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=528056.6666666666, ans=0.125
2024-09-24 15:08:19,312 INFO [train.py:1198] (1/4) Epoch 30, batch 200, loss[loss=0.19, ctc_loss=0.1225, cr_loss=0.3377, over 17103.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.1301, cr_loss=0.3468, over 2130400.79 frames. ], batch size: 49, lr: 4.01e-03, grad_scale: 32.0
2024-09-24 15:08:27,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=528196.6666666666, ans=0.0
2024-09-24 15:08:30,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=528196.6666666666, ans=0.1
2024-09-24 15:08:32,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=528196.6666666666, ans=0.0
2024-09-24 15:08:47,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.78 vs. limit=12.0
2024-09-24 15:08:58,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=528290.0, ans=0.125
2024-09-24 15:09:01,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=528290.0, ans=0.125
2024-09-24 15:09:17,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=528336.6666666666, ans=0.0
2024-09-24 15:09:28,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=528383.3333333334, ans=0.025
2024-09-24 15:09:42,473 INFO [train.py:1198] (1/4) Epoch 30, batch 250, loss[loss=0.2014, ctc_loss=0.1305, cr_loss=0.3546, over 16716.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1311, cr_loss=0.3482, over 2394498.76 frames. ], batch size: 61, lr: 4.00e-03, grad_scale: 32.0
2024-09-24 15:09:52,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=528430.0, ans=0.1
2024-09-24 15:09:55,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=528430.0, ans=0.125
2024-09-24 15:10:11,297 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.266e+02 1.350e+02 1.451e+02 1.821e+02, threshold=2.699e+02, percent-clipped=0.0
2024-09-24 15:10:16,487 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:10:24,870 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.26 vs. limit=15.0
2024-09-24 15:10:30,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=528570.0, ans=0.2
2024-09-24 15:10:34,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=528570.0, ans=0.0
2024-09-24 15:10:36,283 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.33 vs. limit=15.0
2024-09-24 15:10:58,248 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:11:00,174 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.06 vs. limit=6.0
2024-09-24 15:11:02,796 INFO [train.py:1198] (1/4) Epoch 30, batch 300, loss[loss=0.1856, ctc_loss=0.1194, cr_loss=0.3313, over 16272.00 frames. ], tot_loss[loss=0.2008, ctc_loss=0.1312, cr_loss=0.3483, over 2609085.31 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 32.0
2024-09-24 15:11:22,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=528710.0, ans=0.025
2024-09-24 15:11:28,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=528710.0, ans=0.0
2024-09-24 15:11:41,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=528756.6666666666, ans=0.125
2024-09-24 15:12:14,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=528850.0, ans=0.1
2024-09-24 15:12:18,445 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.66 vs. limit=6.0
2024-09-24 15:12:28,505 INFO [train.py:1198] (1/4) Epoch 30, batch 350, loss[loss=0.1976, ctc_loss=0.1284, cr_loss=0.3461, over 17096.00 frames. ], tot_loss[loss=0.2018, ctc_loss=0.1319, cr_loss=0.3496, over 2776997.51 frames. ], batch size: 49, lr: 4.00e-03, grad_scale: 32.0
2024-09-24 15:12:36,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=528896.6666666666, ans=0.125
2024-09-24 15:12:47,224 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=11.17 vs. limit=22.5
2024-09-24 15:13:00,084 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.261e+02 1.365e+02 1.554e+02 1.989e+02, threshold=2.731e+02, percent-clipped=0.0
2024-09-24 15:13:19,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=529036.6666666666, ans=0.1
2024-09-24 15:13:22,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=529036.6666666666, ans=0.125
2024-09-24 15:13:51,536 INFO [train.py:1198] (1/4) Epoch 30, batch 400, loss[loss=0.1925, ctc_loss=0.1225, cr_loss=0.3497, over 17068.00 frames. ], tot_loss[loss=0.2014, ctc_loss=0.1316, cr_loss=0.3494, over 2914578.26 frames. ], batch size: 46, lr: 4.00e-03, grad_scale: 32.0
2024-09-24 15:14:01,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=529130.0, ans=0.125
2024-09-24 15:14:17,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=529176.6666666666, ans=0.0
2024-09-24 15:14:26,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.04 vs. limit=15.0
2024-09-24 15:14:38,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=529270.0, ans=0.1
2024-09-24 15:14:38,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=529270.0, ans=0.0
2024-09-24 15:14:51,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=529270.0, ans=0.125
2024-09-24 15:15:14,475 INFO [train.py:1198] (1/4) Epoch 30, batch 450, loss[loss=0.2528, ctc_loss=0.1677, cr_loss=0.4255, over 16622.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1298, cr_loss=0.3465, over 3020097.85 frames. ], batch size: 66, lr: 4.00e-03, grad_scale: 16.0
2024-09-24 15:15:34,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=529410.0, ans=0.125
2024-09-24 15:15:44,835 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.260e+02 1.335e+02 1.450e+02 2.256e+02, threshold=2.670e+02, percent-clipped=0.0
2024-09-24 15:16:23,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=529550.0, ans=0.125
2024-09-24 15:16:34,281 INFO [train.py:1198] (1/4) Epoch 30, batch 500, loss[loss=0.2123, ctc_loss=0.1355, cr_loss=0.3842, over 17056.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1304, cr_loss=0.3481, over 3099430.06 frames. ], batch size: 56, lr: 4.00e-03, grad_scale: 16.0
2024-09-24 15:16:37,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=529596.6666666666, ans=0.0
2024-09-24 15:17:34,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=529736.6666666666, ans=0.035
2024-09-24 15:17:34,721 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=6.47 vs. limit=12.0
2024-09-24 15:17:42,715 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:18:01,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=529830.0, ans=0.0
2024-09-24 15:18:02,945 INFO [train.py:1198] (1/4) Epoch 30, batch 550, loss[loss=0.2025, ctc_loss=0.1319, cr_loss=0.353, over 17011.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.13, cr_loss=0.3472, over 3161713.04 frames. ], batch size: 53, lr: 4.00e-03, grad_scale: 16.0
2024-09-24 15:18:19,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=529876.6666666666, ans=0.0
2024-09-24 15:18:21,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=529876.6666666666, ans=0.025
2024-09-24 15:18:28,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=529876.6666666666, ans=0.125
2024-09-24 15:18:29,318 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=8.52 vs. limit=15.0
2024-09-24 15:18:33,366 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.259e+02 1.342e+02 1.453e+02 2.055e+02, threshold=2.683e+02, percent-clipped=0.0
2024-09-24 15:18:33,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=529923.3333333334, ans=0.1
2024-09-24 15:18:48,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=529923.3333333334, ans=0.0
2024-09-24 15:19:13,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=530016.6666666666, ans=0.125
2024-09-24 15:19:23,318 INFO [train.py:1198] (1/4) Epoch 30, batch 600, loss[loss=0.2167, ctc_loss=0.1437, cr_loss=0.3652, over 17212.00 frames. ], tot_loss[loss=0.1999, ctc_loss=0.1304, cr_loss=0.3477, over 3196034.80 frames. ], batch size: 50, lr: 4.00e-03, grad_scale: 16.0
2024-09-24 15:19:40,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=530110.0, ans=0.2
2024-09-24 15:19:56,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=530156.6666666666, ans=0.2
2024-09-24 15:19:56,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.min_positive, batch_count=530156.6666666666, ans=0.025
2024-09-24 15:20:01,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=530156.6666666666, ans=0.125
2024-09-24 15:20:06,745 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.64 vs. limit=12.0
2024-09-24 15:20:09,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=530156.6666666666, ans=0.0
2024-09-24 15:20:12,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=530203.3333333334, ans=0.0
2024-09-24 15:20:14,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=530203.3333333334, ans=0.5
2024-09-24 15:20:45,717 INFO [train.py:1198] (1/4) Epoch 30, batch 650, loss[loss=0.1877, ctc_loss=0.1225, cr_loss=0.326, over 16744.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1311, cr_loss=0.3495, over 3237911.76 frames. ], batch size: 61, lr: 4.00e-03, grad_scale: 16.0
2024-09-24 15:20:46,930 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.99 vs. limit=15.0
2024-09-24 15:20:52,950 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.51 vs. limit=6.0
2024-09-24 15:21:11,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=530343.3333333334, ans=0.2
2024-09-24 15:21:15,828 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.275e+02 1.371e+02 1.445e+02 2.518e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-24 15:21:16,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=530390.0, ans=0.125
2024-09-24 15:21:44,857 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.33 vs. limit=15.0
2024-09-24 15:21:47,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=530436.6666666666, ans=0.125
2024-09-24 15:22:10,436 INFO [train.py:1198] (1/4) Epoch 30, batch 700, loss[loss=0.2078, ctc_loss=0.1361, cr_loss=0.3587, over 17027.00 frames. ], tot_loss[loss=0.2005, ctc_loss=0.1307, cr_loss=0.3488, over 3259681.38 frames. ], batch size: 52, lr: 4.00e-03, grad_scale: 16.0
2024-09-24 15:22:13,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=530530.0, ans=0.025
2024-09-24 15:22:14,448 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.20 vs. limit=22.5
2024-09-24 15:22:23,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=530530.0, ans=0.2
2024-09-24 15:22:25,551 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.62 vs. limit=15.0
2024-09-24 15:23:15,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=530716.6666666666, ans=0.025
2024-09-24 15:23:15,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=530716.6666666666, ans=0.125
2024-09-24 15:23:17,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=530716.6666666666, ans=0.125
2024-09-24 15:23:23,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=530716.6666666666, ans=0.125
2024-09-24 15:23:23,977 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:23:33,171 INFO [train.py:1198] (1/4) Epoch 30, batch 750, loss[loss=0.2202, ctc_loss=0.1452, cr_loss=0.3749, over 17305.00 frames. ], tot_loss[loss=0.1992, ctc_loss=0.1297, cr_loss=0.3472, over 3281508.61 frames. ], batch size: 49, lr: 4.00e-03, grad_scale: 16.0
2024-09-24 15:23:48,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=530810.0, ans=0.125
2024-09-24 15:23:51,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=530810.0, ans=0.07
2024-09-24 15:24:04,035 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.243e+02 1.346e+02 1.462e+02 2.306e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-24 15:24:04,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=530856.6666666666, ans=0.035
2024-09-24 15:24:20,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=530903.3333333334, ans=0.125
2024-09-24 15:24:31,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=530903.3333333334, ans=0.125
2024-09-24 15:24:50,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=530950.0, ans=0.125
2024-09-24 15:24:56,185 INFO [train.py:1198] (1/4) Epoch 30, batch 800, loss[loss=0.2237, ctc_loss=0.1465, cr_loss=0.3861, over 17200.00 frames. ], tot_loss[loss=0.2005, ctc_loss=0.1308, cr_loss=0.3484, over 3299461.55 frames. ], batch size: 55, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:25:31,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=531090.0, ans=0.0
2024-09-24 15:25:43,301 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.57 vs. limit=15.0
2024-09-24 15:25:51,164 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.50 vs. limit=15.0
2024-09-24 15:25:52,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=531136.6666666666, ans=0.125
2024-09-24 15:26:16,196 INFO [train.py:1198] (1/4) Epoch 30, batch 850, loss[loss=0.2354, ctc_loss=0.1566, cr_loss=0.3938, over 17025.00 frames. ], tot_loss[loss=0.2013, ctc_loss=0.1314, cr_loss=0.3496, over 3311982.95 frames. ], batch size: 56, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:26:19,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=531230.0, ans=0.125
2024-09-24 15:26:21,473 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.45 vs. limit=12.0
2024-09-24 15:26:27,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=531230.0, ans=0.025
2024-09-24 15:26:53,452 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.016e+02 1.252e+02 1.339e+02 1.409e+02 2.350e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-24 15:27:04,528 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=10.00 vs. limit=10.0
2024-09-24 15:27:33,197 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.11 vs. limit=6.0
2024-09-24 15:27:41,874 INFO [train.py:1198] (1/4) Epoch 30, batch 900, loss[loss=0.1535, ctc_loss=0.09762, cr_loss=0.2794, over 17133.00 frames. ], tot_loss[loss=0.2008, ctc_loss=0.131, cr_loss=0.3489, over 3319724.50 frames. ], batch size: 40, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:27:48,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=531463.3333333334, ans=0.025
2024-09-24 15:27:51,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=531463.3333333334, ans=0.0
2024-09-24 15:28:05,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=531510.0, ans=0.125
2024-09-24 15:28:17,261 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.37 vs. limit=10.0
2024-09-24 15:28:22,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=531556.6666666666, ans=0.0
2024-09-24 15:28:27,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=531556.6666666666, ans=0.125
2024-09-24 15:28:53,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=531650.0, ans=0.1
2024-09-24 15:29:04,215 INFO [train.py:1198] (1/4) Epoch 30, batch 950, loss[loss=0.2284, ctc_loss=0.1545, cr_loss=0.3694, over 17221.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1317, cr_loss=0.3499, over 3327824.24 frames. ], batch size: 50, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:29:12,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=531696.6666666666, ans=0.0
2024-09-24 15:29:15,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=531696.6666666666, ans=0.2
2024-09-24 15:29:18,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=531743.3333333334, ans=0.0
2024-09-24 15:29:32,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=531743.3333333334, ans=0.125
2024-09-24 15:29:34,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=531790.0, ans=0.2
2024-09-24 15:29:35,837 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.249e+02 1.333e+02 1.422e+02 1.714e+02, threshold=2.667e+02, percent-clipped=0.0
2024-09-24 15:29:39,473 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:30:26,509 INFO [train.py:1198] (1/4) Epoch 30, batch 1000, loss[loss=0.2029, ctc_loss=0.1333, cr_loss=0.348, over 17266.00 frames. ], tot_loss[loss=0.2012, ctc_loss=0.1314, cr_loss=0.349, over 3328161.41 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:30:38,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=531930.0, ans=0.0
2024-09-24 15:30:43,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=531976.6666666666, ans=0.125
2024-09-24 15:30:58,389 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.86 vs. limit=15.0
2024-09-24 15:31:15,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=532070.0, ans=0.1
2024-09-24 15:31:36,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=532116.6666666666, ans=0.05
2024-09-24 15:31:51,972 INFO [train.py:1198] (1/4) Epoch 30, batch 1050, loss[loss=0.1589, ctc_loss=0.1019, cr_loss=0.2846, over 17261.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1301, cr_loss=0.3465, over 3345309.45 frames. ], batch size: 42, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:32:03,770 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.54 vs. limit=15.0
2024-09-24 15:32:23,922 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.266e+02 1.377e+02 1.519e+02 2.640e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-24 15:32:24,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=532256.6666666666, ans=0.1
2024-09-24 15:32:25,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=532256.6666666666, ans=0.125
2024-09-24 15:32:29,760 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.40 vs. limit=15.0
2024-09-24 15:32:51,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=532303.3333333334, ans=0.0
2024-09-24 15:33:06,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=532350.0, ans=0.125
2024-09-24 15:33:14,102 INFO [train.py:1198] (1/4) Epoch 30, batch 1100, loss[loss=0.226, ctc_loss=0.1496, cr_loss=0.382, over 16593.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.13, cr_loss=0.3459, over 3347485.98 frames. ], batch size: 66, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:33:28,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=532443.3333333334, ans=0.125
2024-09-24 15:33:39,973 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.46 vs. limit=22.5
2024-09-24 15:33:48,017 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.53 vs. limit=15.0
2024-09-24 15:33:50,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=532490.0, ans=0.2
2024-09-24 15:34:03,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=532536.6666666666, ans=0.95
2024-09-24 15:34:03,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=532536.6666666666, ans=0.125
2024-09-24 15:34:06,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=532536.6666666666, ans=0.125
2024-09-24 15:34:13,641 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.34 vs. limit=15.0
2024-09-24 15:34:33,774 INFO [train.py:1198] (1/4) Epoch 30, batch 1150, loss[loss=0.2128, ctc_loss=0.1397, cr_loss=0.3655, over 17171.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1295, cr_loss=0.346, over 3353840.59 frames. ], batch size: 45, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:35:09,706 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.299e+02 1.402e+02 1.532e+02 2.058e+02, threshold=2.804e+02, percent-clipped=0.0
2024-09-24 15:35:30,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=532770.0, ans=0.2
2024-09-24 15:35:36,227 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.69 vs. limit=15.0
2024-09-24 15:35:40,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=532816.6666666666, ans=0.125
2024-09-24 15:35:41,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=532816.6666666666, ans=0.125
2024-09-24 15:35:54,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=532863.3333333334, ans=0.125
2024-09-24 15:35:56,125 INFO [train.py:1198] (1/4) Epoch 30, batch 1200, loss[loss=0.1929, ctc_loss=0.1251, cr_loss=0.3391, over 17164.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1298, cr_loss=0.3465, over 3359701.01 frames. ], batch size: 45, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:35:56,796 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.77 vs. limit=10.0
2024-09-24 15:36:14,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=532910.0, ans=0.125
2024-09-24 15:36:16,560 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.79 vs. limit=12.0
2024-09-24 15:36:42,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=532956.6666666666, ans=0.0
2024-09-24 15:36:44,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=532956.6666666666, ans=0.0
2024-09-24 15:37:00,024 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.66 vs. limit=10.0
2024-09-24 15:37:10,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=533050.0, ans=0.125
2024-09-24 15:37:21,570 INFO [train.py:1198] (1/4) Epoch 30, batch 1250, loss[loss=0.218, ctc_loss=0.1434, cr_loss=0.3732, over 17019.00 frames. ], tot_loss[loss=0.1996, ctc_loss=0.1302, cr_loss=0.347, over 3362010.34 frames. ], batch size: 52, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:37:28,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=533096.6666666666, ans=0.125
2024-09-24 15:37:57,234 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.280e+02 1.378e+02 1.490e+02 1.846e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-24 15:37:59,209 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:38:32,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=533283.3333333334, ans=0.125
2024-09-24 15:38:32,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=533283.3333333334, ans=0.1
2024-09-24 15:38:34,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=533283.3333333334, ans=0.0
2024-09-24 15:38:35,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=533283.3333333334, ans=0.04949747468305833
2024-09-24 15:38:43,374 INFO [train.py:1198] (1/4) Epoch 30, batch 1300, loss[loss=0.1693, ctc_loss=0.1075, cr_loss=0.3093, over 17083.00 frames. ], tot_loss[loss=0.1998, ctc_loss=0.1304, cr_loss=0.3472, over 3357931.20 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:38:48,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=533330.0, ans=0.125
2024-09-24 15:38:56,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=533330.0, ans=0.07
2024-09-24 15:39:09,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=533376.6666666666, ans=0.2
2024-09-24 15:39:09,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=533376.6666666666, ans=0.2
2024-09-24 15:39:10,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=533376.6666666666, ans=0.1
2024-09-24 15:39:44,611 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.25 vs. limit=15.0
2024-09-24 15:40:05,653 INFO [train.py:1198] (1/4) Epoch 30, batch 1350, loss[loss=0.2409, ctc_loss=0.156, cr_loss=0.4244, over 16460.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1297, cr_loss=0.3464, over 3370934.84 frames. ], batch size: 66, lr: 3.99e-03, grad_scale: 16.0
2024-09-24 15:40:26,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=533610.0, ans=10.0
2024-09-24 15:40:29,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=533610.0, ans=0.2
2024-09-24 15:40:37,887 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.15 vs. limit=15.0
2024-09-24 15:40:38,858 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.284e+02 1.374e+02 1.506e+02 2.096e+02, threshold=2.749e+02, percent-clipped=0.0
2024-09-24 15:40:43,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=533656.6666666666, ans=0.015
2024-09-24 15:41:01,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=533703.3333333334, ans=0.125
2024-09-24 15:41:02,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.78 vs. limit=12.0
2024-09-24 15:41:08,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=533750.0, ans=0.5
2024-09-24 15:41:13,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=533750.0, ans=0.2
2024-09-24 15:41:26,035 INFO [train.py:1198] (1/4) Epoch 30, batch 1400, loss[loss=0.2137, ctc_loss=0.1403, cr_loss=0.3668, over 17011.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1306, cr_loss=0.3475, over 3363482.17 frames. ], batch size: 56, lr: 3.98e-03, grad_scale: 16.0
2024-09-24 15:41:53,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=533843.3333333334, ans=0.125
2024-09-24 15:42:15,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=533890.0, ans=0.125
2024-09-24 15:42:41,846 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.22 vs. limit=15.0
2024-09-24 15:42:54,635 INFO [train.py:1198] (1/4) Epoch 30, batch 1450, loss[loss=0.2474, ctc_loss=0.165, cr_loss=0.4121, over 15954.00 frames. ], tot_loss[loss=0.2004, ctc_loss=0.1308, cr_loss=0.3479, over 3358379.84 frames. ], batch size: 74, lr: 3.98e-03, grad_scale: 16.0
2024-09-24 15:43:28,225 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.340e+02 1.457e+02 1.567e+02 2.615e+02, threshold=2.914e+02, percent-clipped=0.0
2024-09-24 15:43:30,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=534123.3333333334, ans=0.0
2024-09-24 15:43:59,136 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=10.27 vs. limit=22.5
2024-09-24 15:44:06,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=534216.6666666666, ans=0.2
2024-09-24 15:44:14,826 INFO [train.py:1198] (1/4) Epoch 30, batch 1500, loss[loss=0.1887, ctc_loss=0.1204, cr_loss=0.3416, over 17250.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1313, cr_loss=0.3489, over 3360955.52 frames. ], batch size: 44, lr: 3.98e-03, grad_scale: 16.0
2024-09-24 15:44:18,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=534263.3333333334, ans=0.025
2024-09-24 15:44:20,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=534263.3333333334, ans=0.125
2024-09-24 15:44:37,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=534310.0, ans=0.1
2024-09-24 15:44:48,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=534356.6666666666, ans=0.0
2024-09-24 15:45:08,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=534403.3333333334, ans=0.0
2024-09-24 15:45:11,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=534403.3333333334, ans=0.2
2024-09-24 15:45:29,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=534450.0, ans=0.2
2024-09-24 15:45:36,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.42 vs. limit=22.5
2024-09-24 15:45:37,368 INFO [train.py:1198] (1/4) Epoch 30, batch 1550, loss[loss=0.2213, ctc_loss=0.1465, cr_loss=0.3743, over 16440.00 frames. ], tot_loss[loss=0.2025, ctc_loss=0.1324, cr_loss=0.3507, over 3348817.21 frames. ], batch size: 66, lr: 3.98e-03, grad_scale: 16.0
2024-09-24 15:46:03,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=534543.3333333334, ans=0.125
2024-09-24 15:46:05,089 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:46:11,088 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.255e+02 1.331e+02 1.418e+02 1.761e+02, threshold=2.662e+02, percent-clipped=0.0
2024-09-24 15:46:13,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=534590.0, ans=0.2
2024-09-24 15:47:02,574 INFO [train.py:1198] (1/4) Epoch 30, batch 1600, loss[loss=0.2588, ctc_loss=0.1795, cr_loss=0.3967, over 16966.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1325, cr_loss=0.3503, over 3344121.28 frames. ], batch size: 56, lr: 3.98e-03, grad_scale: 32.0
2024-09-24 15:47:06,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=534730.0, ans=0.2
2024-09-24 15:47:47,804 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.25 vs. limit=22.5
2024-09-24 15:47:50,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=534823.3333333334, ans=0.1
2024-09-24 15:47:57,091 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 15:48:16,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=534916.6666666666, ans=0.125
2024-09-24 15:48:25,594 INFO [train.py:1198] (1/4) Epoch 30, batch 1650, loss[loss=0.2305, ctc_loss=0.1535, cr_loss=0.3853, over 17037.00 frames. ], tot_loss[loss=0.2034, ctc_loss=0.133, cr_loss=0.3516, over 3352003.55 frames. ], batch size: 52, lr: 3.98e-03, grad_scale: 32.0
2024-09-24 15:48:49,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=535010.0, ans=0.125
2024-09-24 15:48:59,209 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.272e+02 1.342e+02 1.418e+02 1.836e+02, threshold=2.683e+02, percent-clipped=0.0
2024-09-24 15:49:05,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=535056.6666666666, ans=0.125
2024-09-24 15:49:45,317 INFO [train.py:1198] (1/4) Epoch 30, batch 1700, loss[loss=0.1732, ctc_loss=0.1085, cr_loss=0.3234, over 17191.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.1321, cr_loss=0.3501, over 3349553.62 frames. ], batch size: 41, lr: 3.98e-03, grad_scale: 32.0
2024-09-24 15:49:52,004 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.20 vs. limit=15.0
2024-09-24 15:49:57,186 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.72 vs. limit=15.0
2024-09-24 15:50:01,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=535196.6666666666, ans=0.025
2024-09-24 15:50:36,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=535336.6666666666, ans=0.125
2024-09-24 15:51:08,582 INFO [train.py:1198] (1/4) Epoch 30, batch 1750, loss[loss=0.2349, ctc_loss=0.1533, cr_loss=0.4081, over 16891.00 frames. ], tot_loss[loss=0.2008, ctc_loss=0.131, cr_loss=0.3487, over 3349859.65 frames. ], batch size: 58, lr: 3.98e-03, grad_scale: 32.0
2024-09-24 15:51:10,993 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.77 vs. limit=10.0
2024-09-24 15:51:47,334 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.310e+02 1.375e+02 1.459e+02 2.196e+02, threshold=2.749e+02, percent-clipped=0.0
2024-09-24 15:51:53,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=535523.3333333334, ans=0.0
2024-09-24 15:52:01,220 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.77 vs. limit=6.0
2024-09-24 15:52:33,548 INFO [train.py:1198] (1/4) Epoch 30, batch 1800, loss[loss=0.185, ctc_loss=0.1169, cr_loss=0.3404, over 17103.00 frames. ], tot_loss[loss=0.2015, ctc_loss=0.1316, cr_loss=0.3494, over 3347027.86 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 32.0
2024-09-24 15:52:35,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=535663.3333333334, ans=0.1
2024-09-24 15:52:37,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=535663.3333333334, ans=0.2
2024-09-24 15:52:59,472 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.38 vs. limit=12.0
2024-09-24 15:53:08,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=535756.6666666666, ans=0.025
2024-09-24 15:53:33,617 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=14.06 vs. limit=22.5
2024-09-24 15:53:56,383 INFO [train.py:1198] (1/4) Epoch 30, batch 1850, loss[loss=0.1795, ctc_loss=0.1142, cr_loss=0.3264, over 17194.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1312, cr_loss=0.3484, over 3342945.23 frames. ], batch size: 41, lr: 3.98e-03, grad_scale: 32.0
2024-09-24 15:54:11,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=535943.3333333334, ans=0.95
2024-09-24 15:54:25,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=535943.3333333334, ans=0.125
2024-09-24 15:54:30,249 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.250e+02 1.313e+02 1.396e+02 1.985e+02, threshold=2.627e+02, percent-clipped=0.0
2024-09-24 15:54:40,734 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=7.93 vs. limit=22.5
2024-09-24 15:54:41,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.41 vs. limit=22.5
2024-09-24 15:55:19,283 INFO [train.py:1198] (1/4) Epoch 30, batch 1900, loss[loss=0.2327, ctc_loss=0.1584, cr_loss=0.3719, over 15104.00 frames. ], tot_loss[loss=0.2004, ctc_loss=0.1308, cr_loss=0.3479, over 3344753.96 frames. ], batch size: 89, lr: 3.98e-03, grad_scale: 32.0
2024-09-24 15:55:19,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=536130.0, ans=0.0
2024-09-24 15:55:21,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=536130.0, ans=0.025
2024-09-24 15:55:26,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.84 vs. limit=15.0
2024-09-24 15:56:17,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=536270.0, ans=0.125
2024-09-24 15:56:23,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=536316.6666666666, ans=0.0
2024-09-24 15:56:34,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=536316.6666666666, ans=0.125
2024-09-24 15:56:36,179 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.30 vs. limit=15.0
2024-09-24 15:56:44,474 INFO [train.py:1198] (1/4) Epoch 30, batch 1950, loss[loss=0.1773, ctc_loss=0.1134, cr_loss=0.3196, over 17110.00 frames. ], tot_loss[loss=0.2002, ctc_loss=0.1306, cr_loss=0.3481, over 3356445.14 frames. ], batch size: 40, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 15:56:51,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=536363.3333333334, ans=0.035
2024-09-24 15:57:03,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=536410.0, ans=0.125
2024-09-24 15:57:13,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=536410.0, ans=0.1
2024-09-24 15:57:16,788 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=16.24 vs. limit=22.5
2024-09-24 15:57:19,471 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.270e+02 1.356e+02 1.450e+02 2.095e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-24 15:57:19,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=536456.6666666666, ans=0.125
2024-09-24 15:57:35,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=536503.3333333334, ans=0.2
2024-09-24 15:57:46,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=536503.3333333334, ans=0.09899494936611666
2024-09-24 15:57:53,056 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.96 vs. limit=15.0
2024-09-24 15:57:55,040 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.83 vs. limit=15.0
2024-09-24 15:57:59,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=536550.0, ans=0.025
2024-09-24 15:58:02,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=536550.0, ans=0.125
2024-09-24 15:58:06,788 INFO [train.py:1198] (1/4) Epoch 30, batch 2000, loss[loss=0.1841, ctc_loss=0.1181, cr_loss=0.3299, over 17098.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1299, cr_loss=0.3474, over 3367715.36 frames. ], batch size: 43, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 15:58:07,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=536596.6666666666, ans=0.05
2024-09-24 15:58:16,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=536596.6666666666, ans=0.125
2024-09-24 15:58:20,123 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.05 vs. limit=15.0
2024-09-24 15:58:25,332 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.18 vs. limit=12.0
2024-09-24 15:58:45,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=536690.0, ans=0.125
2024-09-24 15:59:04,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=536736.6666666666, ans=0.125
2024-09-24 15:59:26,920 INFO [train.py:1198] (1/4) Epoch 30, batch 2050, loss[loss=0.2285, ctc_loss=0.1521, cr_loss=0.3823, over 17024.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1297, cr_loss=0.3468, over 3364328.43 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 15:59:36,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=536830.0, ans=0.125
2024-09-24 15:59:51,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.88 vs. limit=15.0
2024-09-24 15:59:53,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=536876.6666666666, ans=0.025
2024-09-24 16:00:00,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=536923.3333333334, ans=0.025
2024-09-24 16:00:04,534 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.272e+02 1.366e+02 1.463e+02 2.391e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-24 16:00:04,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=536923.3333333334, ans=0.2
2024-09-24 16:00:27,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=536970.0, ans=0.125
2024-09-24 16:00:35,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=537016.6666666666, ans=0.125
2024-09-24 16:00:46,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=537016.6666666666, ans=0.125
2024-09-24 16:00:49,745 INFO [train.py:1198] (1/4) Epoch 30, batch 2100, loss[loss=0.1974, ctc_loss=0.1292, cr_loss=0.3408, over 17259.00 frames. ], tot_loss[loss=0.1997, ctc_loss=0.1302, cr_loss=0.3473, over 3367308.01 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:00:53,830 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.39 vs. limit=22.5
2024-09-24 16:01:07,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=537110.0, ans=0.125
2024-09-24 16:01:18,681 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=16.87 vs. limit=22.5
2024-09-24 16:01:40,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=537203.3333333334, ans=0.125
2024-09-24 16:01:54,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=537203.3333333334, ans=0.125
2024-09-24 16:02:13,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=537296.6666666666, ans=0.0
2024-09-24 16:02:14,604 INFO [train.py:1198] (1/4) Epoch 30, batch 2150, loss[loss=0.1896, ctc_loss=0.1197, cr_loss=0.3497, over 17241.00 frames. ], tot_loss[loss=0.201, ctc_loss=0.1312, cr_loss=0.3489, over 3357404.00 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:02:15,225 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=10.81 vs. limit=22.5
2024-09-24 16:02:29,611 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.46 vs. limit=22.5
2024-09-24 16:02:49,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=537390.0, ans=0.125
2024-09-24 16:02:52,196 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.245e+02 1.327e+02 1.449e+02 2.310e+02, threshold=2.654e+02, percent-clipped=0.0
2024-09-24 16:03:19,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=537483.3333333334, ans=0.025
2024-09-24 16:03:36,774 INFO [train.py:1198] (1/4) Epoch 30, batch 2200, loss[loss=0.1914, ctc_loss=0.1245, cr_loss=0.3343, over 17361.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.131, cr_loss=0.3481, over 3361288.53 frames. ], batch size: 48, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:03:40,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=537530.0, ans=0.1
2024-09-24 16:03:48,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=537530.0, ans=0.2
2024-09-24 16:03:57,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=537576.6666666666, ans=0.5
2024-09-24 16:04:12,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=537623.3333333334, ans=0.125
2024-09-24 16:04:44,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.10 vs. limit=15.0
2024-09-24 16:04:47,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=537716.6666666666, ans=0.1
2024-09-24 16:05:00,238 INFO [train.py:1198] (1/4) Epoch 30, batch 2250, loss[loss=0.2137, ctc_loss=0.1406, cr_loss=0.3655, over 16853.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1311, cr_loss=0.348, over 3352691.25 frames. ], batch size: 58, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:05:02,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=537763.3333333334, ans=0.2
2024-09-24 16:05:02,890 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.70 vs. limit=15.0
2024-09-24 16:05:11,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=537763.3333333334, ans=0.1
2024-09-24 16:05:16,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=537810.0, ans=0.125
2024-09-24 16:05:30,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=537856.6666666666, ans=0.0
2024-09-24 16:05:33,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.02 vs. limit=15.0
2024-09-24 16:05:35,467 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.295e+02 1.394e+02 1.566e+02 2.386e+02, threshold=2.787e+02, percent-clipped=0.0
2024-09-24 16:05:51,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=537903.3333333334, ans=0.95
2024-09-24 16:05:53,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=537903.3333333334, ans=0.125
2024-09-24 16:06:04,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=537950.0, ans=0.125
2024-09-24 16:06:04,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=537950.0, ans=0.0
2024-09-24 16:06:14,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=537950.0, ans=0.025
2024-09-24 16:06:20,101 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.09 vs. limit=22.5
2024-09-24 16:06:20,401 INFO [train.py:1198] (1/4) Epoch 30, batch 2300, loss[loss=0.2197, ctc_loss=0.1464, cr_loss=0.3664, over 16936.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1319, cr_loss=0.35, over 3346872.76 frames. ], batch size: 42, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:06:43,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=538043.3333333334, ans=0.125
2024-09-24 16:07:00,093 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.90 vs. limit=15.0
2024-09-24 16:07:31,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.16 vs. limit=15.0
2024-09-24 16:07:36,056 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.87 vs. limit=22.5
2024-09-24 16:07:47,754 INFO [train.py:1198] (1/4) Epoch 30, batch 2350, loss[loss=0.1686, ctc_loss=0.1069, cr_loss=0.309, over 17127.00 frames. ], tot_loss[loss=0.2013, ctc_loss=0.1315, cr_loss=0.3491, over 3352394.36 frames. ], batch size: 40, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:07:54,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=538230.0, ans=0.0
2024-09-24 16:08:15,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=538276.6666666666, ans=0.0
2024-09-24 16:08:23,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.275e+02 1.344e+02 1.471e+02 2.396e+02, threshold=2.688e+02, percent-clipped=0.0
2024-09-24 16:08:41,602 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.80 vs. limit=22.5
2024-09-24 16:08:58,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=538416.6666666666, ans=0.125
2024-09-24 16:09:08,045 INFO [train.py:1198] (1/4) Epoch 30, batch 2400, loss[loss=0.2113, ctc_loss=0.1447, cr_loss=0.333, over 11266.00 frames. ], tot_loss[loss=0.2028, ctc_loss=0.1325, cr_loss=0.3513, over 3344181.22 frames. ], batch size: 123, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:09:20,091 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=9.13 vs. limit=15.0
2024-09-24 16:09:21,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=538463.3333333334, ans=0.0
2024-09-24 16:09:27,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=538510.0, ans=0.05
2024-09-24 16:09:44,830 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.34 vs. limit=10.0
2024-09-24 16:09:47,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=538556.6666666666, ans=0.0
2024-09-24 16:09:49,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=538556.6666666666, ans=0.025
2024-09-24 16:10:20,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=538650.0, ans=0.1
2024-09-24 16:10:30,237 INFO [train.py:1198] (1/4) Epoch 30, batch 2450, loss[loss=0.2271, ctc_loss=0.1502, cr_loss=0.3848, over 17030.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1319, cr_loss=0.3499, over 3344771.87 frames. ], batch size: 53, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:10:51,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=538743.3333333334, ans=0.05
2024-09-24 16:11:05,480 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.261e+02 1.331e+02 1.438e+02 1.772e+02, threshold=2.662e+02, percent-clipped=0.0
2024-09-24 16:11:29,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=538836.6666666666, ans=0.025
2024-09-24 16:11:54,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=538930.0, ans=0.125
2024-09-24 16:11:54,637 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.08 vs. limit=22.5
2024-09-24 16:11:55,427 INFO [train.py:1198] (1/4) Epoch 30, batch 2500, loss[loss=0.1586, ctc_loss=0.1007, cr_loss=0.2895, over 17120.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1318, cr_loss=0.3497, over 3355358.51 frames. ], batch size: 40, lr: 3.97e-03, grad_scale: 32.0
2024-09-24 16:12:06,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=538930.0, ans=0.1
2024-09-24 16:12:10,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=538976.6666666666, ans=0.1
2024-09-24 16:12:15,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.60 vs. limit=22.5
2024-09-24 16:12:25,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=539023.3333333334, ans=0.125
2024-09-24 16:12:27,915 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.55 vs. limit=15.0
2024-09-24 16:13:05,299 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=539116.6666666666, ans=0.07
2024-09-24 16:13:17,973 INFO [train.py:1198] (1/4) Epoch 30, batch 2550, loss[loss=0.2143, ctc_loss=0.1398, cr_loss=0.3724, over 17025.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1309, cr_loss=0.3488, over 3363500.89 frames. ], batch size: 56, lr: 3.96e-03, grad_scale: 32.0
2024-09-24 16:13:24,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=539163.3333333334, ans=10.0
2024-09-24 16:13:26,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=539163.3333333334, ans=0.125
2024-09-24 16:13:29,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=539163.3333333334, ans=0.125
2024-09-24 16:13:42,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=539210.0, ans=0.1
2024-09-24 16:13:42,919 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.02 vs. limit=12.0
2024-09-24 16:13:53,195 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.247e+02 1.346e+02 1.482e+02 2.162e+02, threshold=2.691e+02, percent-clipped=0.0
2024-09-24 16:14:13,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=539303.3333333334, ans=0.025
2024-09-24 16:14:30,525 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.72 vs. limit=15.0
2024-09-24 16:14:35,060 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.35 vs. limit=22.5
2024-09-24 16:14:40,187 INFO [train.py:1198] (1/4) Epoch 30, batch 2600, loss[loss=0.1978, ctc_loss=0.1287, cr_loss=0.3452, over 17017.00 frames. ], tot_loss[loss=0.202, ctc_loss=0.1319, cr_loss=0.3505, over 3349814.55 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:14:53,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.19 vs. limit=22.5
2024-09-24 16:15:37,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=539536.6666666666, ans=0.125
2024-09-24 16:15:44,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=539583.3333333334, ans=0.025
2024-09-24 16:15:46,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=539583.3333333334, ans=0.1
2024-09-24 16:15:48,257 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.41 vs. limit=15.0
2024-09-24 16:16:00,286 INFO [train.py:1198] (1/4) Epoch 30, batch 2650, loss[loss=0.1819, ctc_loss=0.1177, cr_loss=0.3211, over 16716.00 frames. ], tot_loss[loss=0.2022, ctc_loss=0.132, cr_loss=0.3507, over 3350222.04 frames. ], batch size: 37, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:16:30,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=539676.6666666666, ans=0.1
2024-09-24 16:16:42,933 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.284e+02 1.354e+02 1.501e+02 2.171e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-24 16:16:52,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=539770.0, ans=0.0
2024-09-24 16:16:54,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=539770.0, ans=0.0
2024-09-24 16:17:28,543 INFO [train.py:1198] (1/4) Epoch 30, batch 2700, loss[loss=0.199, ctc_loss=0.1299, cr_loss=0.3453, over 16997.00 frames. ], tot_loss[loss=0.2013, ctc_loss=0.1315, cr_loss=0.3493, over 3351435.42 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:17:35,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=539863.3333333334, ans=0.2
2024-09-24 16:17:50,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=539910.0, ans=0.125
2024-09-24 16:17:50,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=539910.0, ans=0.125
2024-09-24 16:17:57,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=539910.0, ans=0.125
2024-09-24 16:18:10,311 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=15.47 vs. limit=15.0
2024-09-24 16:18:31,236 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=9.02 vs. limit=15.0
2024-09-24 16:18:48,019 INFO [train.py:1198] (1/4) Epoch 30, batch 2750, loss[loss=0.2004, ctc_loss=0.1308, cr_loss=0.3478, over 17190.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1317, cr_loss=0.35, over 3356611.71 frames. ], batch size: 41, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:19:17,650 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=7.36 vs. limit=15.0
2024-09-24 16:19:26,136 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.254e+02 1.332e+02 1.454e+02 2.287e+02, threshold=2.664e+02, percent-clipped=0.0
2024-09-24 16:19:55,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=540283.3333333334, ans=0.125
2024-09-24 16:20:10,880 INFO [train.py:1198] (1/4) Epoch 30, batch 2800, loss[loss=0.2219, ctc_loss=0.1459, cr_loss=0.3796, over 17006.00 frames. ], tot_loss[loss=0.2017, ctc_loss=0.1318, cr_loss=0.3495, over 3353615.71 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:20:11,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=540330.0, ans=0.125
2024-09-24 16:20:12,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=540330.0, ans=0.07
2024-09-24 16:20:19,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=540330.0, ans=0.0
2024-09-24 16:21:36,740 INFO [train.py:1198] (1/4) Epoch 30, batch 2850, loss[loss=0.1917, ctc_loss=0.1233, cr_loss=0.3419, over 17020.00 frames. ], tot_loss[loss=0.2018, ctc_loss=0.132, cr_loss=0.3491, over 3340119.78 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:22:00,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.48 vs. limit=12.0
2024-09-24 16:22:12,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=540656.6666666666, ans=0.5
2024-09-24 16:22:15,505 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.298e+02 1.387e+02 1.478e+02 2.436e+02, threshold=2.774e+02, percent-clipped=0.0
2024-09-24 16:22:15,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=540656.6666666666, ans=0.125
2024-09-24 16:22:23,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=540656.6666666666, ans=0.125
2024-09-24 16:22:23,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=540656.6666666666, ans=0.125
2024-09-24 16:22:26,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=540703.3333333334, ans=0.125
2024-09-24 16:22:30,598 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.90 vs. limit=15.0
2024-09-24 16:22:34,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=540703.3333333334, ans=0.1
2024-09-24 16:22:52,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=540750.0, ans=0.1
2024-09-24 16:22:58,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=540796.6666666666, ans=0.04949747468305833
2024-09-24 16:23:00,142 INFO [train.py:1198] (1/4) Epoch 30, batch 2900, loss[loss=0.2309, ctc_loss=0.1496, cr_loss=0.4065, over 16981.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1324, cr_loss=0.3507, over 3349489.31 frames. ], batch size: 56, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:23:52,681 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.70 vs. limit=15.0
2024-09-24 16:24:20,582 INFO [train.py:1198] (1/4) Epoch 30, batch 2950, loss[loss=0.1934, ctc_loss=0.1265, cr_loss=0.3345, over 17285.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1325, cr_loss=0.3508, over 3357686.29 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:24:23,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=541030.0, ans=0.0
2024-09-24 16:24:48,361 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.45 vs. limit=15.0
2024-09-24 16:24:55,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=541123.3333333334, ans=0.09899494936611666
2024-09-24 16:25:01,512 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.276e+02 1.368e+02 1.488e+02 1.810e+02, threshold=2.735e+02, percent-clipped=0.0
2024-09-24 16:25:17,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=541170.0, ans=0.125
2024-09-24 16:25:30,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=541216.6666666666, ans=0.0
2024-09-24 16:25:31,965 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.00 vs. limit=6.0
2024-09-24 16:25:37,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=541216.6666666666, ans=0.125
2024-09-24 16:25:42,514 INFO [train.py:1198] (1/4) Epoch 30, batch 3000, loss[loss=0.2089, ctc_loss=0.1366, cr_loss=0.3617, over 17012.00 frames. ], tot_loss[loss=0.2026, ctc_loss=0.1324, cr_loss=0.3511, over 3364832.15 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:25:42,515 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 16:25:51,122 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.4.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.3126, 4.9240, 4.9182, 4.6055], device='cuda:1')
2024-09-24 16:25:57,426 INFO [train.py:1230] (1/4) Epoch 30, validation: loss=0.03649, ctc_loss=0.03649, cr_loss=8.522e-15, over 944034.00 frames. 
2024-09-24 16:25:57,426 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 16:26:05,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.min_positive, batch_count=541263.3333333334, ans=0.025
2024-09-24 16:26:11,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=541310.0, ans=0.125
2024-09-24 16:26:30,041 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.18 vs. limit=15.0
2024-09-24 16:26:34,355 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.55 vs. limit=15.0
2024-09-24 16:26:45,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=541356.6666666666, ans=0.0
2024-09-24 16:26:48,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=541403.3333333334, ans=0.125
2024-09-24 16:26:49,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=541403.3333333334, ans=0.125
2024-09-24 16:26:51,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=541403.3333333334, ans=0.125
2024-09-24 16:26:54,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=541403.3333333334, ans=0.0
2024-09-24 16:26:59,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=541403.3333333334, ans=0.0
2024-09-24 16:27:11,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=541450.0, ans=0.1
2024-09-24 16:27:23,552 INFO [train.py:1198] (1/4) Epoch 30, batch 3050, loss[loss=0.1826, ctc_loss=0.1166, cr_loss=0.3301, over 17088.00 frames. ], tot_loss[loss=0.2015, ctc_loss=0.1316, cr_loss=0.3494, over 3370702.69 frames. ], batch size: 49, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:27:23,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=541496.6666666666, ans=0.125
2024-09-24 16:27:30,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=541496.6666666666, ans=0.025
2024-09-24 16:28:00,734 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.275e+02 1.345e+02 1.423e+02 1.790e+02, threshold=2.690e+02, percent-clipped=0.0
2024-09-24 16:28:00,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=541590.0, ans=0.1
2024-09-24 16:28:08,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=541636.6666666666, ans=0.1
2024-09-24 16:28:25,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=541683.3333333334, ans=0.125
2024-09-24 16:28:38,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=541683.3333333334, ans=0.1
2024-09-24 16:28:41,094 INFO [train.py:1198] (1/4) Epoch 30, batch 3100, loss[loss=0.2145, ctc_loss=0.1401, cr_loss=0.3722, over 17002.00 frames. ], tot_loss[loss=0.2005, ctc_loss=0.1309, cr_loss=0.348, over 3364276.03 frames. ], batch size: 53, lr: 3.96e-03, grad_scale: 16.0
2024-09-24 16:28:52,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=541730.0, ans=0.125
2024-09-24 16:29:09,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.86 vs. limit=22.5
2024-09-24 16:29:15,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=541823.3333333334, ans=0.2
2024-09-24 16:29:24,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=541823.3333333334, ans=0.1
2024-09-24 16:29:52,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=541916.6666666666, ans=0.1
2024-09-24 16:29:52,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=541916.6666666666, ans=0.125
2024-09-24 16:30:01,866 INFO [train.py:1198] (1/4) Epoch 30, batch 3150, loss[loss=0.2212, ctc_loss=0.1468, cr_loss=0.3718, over 14767.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1306, cr_loss=0.3475, over 3363042.86 frames. ], batch size: 89, lr: 3.95e-03, grad_scale: 16.0
2024-09-24 16:30:05,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=541963.3333333334, ans=0.2
2024-09-24 16:30:12,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=541963.3333333334, ans=0.5
2024-09-24 16:30:29,186 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.84 vs. limit=10.0
2024-09-24 16:30:39,238 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.276e+02 1.352e+02 1.455e+02 1.845e+02, threshold=2.704e+02, percent-clipped=0.0
2024-09-24 16:30:44,929 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.75 vs. limit=15.0
2024-09-24 16:30:47,887 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.21 vs. limit=10.0
2024-09-24 16:31:03,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=542150.0, ans=0.2
2024-09-24 16:31:20,116 INFO [train.py:1198] (1/4) Epoch 30, batch 3200, loss[loss=0.2042, ctc_loss=0.1322, cr_loss=0.36, over 17344.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1309, cr_loss=0.3487, over 3363221.00 frames. ], batch size: 48, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:31:20,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=542196.6666666666, ans=10.0
2024-09-24 16:31:39,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=542243.3333333334, ans=0.125
2024-09-24 16:31:47,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=542243.3333333334, ans=0.0
2024-09-24 16:32:00,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=542290.0, ans=0.125
2024-09-24 16:32:04,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=542290.0, ans=15.0
2024-09-24 16:32:05,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=542336.6666666666, ans=0.125
2024-09-24 16:32:05,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=542336.6666666666, ans=0.2
2024-09-24 16:32:15,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=542336.6666666666, ans=0.125
2024-09-24 16:32:38,559 INFO [train.py:1198] (1/4) Epoch 30, batch 3250, loss[loss=0.2051, ctc_loss=0.1352, cr_loss=0.3495, over 16944.00 frames. ], tot_loss[loss=0.2003, ctc_loss=0.1307, cr_loss=0.3484, over 3361227.99 frames. ], batch size: 58, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:33:03,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=542476.6666666666, ans=0.125
2024-09-24 16:33:13,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=542523.3333333334, ans=0.025
2024-09-24 16:33:16,569 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.257e+02 1.346e+02 1.457e+02 3.617e+02, threshold=2.692e+02, percent-clipped=1.0
2024-09-24 16:33:49,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=542616.6666666666, ans=0.0
2024-09-24 16:33:59,504 INFO [train.py:1198] (1/4) Epoch 30, batch 3300, loss[loss=0.2424, ctc_loss=0.1621, cr_loss=0.4017, over 17038.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1299, cr_loss=0.3478, over 3362824.62 frames. ], batch size: 52, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:34:18,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=542710.0, ans=0.2
2024-09-24 16:34:42,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=542756.6666666666, ans=0.2
2024-09-24 16:34:45,715 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.02 vs. limit=15.0
2024-09-24 16:35:17,843 INFO [train.py:1198] (1/4) Epoch 30, batch 3350, loss[loss=0.214, ctc_loss=0.1392, cr_loss=0.3742, over 17000.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.13, cr_loss=0.3474, over 3361076.76 frames. ], batch size: 51, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:35:22,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=542896.6666666666, ans=0.0
2024-09-24 16:35:24,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=542896.6666666666, ans=0.125
2024-09-24 16:35:38,867 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.56 vs. limit=12.0
2024-09-24 16:35:46,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=542943.3333333334, ans=0.125
2024-09-24 16:35:46,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=542943.3333333334, ans=0.0
2024-09-24 16:35:55,301 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.300e+02 1.382e+02 1.466e+02 2.312e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-24 16:36:18,040 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.13 vs. limit=22.5
2024-09-24 16:36:25,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=543083.3333333334, ans=0.0
2024-09-24 16:36:33,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=543083.3333333334, ans=0.125
2024-09-24 16:36:36,356 INFO [train.py:1198] (1/4) Epoch 30, batch 3400, loss[loss=0.1642, ctc_loss=0.1041, cr_loss=0.3001, over 17086.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.13, cr_loss=0.3474, over 3364779.70 frames. ], batch size: 43, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:36:36,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=543130.0, ans=0.1
2024-09-24 16:36:42,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=543130.0, ans=0.0
2024-09-24 16:36:44,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=543130.0, ans=0.0
2024-09-24 16:37:47,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=543316.6666666666, ans=0.125
2024-09-24 16:37:56,612 INFO [train.py:1198] (1/4) Epoch 30, batch 3450, loss[loss=0.2244, ctc_loss=0.1458, cr_loss=0.3927, over 16646.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1311, cr_loss=0.3492, over 3364574.90 frames. ], batch size: 66, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:37:56,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=543363.3333333334, ans=0.2
2024-09-24 16:38:08,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=543363.3333333334, ans=15.0
2024-09-24 16:38:18,074 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.27 vs. limit=22.5
2024-09-24 16:38:31,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=543456.6666666666, ans=0.125
2024-09-24 16:38:35,837 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.271e+02 1.366e+02 1.468e+02 2.089e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-24 16:39:16,716 INFO [train.py:1198] (1/4) Epoch 30, batch 3500, loss[loss=0.1754, ctc_loss=0.1149, cr_loss=0.3023, over 17241.00 frames. ], tot_loss[loss=0.1998, ctc_loss=0.1303, cr_loss=0.3473, over 3363741.78 frames. ], batch size: 44, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:39:42,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=543643.3333333334, ans=0.125
2024-09-24 16:39:42,418 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.82 vs. limit=6.0
2024-09-24 16:39:48,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=543690.0, ans=0.125
2024-09-24 16:39:55,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=543690.0, ans=0.125
2024-09-24 16:40:09,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=543736.6666666666, ans=0.035
2024-09-24 16:40:19,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=543783.3333333334, ans=0.1
2024-09-24 16:40:36,961 INFO [train.py:1198] (1/4) Epoch 30, batch 3550, loss[loss=0.1936, ctc_loss=0.1269, cr_loss=0.3334, over 17300.00 frames. ], tot_loss[loss=0.1999, ctc_loss=0.1305, cr_loss=0.3471, over 3353257.53 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:41:14,200 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.268e+02 1.349e+02 1.434e+02 2.153e+02, threshold=2.697e+02, percent-clipped=0.0
2024-09-24 16:41:19,375 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 16:41:47,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=544016.6666666666, ans=0.035
2024-09-24 16:41:55,340 INFO [train.py:1198] (1/4) Epoch 30, batch 3600, loss[loss=0.1818, ctc_loss=0.1218, cr_loss=0.3002, over 17230.00 frames. ], tot_loss[loss=0.1998, ctc_loss=0.1304, cr_loss=0.3471, over 3360669.23 frames. ], batch size: 50, lr: 3.95e-03, grad_scale: 32.0
2024-09-24 16:42:04,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=544063.3333333334, ans=0.2
2024-09-24 16:42:06,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=544063.3333333334, ans=0.2
2024-09-24 16:42:09,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=544110.0, ans=0.125
2024-09-24 16:42:30,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=544156.6666666666, ans=0.1
2024-09-24 16:42:45,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=544203.3333333334, ans=0.1
2024-09-24 16:42:53,801 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.35 vs. limit=6.0
2024-09-24 16:43:00,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=544250.0, ans=0.0
2024-09-24 16:43:01,468 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.10 vs. limit=15.0
2024-09-24 16:43:03,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=544250.0, ans=0.1
2024-09-24 16:43:11,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=544296.6666666666, ans=0.0
2024-09-24 16:43:13,172 INFO [train.py:1198] (1/4) Epoch 30, batch 3650, loss[loss=0.2589, ctc_loss=0.182, cr_loss=0.3843, over 11911.00 frames. ], tot_loss[loss=0.2013, ctc_loss=0.1314, cr_loss=0.3494, over 3354431.45 frames. ], batch size: 123, lr: 3.95e-03, grad_scale: 16.0
2024-09-24 16:43:16,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=544296.6666666666, ans=0.5
2024-09-24 16:43:21,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.90 vs. limit=22.5
2024-09-24 16:43:54,824 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.277e+02 1.338e+02 1.462e+02 2.145e+02, threshold=2.677e+02, percent-clipped=0.0
2024-09-24 16:44:14,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=544436.6666666666, ans=0.0
2024-09-24 16:44:35,114 INFO [train.py:1198] (1/4) Epoch 30, batch 3700, loss[loss=0.2279, ctc_loss=0.1494, cr_loss=0.3923, over 16554.00 frames. ], tot_loss[loss=0.2018, ctc_loss=0.1319, cr_loss=0.3496, over 3348116.90 frames. ], batch size: 66, lr: 3.95e-03, grad_scale: 8.0
2024-09-24 16:44:54,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=544576.6666666666, ans=0.1
2024-09-24 16:45:05,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=544623.3333333334, ans=0.0
2024-09-24 16:45:24,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=544670.0, ans=0.125
2024-09-24 16:45:53,702 INFO [train.py:1198] (1/4) Epoch 30, batch 3750, loss[loss=0.1438, ctc_loss=0.09115, cr_loss=0.2632, over 17263.00 frames. ], tot_loss[loss=0.2005, ctc_loss=0.131, cr_loss=0.3476, over 3338964.60 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 8.0
2024-09-24 16:46:08,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=544810.0, ans=0.125
2024-09-24 16:46:09,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=544810.0, ans=0.125
2024-09-24 16:46:11,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=544810.0, ans=0.125
2024-09-24 16:46:23,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=544856.6666666666, ans=0.1
2024-09-24 16:46:29,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=544856.6666666666, ans=0.1
2024-09-24 16:46:35,276 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.265e+02 1.357e+02 1.446e+02 2.507e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-24 16:47:12,372 INFO [train.py:1198] (1/4) Epoch 30, batch 3800, loss[loss=0.2205, ctc_loss=0.1451, cr_loss=0.3769, over 16558.00 frames. ], tot_loss[loss=0.2011, ctc_loss=0.1313, cr_loss=0.3488, over 3342204.37 frames. ], batch size: 66, lr: 3.94e-03, grad_scale: 8.0
2024-09-24 16:47:15,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=544996.6666666666, ans=0.125
2024-09-24 16:48:00,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=545136.6666666666, ans=0.125
2024-09-24 16:48:31,293 INFO [train.py:1198] (1/4) Epoch 30, batch 3850, loss[loss=0.2222, ctc_loss=0.1467, cr_loss=0.3774, over 16932.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.1341, cr_loss=0.3526, over 3295636.14 frames. ], batch size: 58, lr: 3.94e-03, grad_scale: 8.0
2024-09-24 16:48:45,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=545276.6666666666, ans=0.125
2024-09-24 16:49:01,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=545323.3333333334, ans=0.1
2024-09-24 16:49:02,913 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=11.81 vs. limit=15.0
2024-09-24 16:49:05,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=545323.3333333334, ans=0.0
2024-09-24 16:49:11,169 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.380e+02 1.551e+02 1.671e+02 2.623e+02, threshold=3.102e+02, percent-clipped=0.0
2024-09-24 16:49:13,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=14.75 vs. limit=15.0
2024-09-24 16:49:25,466 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.05 vs. limit=6.0
2024-09-24 16:49:30,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.max_abs, batch_count=545416.6666666666, ans=10.0
2024-09-24 16:49:38,690 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.74 vs. limit=15.0
2024-09-24 16:50:31,769 INFO [train.py:1198] (1/4) Epoch 31, batch 0, loss[loss=0.1766, ctc_loss=0.1161, cr_loss=0.3023, over 17269.00 frames. ], tot_loss[loss=0.1766, ctc_loss=0.1161, cr_loss=0.3023, over 17269.00 frames. ], batch size: 42, lr: 3.88e-03, grad_scale: 16.0
2024-09-24 16:50:31,770 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 16:50:44,466 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.4.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([2.8786, 3.7927, 3.7139, 3.1040], device='cuda:1')
2024-09-24 16:50:45,712 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.1.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([4.8675, 4.7674, 4.3924, 4.8525], device='cuda:1')
2024-09-24 16:50:47,107 INFO [train.py:1230] (1/4) Epoch 31, validation: loss=0.03594, ctc_loss=0.03594, cr_loss=9.065e-15, over 944034.00 frames. 
2024-09-24 16:50:47,107 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 16:50:50,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=545444.6666666666, ans=0.125
2024-09-24 16:50:55,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=545444.6666666666, ans=0.125
2024-09-24 16:50:56,035 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.83 vs. limit=10.0
2024-09-24 16:51:27,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=545538.0, ans=0.05
2024-09-24 16:51:31,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=545538.0, ans=0.0
2024-09-24 16:51:40,059 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 16:51:41,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=545584.6666666666, ans=0.125
2024-09-24 16:51:48,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=545584.6666666666, ans=0.125
2024-09-24 16:51:56,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=545631.3333333334, ans=0.1
2024-09-24 16:51:58,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=545631.3333333334, ans=0.125
2024-09-24 16:52:06,491 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.20 vs. limit=12.0
2024-09-24 16:52:12,148 INFO [train.py:1198] (1/4) Epoch 31, batch 50, loss[loss=0.2103, ctc_loss=0.1364, cr_loss=0.3695, over 16997.00 frames. ], tot_loss[loss=0.2032, ctc_loss=0.1329, cr_loss=0.3515, over 747709.42 frames. ], batch size: 56, lr: 3.88e-03, grad_scale: 16.0
2024-09-24 16:52:28,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=545724.6666666666, ans=0.125
2024-09-24 16:52:31,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=545724.6666666666, ans=0.0
2024-09-24 16:52:55,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=545771.3333333334, ans=0.0
2024-09-24 16:53:01,962 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.275e+02 1.374e+02 1.534e+02 1.966e+02, threshold=2.749e+02, percent-clipped=0.0
2024-09-24 16:53:12,214 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.02 vs. limit=10.0
2024-09-24 16:53:21,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten.whitening_limit, batch_count=545864.6666666666, ans=15.0
2024-09-24 16:53:34,044 INFO [train.py:1198] (1/4) Epoch 31, batch 100, loss[loss=0.2128, ctc_loss=0.1409, cr_loss=0.3597, over 16806.00 frames. ], tot_loss[loss=0.2046, ctc_loss=0.134, cr_loss=0.3527, over 1309437.00 frames. ], batch size: 58, lr: 3.87e-03, grad_scale: 16.0
2024-09-24 16:53:54,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=545958.0, ans=0.1
2024-09-24 16:54:02,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=545958.0, ans=0.125
2024-09-24 16:54:17,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=546004.6666666666, ans=0.125
2024-09-24 16:54:19,147 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 16:54:20,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=546051.3333333334, ans=0.2
2024-09-24 16:54:49,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.22 vs. limit=15.0
2024-09-24 16:54:56,692 INFO [train.py:1198] (1/4) Epoch 31, batch 150, loss[loss=0.186, ctc_loss=0.1227, cr_loss=0.3162, over 17097.00 frames. ], tot_loss[loss=0.2023, ctc_loss=0.1323, cr_loss=0.35, over 1758915.47 frames. ], batch size: 49, lr: 3.87e-03, grad_scale: 16.0
2024-09-24 16:55:08,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=546144.6666666666, ans=0.125
2024-09-24 16:55:13,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=546191.3333333334, ans=0.2
2024-09-24 16:55:18,364 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.43 vs. limit=15.0
2024-09-24 16:55:29,523 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.80 vs. limit=15.0
2024-09-24 16:55:44,680 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.256e+02 1.336e+02 1.452e+02 1.976e+02, threshold=2.673e+02, percent-clipped=0.0
2024-09-24 16:56:16,051 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.95 vs. limit=15.0
2024-09-24 16:56:16,760 INFO [train.py:1198] (1/4) Epoch 31, batch 200, loss[loss=0.1975, ctc_loss=0.1316, cr_loss=0.3295, over 17275.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1302, cr_loss=0.3461, over 2109755.75 frames. ], batch size: 46, lr: 3.87e-03, grad_scale: 16.0
2024-09-24 16:56:38,274 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.66 vs. limit=10.0
2024-09-24 16:56:39,435 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.29 vs. limit=12.0
2024-09-24 16:56:48,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=546424.6666666666, ans=0.125
2024-09-24 16:57:27,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=546564.6666666666, ans=0.0
2024-09-24 16:57:31,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=546564.6666666666, ans=0.125
2024-09-24 16:57:44,192 INFO [train.py:1198] (1/4) Epoch 31, batch 250, loss[loss=0.1742, ctc_loss=0.1101, cr_loss=0.3209, over 17069.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1298, cr_loss=0.3454, over 2378401.83 frames. ], batch size: 46, lr: 3.87e-03, grad_scale: 16.0
2024-09-24 16:57:44,614 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 16:57:44,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=546611.3333333334, ans=0.0
2024-09-24 16:58:14,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=546704.6666666666, ans=0.125
2024-09-24 16:58:31,699 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.278e+02 1.364e+02 1.509e+02 2.036e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-24 16:58:46,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=546798.0, ans=0.025
2024-09-24 16:58:46,486 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 16:58:49,866 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.80 vs. limit=6.0
2024-09-24 16:59:03,343 INFO [train.py:1198] (1/4) Epoch 31, batch 300, loss[loss=0.1721, ctc_loss=0.1114, cr_loss=0.3034, over 17030.00 frames. ], tot_loss[loss=0.1993, ctc_loss=0.13, cr_loss=0.3467, over 2603112.63 frames. ], batch size: 44, lr: 3.87e-03, grad_scale: 16.0
2024-09-24 16:59:22,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=546891.3333333334, ans=0.125
2024-09-24 16:59:27,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=546891.3333333334, ans=0.125
2024-09-24 16:59:33,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=546891.3333333334, ans=0.1
2024-09-24 16:59:35,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=546891.3333333334, ans=0.0
2024-09-24 16:59:48,696 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.59 vs. limit=22.5
2024-09-24 17:00:26,013 INFO [train.py:1198] (1/4) Epoch 31, batch 350, loss[loss=0.2259, ctc_loss=0.1496, cr_loss=0.3816, over 16914.00 frames. ], tot_loss[loss=0.2002, ctc_loss=0.1307, cr_loss=0.3476, over 2770166.53 frames. ], batch size: 58, lr: 3.87e-03, grad_scale: 16.0
2024-09-24 17:00:35,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=547078.0, ans=0.1
2024-09-24 17:00:42,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=547124.6666666666, ans=0.0
2024-09-24 17:00:56,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=547171.3333333334, ans=0.0
2024-09-24 17:01:07,007 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.79 vs. limit=22.5
2024-09-24 17:01:14,094 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.224e+02 1.318e+02 1.423e+02 1.795e+02, threshold=2.635e+02, percent-clipped=0.0
2024-09-24 17:01:25,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=547218.0, ans=0.1
2024-09-24 17:01:28,659 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.53 vs. limit=12.0
2024-09-24 17:01:39,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=547264.6666666666, ans=0.025
2024-09-24 17:01:51,999 INFO [train.py:1198] (1/4) Epoch 31, batch 400, loss[loss=0.1809, ctc_loss=0.1206, cr_loss=0.3017, over 17335.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1297, cr_loss=0.3469, over 2908424.70 frames. ], batch size: 52, lr: 3.87e-03, grad_scale: 32.0
2024-09-24 17:02:05,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=547311.3333333334, ans=0.2
2024-09-24 17:02:05,414 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.18 vs. limit=22.5
2024-09-24 17:02:16,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=547358.0, ans=0.125
2024-09-24 17:03:14,668 INFO [train.py:1198] (1/4) Epoch 31, batch 450, loss[loss=0.1791, ctc_loss=0.1169, cr_loss=0.3109, over 17094.00 frames. ], tot_loss[loss=0.1996, ctc_loss=0.1302, cr_loss=0.3469, over 3001369.17 frames. ], batch size: 49, lr: 3.87e-03, grad_scale: 32.0
2024-09-24 17:03:40,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=547591.3333333334, ans=0.125
2024-09-24 17:04:00,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=547638.0, ans=0.1
2024-09-24 17:04:02,912 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.314e+02 1.420e+02 1.522e+02 1.945e+02, threshold=2.840e+02, percent-clipped=0.0
2024-09-24 17:04:31,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=547731.3333333334, ans=0.125
2024-09-24 17:04:37,644 INFO [train.py:1198] (1/4) Epoch 31, batch 500, loss[loss=0.1916, ctc_loss=0.1233, cr_loss=0.3414, over 17167.00 frames. ], tot_loss[loss=0.1985, ctc_loss=0.1294, cr_loss=0.3455, over 3083105.34 frames. ], batch size: 45, lr: 3.87e-03, grad_scale: 32.0
2024-09-24 17:05:02,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=547824.6666666666, ans=0.0
2024-09-24 17:05:10,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.min_positive, batch_count=547871.3333333334, ans=0.025
2024-09-24 17:05:22,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=547871.3333333334, ans=0.125
2024-09-24 17:05:27,441 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.45 vs. limit=15.0
2024-09-24 17:05:32,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=547918.0, ans=0.125
2024-09-24 17:05:39,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=547918.0, ans=0.1
2024-09-24 17:05:58,745 INFO [train.py:1198] (1/4) Epoch 31, batch 550, loss[loss=0.1954, ctc_loss=0.1263, cr_loss=0.3451, over 17303.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1298, cr_loss=0.3462, over 3140643.50 frames. ], batch size: 49, lr: 3.87e-03, grad_scale: 32.0
2024-09-24 17:06:10,840 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.56 vs. limit=15.0
2024-09-24 17:06:35,428 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:06:43,117 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:06:52,446 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.269e+02 1.346e+02 1.471e+02 2.079e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-24 17:06:59,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.59 vs. limit=12.0
2024-09-24 17:07:05,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=548151.3333333334, ans=0.0
2024-09-24 17:07:08,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=548198.0, ans=0.0
2024-09-24 17:07:21,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=548198.0, ans=0.025
2024-09-24 17:07:23,920 INFO [train.py:1198] (1/4) Epoch 31, batch 600, loss[loss=0.2031, ctc_loss=0.1303, cr_loss=0.3638, over 17369.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.13, cr_loss=0.3469, over 3188178.55 frames. ], batch size: 48, lr: 3.87e-03, grad_scale: 32.0
2024-09-24 17:07:37,075 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.12 vs. limit=15.0
2024-09-24 17:07:55,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=548291.3333333334, ans=0.0
2024-09-24 17:07:55,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=548291.3333333334, ans=0.0
2024-09-24 17:08:11,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=548338.0, ans=0.125
2024-09-24 17:08:44,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=548478.0, ans=0.125
2024-09-24 17:08:46,119 INFO [train.py:1198] (1/4) Epoch 31, batch 650, loss[loss=0.2071, ctc_loss=0.1343, cr_loss=0.364, over 17303.00 frames. ], tot_loss[loss=0.1999, ctc_loss=0.1303, cr_loss=0.3477, over 3229914.18 frames. ], batch size: 49, lr: 3.87e-03, grad_scale: 32.0
2024-09-24 17:09:10,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=548524.6666666666, ans=0.1
2024-09-24 17:09:16,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=548571.3333333334, ans=0.125
2024-09-24 17:09:21,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=548571.3333333334, ans=0.125
2024-09-24 17:09:27,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=548571.3333333334, ans=10.0
2024-09-24 17:09:27,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=548571.3333333334, ans=0.125
2024-09-24 17:09:35,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=548618.0, ans=0.1
2024-09-24 17:09:36,766 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.261e+02 1.352e+02 1.465e+02 2.749e+02, threshold=2.705e+02, percent-clipped=1.0
2024-09-24 17:09:45,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=548618.0, ans=0.1
2024-09-24 17:09:50,122 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.17 vs. limit=15.0
2024-09-24 17:10:08,669 INFO [train.py:1198] (1/4) Epoch 31, batch 700, loss[loss=0.2557, ctc_loss=0.175, cr_loss=0.4035, over 15304.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1311, cr_loss=0.3481, over 3252364.48 frames. ], batch size: 89, lr: 3.86e-03, grad_scale: 32.0
2024-09-24 17:10:15,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=548711.3333333334, ans=0.1
2024-09-24 17:10:34,654 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.15 vs. limit=6.0
2024-09-24 17:11:31,808 INFO [train.py:1198] (1/4) Epoch 31, batch 750, loss[loss=0.2099, ctc_loss=0.137, cr_loss=0.3643, over 17072.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1312, cr_loss=0.3485, over 3272892.91 frames. ], batch size: 46, lr: 3.86e-03, grad_scale: 32.0
2024-09-24 17:11:57,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=548991.3333333334, ans=0.0
2024-09-24 17:12:21,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=549084.6666666666, ans=0.125
2024-09-24 17:12:25,358 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.270e+02 1.339e+02 1.427e+02 2.075e+02, threshold=2.677e+02, percent-clipped=0.0
2024-09-24 17:12:43,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=549131.3333333334, ans=0.125
2024-09-24 17:12:54,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=549131.3333333334, ans=0.035
2024-09-24 17:12:57,346 INFO [train.py:1198] (1/4) Epoch 31, batch 800, loss[loss=0.2302, ctc_loss=0.1574, cr_loss=0.364, over 11137.00 frames. ], tot_loss[loss=0.2018, ctc_loss=0.1318, cr_loss=0.3499, over 3279992.89 frames. ], batch size: 123, lr: 3.86e-03, grad_scale: 32.0
2024-09-24 17:13:05,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=549178.0, ans=0.0
2024-09-24 17:13:12,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=549224.6666666666, ans=0.0
2024-09-24 17:13:13,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=549224.6666666666, ans=0.1
2024-09-24 17:13:46,257 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.00 vs. limit=15.0
2024-09-24 17:14:03,232 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:14:12,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=549364.6666666666, ans=0.125
2024-09-24 17:14:17,350 INFO [train.py:1198] (1/4) Epoch 31, batch 850, loss[loss=0.2005, ctc_loss=0.1296, cr_loss=0.3545, over 17018.00 frames. ], tot_loss[loss=0.2011, ctc_loss=0.1313, cr_loss=0.3491, over 3302841.67 frames. ], batch size: 44, lr: 3.86e-03, grad_scale: 32.0
2024-09-24 17:14:24,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.72 vs. limit=22.5
2024-09-24 17:14:39,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=549458.0, ans=0.125
2024-09-24 17:14:42,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=549458.0, ans=0.125
2024-09-24 17:14:45,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=549458.0, ans=0.1
2024-09-24 17:15:09,432 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.257e+02 1.348e+02 1.447e+02 2.367e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-24 17:15:19,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=549551.3333333334, ans=0.125
2024-09-24 17:15:23,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=549598.0, ans=0.125
2024-09-24 17:15:33,632 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:15:39,448 INFO [train.py:1198] (1/4) Epoch 31, batch 900, loss[loss=0.2035, ctc_loss=0.1302, cr_loss=0.3664, over 17201.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1311, cr_loss=0.3492, over 3321465.60 frames. ], batch size: 55, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:15:47,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=549644.6666666666, ans=0.0
2024-09-24 17:15:50,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.80 vs. limit=12.0
2024-09-24 17:16:01,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.24 vs. limit=15.0
2024-09-24 17:16:46,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=549784.6666666666, ans=0.0
2024-09-24 17:16:57,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=549831.3333333334, ans=0.0
2024-09-24 17:17:02,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.29 vs. limit=22.5
2024-09-24 17:17:05,204 INFO [train.py:1198] (1/4) Epoch 31, batch 950, loss[loss=0.1543, ctc_loss=0.09589, cr_loss=0.2921, over 17079.00 frames. ], tot_loss[loss=0.2011, ctc_loss=0.1311, cr_loss=0.3498, over 3321265.17 frames. ], batch size: 43, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:17:12,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=549878.0, ans=0.125
2024-09-24 17:17:15,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=549878.0, ans=0.0
2024-09-24 17:17:44,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=549971.3333333334, ans=0.04949747468305833
2024-09-24 17:17:54,490 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.20 vs. limit=22.5
2024-09-24 17:17:57,027 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.281e+02 1.359e+02 1.441e+02 1.895e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-24 17:18:27,477 INFO [train.py:1198] (1/4) Epoch 31, batch 1000, loss[loss=0.2196, ctc_loss=0.1458, cr_loss=0.3693, over 16543.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1309, cr_loss=0.3498, over 3334528.09 frames. ], batch size: 66, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:19:15,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=550251.3333333334, ans=0.125
2024-09-24 17:19:45,880 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.75 vs. limit=15.0
2024-09-24 17:19:50,156 INFO [train.py:1198] (1/4) Epoch 31, batch 1050, loss[loss=0.1817, ctc_loss=0.1162, cr_loss=0.3272, over 17039.00 frames. ], tot_loss[loss=0.2002, ctc_loss=0.1303, cr_loss=0.349, over 3346583.23 frames. ], batch size: 39, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:20:06,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=550391.3333333334, ans=0.2
2024-09-24 17:20:24,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=550438.0, ans=0.2
2024-09-24 17:20:30,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=550438.0, ans=0.0
2024-09-24 17:20:39,886 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.258e+02 1.325e+02 1.421e+02 1.846e+02, threshold=2.651e+02, percent-clipped=0.0
2024-09-24 17:20:41,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=550484.6666666666, ans=0.07
2024-09-24 17:20:49,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=550484.6666666666, ans=0.025
2024-09-24 17:20:54,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=550531.3333333334, ans=0.125
2024-09-24 17:20:59,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=550531.3333333334, ans=0.125
2024-09-24 17:21:10,256 INFO [train.py:1198] (1/4) Epoch 31, batch 1100, loss[loss=0.1861, ctc_loss=0.1176, cr_loss=0.3426, over 17243.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1299, cr_loss=0.3475, over 3357929.41 frames. ], batch size: 44, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:21:18,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=550578.0, ans=0.0
2024-09-24 17:21:27,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=550624.6666666666, ans=0.0
2024-09-24 17:21:31,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.87 vs. limit=15.0
2024-09-24 17:22:04,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=550718.0, ans=0.025
2024-09-24 17:22:10,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=550718.0, ans=0.2
2024-09-24 17:22:17,570 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:22:39,182 INFO [train.py:1198] (1/4) Epoch 31, batch 1150, loss[loss=0.1654, ctc_loss=0.1047, cr_loss=0.3036, over 16667.00 frames. ], tot_loss[loss=0.1992, ctc_loss=0.1298, cr_loss=0.3473, over 3363213.52 frames. ], batch size: 37, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:22:41,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=550811.3333333334, ans=0.0
2024-09-24 17:22:42,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=550811.3333333334, ans=0.5
2024-09-24 17:23:28,709 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.270e+02 1.354e+02 1.493e+02 2.055e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-24 17:23:29,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=550951.3333333334, ans=0.0
2024-09-24 17:23:30,769 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:23:54,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=550998.0, ans=0.2
2024-09-24 17:23:56,774 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.90 vs. limit=22.5
2024-09-24 17:23:57,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=551044.6666666666, ans=0.0
2024-09-24 17:23:59,053 INFO [train.py:1198] (1/4) Epoch 31, batch 1200, loss[loss=0.1665, ctc_loss=0.1039, cr_loss=0.313, over 16325.00 frames. ], tot_loss[loss=0.1983, ctc_loss=0.129, cr_loss=0.3465, over 3367372.72 frames. ], batch size: 36, lr: 3.86e-03, grad_scale: 32.0
2024-09-24 17:24:39,203 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:24:58,708 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.19 vs. limit=15.0
2024-09-24 17:24:59,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=551184.6666666666, ans=0.125
2024-09-24 17:25:06,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=551231.3333333334, ans=0.125
2024-09-24 17:25:15,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=551231.3333333334, ans=0.0
2024-09-24 17:25:17,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=551231.3333333334, ans=0.2
2024-09-24 17:25:21,965 INFO [train.py:1198] (1/4) Epoch 31, batch 1250, loss[loss=0.2115, ctc_loss=0.1355, cr_loss=0.3802, over 17226.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1298, cr_loss=0.3477, over 3362220.12 frames. ], batch size: 50, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:25:43,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=551324.6666666666, ans=0.125
2024-09-24 17:25:43,564 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.82 vs. limit=10.0
2024-09-24 17:26:12,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=551418.0, ans=0.0
2024-09-24 17:26:13,266 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.296e+02 1.398e+02 1.491e+02 2.236e+02, threshold=2.796e+02, percent-clipped=0.0
2024-09-24 17:26:24,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=551418.0, ans=0.125
2024-09-24 17:26:46,959 INFO [train.py:1198] (1/4) Epoch 31, batch 1300, loss[loss=0.1718, ctc_loss=0.1112, cr_loss=0.3029, over 16289.00 frames. ], tot_loss[loss=0.1996, ctc_loss=0.13, cr_loss=0.3482, over 3359913.39 frames. ], batch size: 36, lr: 3.86e-03, grad_scale: 16.0
2024-09-24 17:27:13,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=551558.0, ans=0.09899494936611666
2024-09-24 17:27:22,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=551604.6666666666, ans=0.0
2024-09-24 17:27:33,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=551604.6666666666, ans=0.0
2024-09-24 17:27:54,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=551698.0, ans=0.1
2024-09-24 17:28:06,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=551698.0, ans=0.0
2024-09-24 17:28:09,625 INFO [train.py:1198] (1/4) Epoch 31, batch 1350, loss[loss=0.2108, ctc_loss=0.1349, cr_loss=0.3797, over 17200.00 frames. ], tot_loss[loss=0.1988, ctc_loss=0.1293, cr_loss=0.3472, over 3362661.48 frames. ], batch size: 47, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:28:29,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.max_positive, batch_count=551791.3333333334, ans=0.95
2024-09-24 17:28:40,854 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.32 vs. limit=15.0
2024-09-24 17:29:00,716 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.265e+02 1.360e+02 1.447e+02 1.964e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-24 17:29:17,523 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.34 vs. limit=15.0
2024-09-24 17:29:32,591 INFO [train.py:1198] (1/4) Epoch 31, batch 1400, loss[loss=0.2219, ctc_loss=0.1518, cr_loss=0.3506, over 14952.00 frames. ], tot_loss[loss=0.1997, ctc_loss=0.1301, cr_loss=0.3479, over 3362136.97 frames. ], batch size: 89, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:29:50,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=552024.6666666666, ans=0.0
2024-09-24 17:30:22,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=552118.0, ans=0.125
2024-09-24 17:30:24,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=552118.0, ans=0.125
2024-09-24 17:30:24,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=552118.0, ans=0.05
2024-09-24 17:30:24,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=552118.0, ans=0.0
2024-09-24 17:30:24,380 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.04 vs. limit=15.0
2024-09-24 17:30:26,001 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=7.32 vs. limit=12.0
2024-09-24 17:30:36,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=552164.6666666666, ans=0.2
2024-09-24 17:30:43,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=552164.6666666666, ans=0.125
2024-09-24 17:30:43,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=552164.6666666666, ans=0.1
2024-09-24 17:30:48,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=552164.6666666666, ans=0.125
2024-09-24 17:30:52,825 INFO [train.py:1198] (1/4) Epoch 31, batch 1450, loss[loss=0.205, ctc_loss=0.1361, cr_loss=0.3442, over 17232.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1301, cr_loss=0.3464, over 3351955.68 frames. ], batch size: 50, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:31:02,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=552211.3333333334, ans=0.125
2024-09-24 17:31:09,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=552258.0, ans=0.025
2024-09-24 17:31:25,127 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.05 vs. limit=10.0
2024-09-24 17:31:48,822 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.229e+02 1.325e+02 1.469e+02 3.189e+02, threshold=2.649e+02, percent-clipped=1.0
2024-09-24 17:31:57,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=552351.3333333334, ans=0.125
2024-09-24 17:32:19,916 INFO [train.py:1198] (1/4) Epoch 31, batch 1500, loss[loss=0.2145, ctc_loss=0.141, cr_loss=0.3675, over 16859.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.13, cr_loss=0.3468, over 3359013.20 frames. ], batch size: 58, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:32:33,676 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.12 vs. limit=22.5
2024-09-24 17:33:19,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten.whitening_limit, batch_count=552584.6666666666, ans=22.5
2024-09-24 17:33:27,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=552631.3333333334, ans=0.1
2024-09-24 17:33:39,665 INFO [train.py:1198] (1/4) Epoch 31, batch 1550, loss[loss=0.2251, ctc_loss=0.1475, cr_loss=0.3882, over 16987.00 frames. ], tot_loss[loss=0.2003, ctc_loss=0.1307, cr_loss=0.3481, over 3356400.22 frames. ], batch size: 56, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:34:16,000 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:34:30,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=552818.0, ans=0.0
2024-09-24 17:34:33,295 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.274e+02 1.344e+02 1.453e+02 2.165e+02, threshold=2.688e+02, percent-clipped=0.0
2024-09-24 17:34:57,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=552864.6666666666, ans=0.125
2024-09-24 17:34:59,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=552864.6666666666, ans=0.025
2024-09-24 17:35:02,239 INFO [train.py:1198] (1/4) Epoch 31, batch 1600, loss[loss=0.1411, ctc_loss=0.09057, cr_loss=0.2526, over 17029.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1297, cr_loss=0.3459, over 3353793.60 frames. ], batch size: 39, lr: 3.85e-03, grad_scale: 32.0
2024-09-24 17:35:04,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=552911.3333333334, ans=0.125
2024-09-24 17:35:44,173 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:35:55,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=553051.3333333334, ans=0.0
2024-09-24 17:36:02,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=553051.3333333334, ans=0.0
2024-09-24 17:36:25,120 INFO [train.py:1198] (1/4) Epoch 31, batch 1650, loss[loss=0.1649, ctc_loss=0.1046, cr_loss=0.3017, over 17256.00 frames. ], tot_loss[loss=0.1997, ctc_loss=0.1303, cr_loss=0.3471, over 3353220.39 frames. ], batch size: 42, lr: 3.85e-03, grad_scale: 32.0
2024-09-24 17:36:43,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=553191.3333333334, ans=0.0
2024-09-24 17:36:44,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=553191.3333333334, ans=0.0
2024-09-24 17:36:50,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=553191.3333333334, ans=0.125
2024-09-24 17:37:00,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=553238.0, ans=0.125
2024-09-24 17:37:21,451 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.255e+02 1.333e+02 1.430e+02 2.111e+02, threshold=2.665e+02, percent-clipped=0.0
2024-09-24 17:37:21,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=553284.6666666666, ans=0.0
2024-09-24 17:37:31,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=553284.6666666666, ans=0.0
2024-09-24 17:37:34,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=553331.3333333334, ans=0.125
2024-09-24 17:37:49,564 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.94 vs. limit=10.0
2024-09-24 17:37:50,371 INFO [train.py:1198] (1/4) Epoch 31, batch 1700, loss[loss=0.2001, ctc_loss=0.1317, cr_loss=0.3423, over 17257.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1307, cr_loss=0.3472, over 3351377.77 frames. ], batch size: 44, lr: 3.85e-03, grad_scale: 32.0
2024-09-24 17:37:53,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=553378.0, ans=0.0
2024-09-24 17:38:16,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=553424.6666666666, ans=0.125
2024-09-24 17:38:36,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=553471.3333333334, ans=0.0
2024-09-24 17:38:37,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=553518.0, ans=0.025
2024-09-24 17:39:13,900 INFO [train.py:1198] (1/4) Epoch 31, batch 1750, loss[loss=0.1796, ctc_loss=0.115, cr_loss=0.3234, over 17035.00 frames. ], tot_loss[loss=0.1997, ctc_loss=0.1303, cr_loss=0.3468, over 3351943.25 frames. ], batch size: 44, lr: 3.85e-03, grad_scale: 32.0
2024-09-24 17:39:23,955 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.12 vs. limit=22.5
2024-09-24 17:39:28,465 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 17:39:28,697 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.11 vs. limit=22.5
2024-09-24 17:39:31,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=553658.0, ans=0.125
2024-09-24 17:39:33,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=553658.0, ans=0.0
2024-09-24 17:39:36,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=553658.0, ans=0.025
2024-09-24 17:39:52,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=553704.6666666666, ans=0.2
2024-09-24 17:40:04,937 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.305e+02 1.399e+02 1.556e+02 3.053e+02, threshold=2.798e+02, percent-clipped=1.0
2024-09-24 17:40:11,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=553751.3333333334, ans=0.0
2024-09-24 17:40:11,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.max_positive, batch_count=553751.3333333334, ans=0.95
2024-09-24 17:40:11,927 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.92 vs. limit=15.0
2024-09-24 17:40:12,275 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=7.47 vs. limit=15.0
2024-09-24 17:40:33,711 INFO [train.py:1198] (1/4) Epoch 31, batch 1800, loss[loss=0.204, ctc_loss=0.1322, cr_loss=0.3593, over 15982.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1299, cr_loss=0.3463, over 3350840.58 frames. ], batch size: 74, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:40:34,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=553844.6666666666, ans=0.1
2024-09-24 17:40:40,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=553844.6666666666, ans=0.125
2024-09-24 17:41:05,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.06 vs. limit=22.5
2024-09-24 17:41:46,277 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.18 vs. limit=15.0
2024-09-24 17:42:01,014 INFO [train.py:1198] (1/4) Epoch 31, batch 1850, loss[loss=0.2124, ctc_loss=0.1378, cr_loss=0.3728, over 17128.00 frames. ], tot_loss[loss=0.1993, ctc_loss=0.13, cr_loss=0.3467, over 3352075.60 frames. ], batch size: 48, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:42:08,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.08 vs. limit=12.0
2024-09-24 17:42:14,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=554078.0, ans=0.1
2024-09-24 17:42:15,039 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.17 vs. limit=15.0
2024-09-24 17:42:28,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=554124.6666666666, ans=0.0
2024-09-24 17:42:32,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=554171.3333333334, ans=0.125
2024-09-24 17:42:36,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=554171.3333333334, ans=0.025
2024-09-24 17:42:53,413 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.284e+02 1.348e+02 1.466e+02 2.241e+02, threshold=2.697e+02, percent-clipped=0.0
2024-09-24 17:42:53,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=554218.0, ans=0.1
2024-09-24 17:42:53,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=554218.0, ans=0.09899494936611666
2024-09-24 17:43:01,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=554218.0, ans=0.1
2024-09-24 17:43:11,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=554264.6666666666, ans=0.2
2024-09-24 17:43:20,557 INFO [train.py:1198] (1/4) Epoch 31, batch 1900, loss[loss=0.1796, ctc_loss=0.116, cr_loss=0.318, over 17145.00 frames. ], tot_loss[loss=0.1992, ctc_loss=0.1299, cr_loss=0.3462, over 3341247.43 frames. ], batch size: 48, lr: 3.85e-03, grad_scale: 16.0
2024-09-24 17:43:20,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=554311.3333333334, ans=0.0
2024-09-24 17:43:37,177 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.15 vs. limit=22.5
2024-09-24 17:43:46,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=554358.0, ans=0.125
2024-09-24 17:43:51,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=554404.6666666666, ans=0.125
2024-09-24 17:44:32,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=554498.0, ans=0.0
2024-09-24 17:44:37,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn1.whiten.whitening_limit, batch_count=554498.0, ans=22.5
2024-09-24 17:44:42,948 INFO [train.py:1198] (1/4) Epoch 31, batch 1950, loss[loss=0.2274, ctc_loss=0.1502, cr_loss=0.3863, over 16459.00 frames. ], tot_loss[loss=0.2003, ctc_loss=0.1307, cr_loss=0.3482, over 3342556.13 frames. ], batch size: 66, lr: 3.84e-03, grad_scale: 16.0
2024-09-24 17:44:48,349 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.84 vs. limit=15.0
2024-09-24 17:44:59,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=554591.3333333334, ans=0.1
2024-09-24 17:45:35,729 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.290e+02 1.379e+02 1.473e+02 2.110e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-24 17:45:47,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=554731.3333333334, ans=0.2
2024-09-24 17:45:54,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=554731.3333333334, ans=0.125
2024-09-24 17:45:59,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=554731.3333333334, ans=0.1
2024-09-24 17:46:05,598 INFO [train.py:1198] (1/4) Epoch 31, batch 2000, loss[loss=0.1817, ctc_loss=0.1191, cr_loss=0.3131, over 16908.00 frames. ], tot_loss[loss=0.1996, ctc_loss=0.1302, cr_loss=0.3471, over 3342395.46 frames. ], batch size: 58, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:46:07,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=554778.0, ans=0.0
2024-09-24 17:46:07,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=554778.0, ans=0.0
2024-09-24 17:46:09,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.56 vs. limit=22.5
2024-09-24 17:46:32,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=554824.6666666666, ans=0.1
2024-09-24 17:46:36,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.79 vs. limit=15.0
2024-09-24 17:47:23,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=554964.6666666666, ans=0.125
2024-09-24 17:47:31,371 INFO [train.py:1198] (1/4) Epoch 31, batch 2050, loss[loss=0.1668, ctc_loss=0.1069, cr_loss=0.2998, over 17289.00 frames. ], tot_loss[loss=0.1993, ctc_loss=0.13, cr_loss=0.3469, over 3346628.12 frames. ], batch size: 42, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:47:33,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=555011.3333333334, ans=0.125
2024-09-24 17:48:14,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=555104.6666666666, ans=0.0
2024-09-24 17:48:18,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=555151.3333333334, ans=0.125
2024-09-24 17:48:24,148 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.242e+02 1.319e+02 1.408e+02 1.614e+02, threshold=2.639e+02, percent-clipped=0.0
2024-09-24 17:48:24,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=555151.3333333334, ans=0.09899494936611666
2024-09-24 17:48:24,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=555151.3333333334, ans=0.0
2024-09-24 17:48:27,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=555151.3333333334, ans=0.0
2024-09-24 17:48:29,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=555151.3333333334, ans=0.015
2024-09-24 17:48:31,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=555151.3333333334, ans=0.0
2024-09-24 17:48:34,712 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.59 vs. limit=15.0
2024-09-24 17:48:48,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=555198.0, ans=0.0
2024-09-24 17:48:50,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.53 vs. limit=15.0
2024-09-24 17:48:51,333 INFO [train.py:1198] (1/4) Epoch 31, batch 2100, loss[loss=0.1754, ctc_loss=0.1109, cr_loss=0.3225, over 17278.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.1301, cr_loss=0.3474, over 3336178.30 frames. ], batch size: 42, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:49:48,442 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.90 vs. limit=15.0
2024-09-24 17:49:54,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=11.66 vs. limit=15.0
2024-09-24 17:50:14,773 INFO [train.py:1198] (1/4) Epoch 31, batch 2150, loss[loss=0.1906, ctc_loss=0.1239, cr_loss=0.3338, over 16940.00 frames. ], tot_loss[loss=0.1997, ctc_loss=0.1301, cr_loss=0.3477, over 3323033.99 frames. ], batch size: 42, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:50:37,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=555524.6666666666, ans=0.125
2024-09-24 17:50:38,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=555524.6666666666, ans=0.015
2024-09-24 17:50:46,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=555524.6666666666, ans=0.5
2024-09-24 17:51:00,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=555571.3333333334, ans=0.0
2024-09-24 17:51:10,125 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.261e+02 1.344e+02 1.456e+02 2.277e+02, threshold=2.688e+02, percent-clipped=0.0
2024-09-24 17:51:13,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=555618.0, ans=0.0
2024-09-24 17:51:34,311 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.21 vs. limit=15.0
2024-09-24 17:51:36,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=555664.6666666666, ans=0.125
2024-09-24 17:51:39,714 INFO [train.py:1198] (1/4) Epoch 31, batch 2200, loss[loss=0.1911, ctc_loss=0.124, cr_loss=0.3353, over 17197.00 frames. ], tot_loss[loss=0.1986, ctc_loss=0.1293, cr_loss=0.3464, over 3333374.70 frames. ], batch size: 41, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:51:55,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=555711.3333333334, ans=0.0
2024-09-24 17:52:03,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten.whitening_limit, batch_count=555758.0, ans=22.5
2024-09-24 17:53:02,094 INFO [train.py:1198] (1/4) Epoch 31, batch 2250, loss[loss=0.2136, ctc_loss=0.1417, cr_loss=0.3595, over 16880.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1284, cr_loss=0.3451, over 3339574.89 frames. ], batch size: 58, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:53:03,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=555944.6666666666, ans=0.0
2024-09-24 17:53:10,755 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.72 vs. limit=15.0
2024-09-24 17:53:40,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=556038.0, ans=0.07
2024-09-24 17:53:41,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=556038.0, ans=0.025
2024-09-24 17:53:57,559 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.242e+02 1.333e+02 1.428e+02 2.130e+02, threshold=2.666e+02, percent-clipped=0.0
2024-09-24 17:53:59,840 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.09 vs. limit=15.0
2024-09-24 17:54:06,440 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.96 vs. limit=15.0
2024-09-24 17:54:24,673 INFO [train.py:1198] (1/4) Epoch 31, batch 2300, loss[loss=0.1582, ctc_loss=0.1016, cr_loss=0.2827, over 17110.00 frames. ], tot_loss[loss=0.1975, ctc_loss=0.1284, cr_loss=0.3456, over 3339292.11 frames. ], batch size: 40, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:55:00,299 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=556271.3333333334, ans=0.0
2024-09-24 17:55:40,213 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.43 vs. limit=6.0
2024-09-24 17:55:47,357 INFO [train.py:1198] (1/4) Epoch 31, batch 2350, loss[loss=0.2078, ctc_loss=0.1382, cr_loss=0.3479, over 16537.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.1289, cr_loss=0.3471, over 3350875.42 frames. ], batch size: 66, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:56:01,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=556458.0, ans=0.0
2024-09-24 17:56:09,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=556458.0, ans=0.2
2024-09-24 17:56:18,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.55 vs. limit=6.0
2024-09-24 17:56:41,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=556551.3333333334, ans=0.0
2024-09-24 17:56:43,026 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.265e+02 1.349e+02 1.461e+02 1.759e+02, threshold=2.698e+02, percent-clipped=0.0
2024-09-24 17:56:54,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=556551.3333333334, ans=0.5
2024-09-24 17:57:12,807 INFO [train.py:1198] (1/4) Epoch 31, batch 2400, loss[loss=0.1881, ctc_loss=0.1244, cr_loss=0.3185, over 16740.00 frames. ], tot_loss[loss=0.199, ctc_loss=0.1295, cr_loss=0.3474, over 3346189.45 frames. ], batch size: 61, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:57:20,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=556644.6666666666, ans=0.0
2024-09-24 17:58:32,915 INFO [train.py:1198] (1/4) Epoch 31, batch 2450, loss[loss=0.211, ctc_loss=0.1391, cr_loss=0.3594, over 17051.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1294, cr_loss=0.3475, over 3354385.45 frames. ], batch size: 52, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 17:58:45,239 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.83 vs. limit=15.0
2024-09-24 17:59:06,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=556971.3333333334, ans=0.125
2024-09-24 17:59:17,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=556971.3333333334, ans=0.125
2024-09-24 17:59:27,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=557018.0, ans=0.125
2024-09-24 17:59:28,402 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.235e+02 1.328e+02 1.443e+02 2.383e+02, threshold=2.655e+02, percent-clipped=0.0
2024-09-24 17:59:55,317 INFO [train.py:1198] (1/4) Epoch 31, batch 2500, loss[loss=0.2115, ctc_loss=0.1385, cr_loss=0.365, over 17146.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.1291, cr_loss=0.3464, over 3363202.32 frames. ], batch size: 48, lr: 3.84e-03, grad_scale: 32.0
2024-09-24 18:00:00,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=557111.3333333334, ans=0.2
2024-09-24 18:00:20,421 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.57 vs. limit=6.0
2024-09-24 18:00:24,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=557158.0, ans=0.125
2024-09-24 18:00:36,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=557204.6666666666, ans=0.125
2024-09-24 18:00:53,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=557251.3333333334, ans=0.125
2024-09-24 18:00:59,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=557251.3333333334, ans=0.0
2024-09-24 18:01:14,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=557298.0, ans=0.125
2024-09-24 18:01:18,463 INFO [train.py:1198] (1/4) Epoch 31, batch 2550, loss[loss=0.1928, ctc_loss=0.1261, cr_loss=0.3335, over 17348.00 frames. ], tot_loss[loss=0.1982, ctc_loss=0.1289, cr_loss=0.3462, over 3368953.86 frames. ], batch size: 48, lr: 3.83e-03, grad_scale: 32.0
2024-09-24 18:01:28,086 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:01:42,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=557391.3333333334, ans=0.1
2024-09-24 18:01:42,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=557391.3333333334, ans=0.1
2024-09-24 18:02:16,588 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.233e+02 1.314e+02 1.407e+02 1.652e+02, threshold=2.628e+02, percent-clipped=0.0
2024-09-24 18:02:39,290 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:02:43,700 INFO [train.py:1198] (1/4) Epoch 31, batch 2600, loss[loss=0.2047, ctc_loss=0.1354, cr_loss=0.3469, over 17144.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1294, cr_loss=0.3475, over 3364518.24 frames. ], batch size: 48, lr: 3.83e-03, grad_scale: 32.0
2024-09-24 18:03:06,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=557624.6666666666, ans=10.0
2024-09-24 18:03:21,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=557671.3333333334, ans=0.05
2024-09-24 18:03:27,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=557671.3333333334, ans=0.1
2024-09-24 18:04:06,715 INFO [train.py:1198] (1/4) Epoch 31, batch 2650, loss[loss=0.2021, ctc_loss=0.1334, cr_loss=0.3439, over 17035.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1293, cr_loss=0.347, over 3363847.86 frames. ], batch size: 52, lr: 3.83e-03, grad_scale: 32.0
2024-09-24 18:04:07,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=557811.3333333334, ans=0.0
2024-09-24 18:04:07,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=557811.3333333334, ans=0.125
2024-09-24 18:04:10,621 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.64 vs. limit=15.0
2024-09-24 18:04:23,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=557858.0, ans=0.1
2024-09-24 18:04:32,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=557858.0, ans=0.125
2024-09-24 18:04:35,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=557858.0, ans=0.1
2024-09-24 18:04:59,304 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.276e+02 1.391e+02 1.491e+02 3.639e+02, threshold=2.781e+02, percent-clipped=1.0
2024-09-24 18:05:04,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=557951.3333333334, ans=0.125
2024-09-24 18:05:05,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=557951.3333333334, ans=0.025
2024-09-24 18:05:07,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=557951.3333333334, ans=0.0
2024-09-24 18:05:07,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=557951.3333333334, ans=0.025
2024-09-24 18:05:09,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=557998.0, ans=0.0
2024-09-24 18:05:26,482 INFO [train.py:1198] (1/4) Epoch 31, batch 2700, loss[loss=0.2295, ctc_loss=0.1524, cr_loss=0.3853, over 17241.00 frames. ], tot_loss[loss=0.2, ctc_loss=0.1303, cr_loss=0.3488, over 3367430.62 frames. ], batch size: 55, lr: 3.83e-03, grad_scale: 32.0
2024-09-24 18:06:01,795 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:06:07,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=558138.0, ans=0.0
2024-09-24 18:06:09,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=558138.0, ans=0.1
2024-09-24 18:06:17,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=558184.6666666666, ans=0.0
2024-09-24 18:06:37,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=558231.3333333334, ans=0.0
2024-09-24 18:06:54,369 INFO [train.py:1198] (1/4) Epoch 31, batch 2750, loss[loss=0.2246, ctc_loss=0.1495, cr_loss=0.3757, over 11613.00 frames. ], tot_loss[loss=0.2, ctc_loss=0.1302, cr_loss=0.3489, over 3359874.05 frames. ], batch size: 123, lr: 3.83e-03, grad_scale: 16.0
2024-09-24 18:06:54,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=558278.0, ans=0.04949747468305833
2024-09-24 18:07:13,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=558324.6666666666, ans=0.1
2024-09-24 18:07:15,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=558324.6666666666, ans=0.025
2024-09-24 18:07:25,429 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=8.66 vs. limit=15.0
2024-09-24 18:07:39,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=558371.3333333334, ans=0.1
2024-09-24 18:07:42,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=558418.0, ans=0.0
2024-09-24 18:07:47,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=558418.0, ans=0.025
2024-09-24 18:07:48,590 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.267e+02 1.363e+02 1.488e+02 1.840e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-24 18:07:50,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=558418.0, ans=0.125
2024-09-24 18:08:14,311 INFO [train.py:1198] (1/4) Epoch 31, batch 2800, loss[loss=0.2193, ctc_loss=0.1429, cr_loss=0.3821, over 16998.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1308, cr_loss=0.3503, over 3356063.57 frames. ], batch size: 58, lr: 3.83e-03, grad_scale: 32.0
2024-09-24 18:08:45,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=558558.0, ans=0.2
2024-09-24 18:08:47,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=558604.6666666666, ans=0.0
2024-09-24 18:08:49,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=558604.6666666666, ans=0.0
2024-09-24 18:08:54,057 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.24 vs. limit=15.0
2024-09-24 18:09:12,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=558651.3333333334, ans=0.125
2024-09-24 18:09:33,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=558698.0, ans=0.0
2024-09-24 18:09:36,683 INFO [train.py:1198] (1/4) Epoch 31, batch 2850, loss[loss=0.1924, ctc_loss=0.1261, cr_loss=0.3315, over 17282.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.1307, cr_loss=0.3495, over 3353137.27 frames. ], batch size: 51, lr: 3.83e-03, grad_scale: 32.0
2024-09-24 18:09:57,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=558791.3333333334, ans=0.09899494936611666
2024-09-24 18:10:21,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=558838.0, ans=0.125
2024-09-24 18:10:25,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=558884.6666666666, ans=0.0
2024-09-24 18:10:35,374 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.299e+02 1.361e+02 1.459e+02 2.038e+02, threshold=2.722e+02, percent-clipped=0.0
2024-09-24 18:10:37,691 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.74 vs. limit=6.0
2024-09-24 18:10:59,635 INFO [train.py:1198] (1/4) Epoch 31, batch 2900, loss[loss=0.1991, ctc_loss=0.13, cr_loss=0.3456, over 16730.00 frames. ], tot_loss[loss=0.199, ctc_loss=0.1295, cr_loss=0.3475, over 3363005.16 frames. ], batch size: 61, lr: 3.83e-03, grad_scale: 16.0
2024-09-24 18:11:05,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.86 vs. limit=15.0
2024-09-24 18:11:16,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.22 vs. limit=22.5
2024-09-24 18:11:23,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer_na.min_abs, batch_count=559024.6666666666, ans=0.02
2024-09-24 18:11:28,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=559024.6666666666, ans=0.125
2024-09-24 18:11:37,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=559071.3333333334, ans=0.1
2024-09-24 18:11:39,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=559071.3333333334, ans=0.125
2024-09-24 18:12:25,454 INFO [train.py:1198] (1/4) Epoch 31, batch 2950, loss[loss=0.1763, ctc_loss=0.1118, cr_loss=0.3227, over 16337.00 frames. ], tot_loss[loss=0.1993, ctc_loss=0.1297, cr_loss=0.3478, over 3364506.18 frames. ], batch size: 36, lr: 3.83e-03, grad_scale: 16.0
2024-09-24 18:12:25,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=559211.3333333334, ans=0.0
2024-09-24 18:12:37,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=559211.3333333334, ans=0.125
2024-09-24 18:13:15,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=559351.3333333334, ans=0.1
2024-09-24 18:13:21,739 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.279e+02 1.362e+02 1.485e+02 2.605e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-24 18:13:31,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=559398.0, ans=0.025
2024-09-24 18:13:41,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=559398.0, ans=0.125
2024-09-24 18:13:45,364 INFO [train.py:1198] (1/4) Epoch 31, batch 3000, loss[loss=0.1946, ctc_loss=0.1255, cr_loss=0.3457, over 17297.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1295, cr_loss=0.347, over 3364193.42 frames. ], batch size: 51, lr: 3.83e-03, grad_scale: 16.0
2024-09-24 18:13:45,365 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 18:14:00,848 INFO [train.py:1230] (1/4) Epoch 31, validation: loss=0.03667, ctc_loss=0.03667, cr_loss=9.013e-15, over 944034.00 frames. 
2024-09-24 18:14:00,848 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 18:14:59,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=559584.6666666666, ans=0.1
2024-09-24 18:15:10,563 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.67 vs. limit=15.0
2024-09-24 18:15:18,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten.whitening_limit, batch_count=559678.0, ans=15.0
2024-09-24 18:15:19,066 INFO [train.py:1198] (1/4) Epoch 31, batch 3050, loss[loss=0.2376, ctc_loss=0.1648, cr_loss=0.364, over 12090.00 frames. ], tot_loss[loss=0.1995, ctc_loss=0.13, cr_loss=0.3478, over 3356904.22 frames. ], batch size: 123, lr: 3.83e-03, grad_scale: 16.0
2024-09-24 18:15:36,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=559724.6666666666, ans=0.07
2024-09-24 18:16:00,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.56 vs. limit=6.0
2024-09-24 18:16:07,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=559818.0, ans=0.125
2024-09-24 18:16:13,651 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.275e+02 1.357e+02 1.505e+02 2.128e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-24 18:16:17,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=559818.0, ans=0.125
2024-09-24 18:16:37,063 INFO [train.py:1198] (1/4) Epoch 31, batch 3100, loss[loss=0.1972, ctc_loss=0.126, cr_loss=0.3559, over 16776.00 frames. ], tot_loss[loss=0.199, ctc_loss=0.1297, cr_loss=0.3466, over 3353617.10 frames. ], batch size: 61, lr: 3.83e-03, grad_scale: 16.0
2024-09-24 18:16:40,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=559911.3333333334, ans=0.0
2024-09-24 18:16:48,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=559911.3333333334, ans=0.1
2024-09-24 18:16:54,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=559958.0, ans=0.025
2024-09-24 18:16:54,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=559958.0, ans=0.125
2024-09-24 18:16:59,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=559958.0, ans=0.125
2024-09-24 18:17:12,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=560004.6666666666, ans=0.125
2024-09-24 18:17:34,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=560051.3333333334, ans=0.125
2024-09-24 18:17:53,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=560098.0, ans=0.125
2024-09-24 18:17:57,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=560098.0, ans=0.125
2024-09-24 18:17:58,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=560144.6666666666, ans=0.025
2024-09-24 18:18:00,009 INFO [train.py:1198] (1/4) Epoch 31, batch 3150, loss[loss=0.1505, ctc_loss=0.09582, cr_loss=0.2734, over 17269.00 frames. ], tot_loss[loss=0.1982, ctc_loss=0.1291, cr_loss=0.3456, over 3362029.64 frames. ], batch size: 42, lr: 3.83e-03, grad_scale: 16.0
2024-09-24 18:18:12,189 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.98 vs. limit=15.0
2024-09-24 18:18:28,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=560191.3333333334, ans=0.1
2024-09-24 18:18:31,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=560238.0, ans=0.0
2024-09-24 18:18:38,686 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.63 vs. limit=22.5
2024-09-24 18:18:55,087 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.279e+02 1.355e+02 1.423e+02 2.365e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-24 18:18:59,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=560284.6666666666, ans=0.125
2024-09-24 18:19:11,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=560331.3333333334, ans=0.07
2024-09-24 18:19:18,857 INFO [train.py:1198] (1/4) Epoch 31, batch 3200, loss[loss=0.2248, ctc_loss=0.1548, cr_loss=0.3501, over 11641.00 frames. ], tot_loss[loss=0.1985, ctc_loss=0.1294, cr_loss=0.3459, over 3351398.73 frames. ], batch size: 123, lr: 3.82e-03, grad_scale: 32.0
2024-09-24 18:20:24,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=560564.6666666666, ans=0.125
2024-09-24 18:20:24,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=560564.6666666666, ans=0.1
2024-09-24 18:20:25,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=560564.6666666666, ans=0.125
2024-09-24 18:20:27,496 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:20:38,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=560564.6666666666, ans=0.125
2024-09-24 18:20:41,360 INFO [train.py:1198] (1/4) Epoch 31, batch 3250, loss[loss=0.1904, ctc_loss=0.1246, cr_loss=0.329, over 17265.00 frames. ], tot_loss[loss=0.1989, ctc_loss=0.1296, cr_loss=0.3465, over 3355885.03 frames. ], batch size: 44, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:21:10,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.36 vs. limit=22.5
2024-09-24 18:21:17,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=560704.6666666666, ans=0.125
2024-09-24 18:21:33,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=560751.3333333334, ans=0.125
2024-09-24 18:21:37,469 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.237e+02 1.355e+02 1.495e+02 1.938e+02, threshold=2.711e+02, percent-clipped=0.0
2024-09-24 18:21:55,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=560798.0, ans=0.0
2024-09-24 18:21:59,317 INFO [train.py:1198] (1/4) Epoch 31, batch 3300, loss[loss=0.1827, ctc_loss=0.1212, cr_loss=0.3077, over 17170.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1297, cr_loss=0.347, over 3355028.13 frames. ], batch size: 45, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:22:38,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=560938.0, ans=0.04949747468305833
2024-09-24 18:22:40,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=560938.0, ans=0.2
2024-09-24 18:23:07,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=561031.3333333334, ans=0.1
2024-09-24 18:23:18,163 INFO [train.py:1198] (1/4) Epoch 31, batch 3350, loss[loss=0.1591, ctc_loss=0.09995, cr_loss=0.2956, over 17176.00 frames. ], tot_loss[loss=0.1988, ctc_loss=0.1294, cr_loss=0.3468, over 3356952.22 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:23:31,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=561078.0, ans=10.0
2024-09-24 18:23:41,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=561124.6666666666, ans=0.0
2024-09-24 18:23:46,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=561124.6666666666, ans=0.0
2024-09-24 18:24:00,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=561171.3333333334, ans=0.0
2024-09-24 18:24:07,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=561218.0, ans=0.125
2024-09-24 18:24:11,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=561218.0, ans=0.1
2024-09-24 18:24:11,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=561218.0, ans=0.0
2024-09-24 18:24:14,288 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.283e+02 1.367e+02 1.477e+02 4.748e+02, threshold=2.734e+02, percent-clipped=1.0
2024-09-24 18:24:36,278 INFO [train.py:1198] (1/4) Epoch 31, batch 3400, loss[loss=0.2212, ctc_loss=0.1445, cr_loss=0.383, over 17114.00 frames. ], tot_loss[loss=0.1988, ctc_loss=0.1294, cr_loss=0.347, over 3367435.94 frames. ], batch size: 49, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:24:36,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=561311.3333333334, ans=0.0
2024-09-24 18:25:11,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=561404.6666666666, ans=0.2
2024-09-24 18:25:15,120 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.15 vs. limit=22.5
2024-09-24 18:25:42,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=561498.0, ans=0.2
2024-09-24 18:25:51,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=561498.0, ans=0.0
2024-09-24 18:25:56,288 INFO [train.py:1198] (1/4) Epoch 31, batch 3450, loss[loss=0.1867, ctc_loss=0.1232, cr_loss=0.3176, over 16657.00 frames. ], tot_loss[loss=0.199, ctc_loss=0.1296, cr_loss=0.3469, over 3354175.10 frames. ], batch size: 66, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:25:58,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=561544.6666666666, ans=0.5
2024-09-24 18:26:07,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=561544.6666666666, ans=0.125
2024-09-24 18:26:09,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=561544.6666666666, ans=0.2
2024-09-24 18:26:20,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=561591.3333333334, ans=0.0
2024-09-24 18:26:23,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=561591.3333333334, ans=0.2
2024-09-24 18:26:27,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=561638.0, ans=0.125
2024-09-24 18:26:31,126 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.49 vs. limit=15.0
2024-09-24 18:26:46,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=561684.6666666666, ans=0.05
2024-09-24 18:26:46,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=561684.6666666666, ans=0.0
2024-09-24 18:26:52,252 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.286e+02 1.363e+02 1.449e+02 2.372e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-24 18:27:14,199 INFO [train.py:1198] (1/4) Epoch 31, batch 3500, loss[loss=0.2069, ctc_loss=0.1341, cr_loss=0.3641, over 17004.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1293, cr_loss=0.3468, over 3360598.91 frames. ], batch size: 56, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:27:46,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=561871.3333333334, ans=0.015
2024-09-24 18:28:11,695 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.47 vs. limit=12.0
2024-09-24 18:28:34,276 INFO [train.py:1198] (1/4) Epoch 31, batch 3550, loss[loss=0.1923, ctc_loss=0.1258, cr_loss=0.3323, over 17206.00 frames. ], tot_loss[loss=0.1983, ctc_loss=0.1291, cr_loss=0.3463, over 3361907.47 frames. ], batch size: 47, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:28:36,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=562011.3333333334, ans=0.125
2024-09-24 18:28:53,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=562058.0, ans=0.0
2024-09-24 18:29:04,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=562104.6666666666, ans=0.125
2024-09-24 18:29:12,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer_ff2.min_abs, batch_count=562104.6666666666, ans=0.1
2024-09-24 18:29:23,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=562151.3333333334, ans=0.1
2024-09-24 18:29:34,934 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.302e+02 1.378e+02 1.478e+02 2.528e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-24 18:29:56,785 INFO [train.py:1198] (1/4) Epoch 31, batch 3600, loss[loss=0.1932, ctc_loss=0.1298, cr_loss=0.3172, over 16785.00 frames. ], tot_loss[loss=0.1983, ctc_loss=0.129, cr_loss=0.3464, over 3363655.39 frames. ], batch size: 61, lr: 3.82e-03, grad_scale: 32.0
2024-09-24 18:30:00,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=562244.6666666666, ans=0.125
2024-09-24 18:30:17,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=562291.3333333334, ans=0.025
2024-09-24 18:30:32,968 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.91 vs. limit=22.5
2024-09-24 18:30:43,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=562384.6666666666, ans=0.125
2024-09-24 18:31:00,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=562431.3333333334, ans=0.125
2024-09-24 18:31:02,582 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.11 vs. limit=15.0
2024-09-24 18:31:14,555 INFO [train.py:1198] (1/4) Epoch 31, batch 3650, loss[loss=0.1754, ctc_loss=0.1123, cr_loss=0.3156, over 17037.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1294, cr_loss=0.3467, over 3353891.85 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 32.0
2024-09-24 18:31:26,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=562478.0, ans=0.0
2024-09-24 18:31:29,561 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.73 vs. limit=12.0
2024-09-24 18:31:36,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=562524.6666666666, ans=0.125
2024-09-24 18:31:38,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=562524.6666666666, ans=0.125
2024-09-24 18:31:46,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=562571.3333333334, ans=0.125
2024-09-24 18:32:11,872 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.260e+02 1.374e+02 1.513e+02 2.191e+02, threshold=2.748e+02, percent-clipped=0.0
2024-09-24 18:32:33,922 INFO [train.py:1198] (1/4) Epoch 31, batch 3700, loss[loss=0.2112, ctc_loss=0.1342, cr_loss=0.3849, over 16944.00 frames. ], tot_loss[loss=0.1997, ctc_loss=0.1301, cr_loss=0.3478, over 3343582.41 frames. ], batch size: 58, lr: 3.82e-03, grad_scale: 32.0
2024-09-24 18:32:37,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten.whitening_limit, batch_count=562711.3333333334, ans=15.0
2024-09-24 18:32:38,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=562711.3333333334, ans=0.0
2024-09-24 18:32:43,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=562711.3333333334, ans=0.5
2024-09-24 18:32:44,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=562711.3333333334, ans=0.125
2024-09-24 18:32:45,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=562711.3333333334, ans=0.125
2024-09-24 18:33:10,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.20 vs. limit=6.0
2024-09-24 18:33:26,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=562851.3333333334, ans=0.125
2024-09-24 18:33:29,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=562851.3333333334, ans=0.125
2024-09-24 18:33:52,140 INFO [train.py:1198] (1/4) Epoch 31, batch 3750, loss[loss=0.1697, ctc_loss=0.1097, cr_loss=0.2998, over 17132.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1298, cr_loss=0.3464, over 3331357.37 frames. ], batch size: 40, lr: 3.82e-03, grad_scale: 32.0
2024-09-24 18:33:58,908 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.16 vs. limit=15.0
2024-09-24 18:34:03,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=562944.6666666666, ans=0.125
2024-09-24 18:34:13,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten.whitening_limit, batch_count=562991.3333333334, ans=15.0
2024-09-24 18:34:37,254 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.80 vs. limit=22.5
2024-09-24 18:34:44,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=563084.6666666666, ans=0.0
2024-09-24 18:34:48,370 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.261e+02 1.326e+02 1.431e+02 2.182e+02, threshold=2.652e+02, percent-clipped=0.0
2024-09-24 18:34:50,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=563084.6666666666, ans=0.125
2024-09-24 18:34:56,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=563131.3333333334, ans=0.04949747468305833
2024-09-24 18:35:09,880 INFO [train.py:1198] (1/4) Epoch 31, batch 3800, loss[loss=0.1795, ctc_loss=0.1147, cr_loss=0.3238, over 17244.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1299, cr_loss=0.3462, over 3312088.28 frames. ], batch size: 44, lr: 3.82e-03, grad_scale: 16.0
2024-09-24 18:35:42,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=563271.3333333334, ans=0.125
2024-09-24 18:35:46,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=563271.3333333334, ans=0.125
2024-09-24 18:35:54,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=563318.0, ans=0.0
2024-09-24 18:35:56,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=563318.0, ans=0.125
2024-09-24 18:35:56,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.75 vs. limit=15.0
2024-09-24 18:36:19,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=563364.6666666666, ans=0.125
2024-09-24 18:36:27,169 INFO [train.py:1198] (1/4) Epoch 31, batch 3850, loss[loss=0.2121, ctc_loss=0.1378, cr_loss=0.3715, over 17326.00 frames. ], tot_loss[loss=0.2005, ctc_loss=0.1311, cr_loss=0.347, over 3276382.19 frames. ], batch size: 51, lr: 3.81e-03, grad_scale: 16.0
2024-09-24 18:36:35,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=563411.3333333334, ans=0.0
2024-09-24 18:36:57,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_abs, batch_count=563504.6666666666, ans=0.5
2024-09-24 18:37:10,506 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.68 vs. limit=22.5
2024-09-24 18:37:17,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=563551.3333333334, ans=0.0
2024-09-24 18:37:24,225 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.313e+02 1.427e+02 1.596e+02 2.274e+02, threshold=2.854e+02, percent-clipped=0.0
2024-09-24 18:38:29,743 INFO [train.py:1198] (1/4) Epoch 32, batch 0, loss[loss=0.2012, ctc_loss=0.1308, cr_loss=0.3518, over 17146.00 frames. ], tot_loss[loss=0.2012, ctc_loss=0.1308, cr_loss=0.3518, over 17146.00 frames. ], batch size: 48, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:38:29,743 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 18:38:45,169 INFO [train.py:1230] (1/4) Epoch 32, validation: loss=0.03599, ctc_loss=0.03599, cr_loss=9.022e-15, over 944034.00 frames. 
2024-09-24 18:38:45,169 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 18:38:58,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=563626.0, ans=0.125
2024-09-24 18:39:20,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=563719.3333333334, ans=0.1
2024-09-24 18:39:24,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=563719.3333333334, ans=0.125
2024-09-24 18:39:27,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=563719.3333333334, ans=0.125
2024-09-24 18:39:33,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=563766.0, ans=0.125
2024-09-24 18:39:33,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=563766.0, ans=0.125
2024-09-24 18:39:35,665 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.98 vs. limit=15.0
2024-09-24 18:40:14,197 INFO [train.py:1198] (1/4) Epoch 32, batch 50, loss[loss=0.2089, ctc_loss=0.1389, cr_loss=0.35, over 17026.00 frames. ], tot_loss[loss=0.2, ctc_loss=0.1303, cr_loss=0.3486, over 757268.02 frames. ], batch size: 56, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:40:19,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=563859.3333333334, ans=0.125
2024-09-24 18:40:43,236 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:41:19,567 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.253e+02 1.338e+02 1.477e+02 2.326e+02, threshold=2.677e+02, percent-clipped=0.0
2024-09-24 18:41:31,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=564046.0, ans=0.025
2024-09-24 18:41:33,986 INFO [train.py:1198] (1/4) Epoch 32, batch 100, loss[loss=0.1716, ctc_loss=0.1099, cr_loss=0.3088, over 16980.00 frames. ], tot_loss[loss=0.199, ctc_loss=0.1297, cr_loss=0.3465, over 1333671.42 frames. ], batch size: 42, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:41:34,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=564092.6666666666, ans=0.0
2024-09-24 18:41:48,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=564139.3333333334, ans=0.1
2024-09-24 18:42:04,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=564186.0, ans=0.1
2024-09-24 18:42:11,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=564186.0, ans=0.2
2024-09-24 18:42:11,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=564186.0, ans=0.125
2024-09-24 18:42:53,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=564279.3333333334, ans=0.125
2024-09-24 18:42:56,206 INFO [train.py:1198] (1/4) Epoch 32, batch 150, loss[loss=0.1863, ctc_loss=0.1187, cr_loss=0.3379, over 17252.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1279, cr_loss=0.3443, over 1791617.61 frames. ], batch size: 44, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:43:41,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=564419.3333333334, ans=0.2
2024-09-24 18:43:42,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=564466.0, ans=0.125
2024-09-24 18:43:59,222 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.79 vs. limit=10.0
2024-09-24 18:44:01,705 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.312e+02 1.410e+02 1.549e+02 1.890e+02, threshold=2.819e+02, percent-clipped=0.0
2024-09-24 18:44:10,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=564512.6666666666, ans=0.0
2024-09-24 18:44:10,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=564512.6666666666, ans=0.0
2024-09-24 18:44:14,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=564559.3333333334, ans=0.125
2024-09-24 18:44:16,131 INFO [train.py:1198] (1/4) Epoch 32, batch 200, loss[loss=0.207, ctc_loss=0.1364, cr_loss=0.3528, over 17228.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.128, cr_loss=0.3448, over 2141815.93 frames. ], batch size: 50, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:44:45,408 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:45:08,266 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:45:11,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=564699.3333333334, ans=0.0
2024-09-24 18:45:24,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=564699.3333333334, ans=0.125
2024-09-24 18:45:46,132 INFO [train.py:1198] (1/4) Epoch 32, batch 250, loss[loss=0.1671, ctc_loss=0.1033, cr_loss=0.3193, over 17239.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1283, cr_loss=0.3455, over 2414449.21 frames. ], batch size: 42, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:45:47,303 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.84 vs. limit=15.0
2024-09-24 18:45:59,154 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:46:01,104 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.12 vs. limit=15.0
2024-09-24 18:46:14,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=6.79 vs. limit=15.0
2024-09-24 18:46:27,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=564886.0, ans=0.125
2024-09-24 18:46:34,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=564932.6666666666, ans=0.035
2024-09-24 18:46:34,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=564932.6666666666, ans=0.1
2024-09-24 18:46:44,135 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.25 vs. limit=15.0
2024-09-24 18:46:52,819 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.269e+02 1.327e+02 1.431e+02 2.007e+02, threshold=2.654e+02, percent-clipped=0.0
2024-09-24 18:47:05,423 INFO [train.py:1198] (1/4) Epoch 32, batch 300, loss[loss=0.1592, ctc_loss=0.104, cr_loss=0.2761, over 17199.00 frames. ], tot_loss[loss=0.1975, ctc_loss=0.1284, cr_loss=0.3453, over 2625725.00 frames. ], batch size: 41, lr: 3.75e-03, grad_scale: 16.0
2024-09-24 18:47:25,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=565072.6666666666, ans=0.125
2024-09-24 18:48:10,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=565166.0, ans=0.125
2024-09-24 18:48:29,190 INFO [train.py:1198] (1/4) Epoch 32, batch 350, loss[loss=0.2092, ctc_loss=0.1354, cr_loss=0.3693, over 17156.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1287, cr_loss=0.3456, over 2795359.91 frames. ], batch size: 45, lr: 3.75e-03, grad_scale: 16.0
2024-09-24 18:49:14,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=565352.6666666666, ans=0.025
2024-09-24 18:49:22,954 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.47 vs. limit=15.0
2024-09-24 18:49:36,802 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.263e+02 1.319e+02 1.416e+02 2.178e+02, threshold=2.638e+02, percent-clipped=0.0
2024-09-24 18:49:44,970 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.22 vs. limit=15.0
2024-09-24 18:49:52,387 INFO [train.py:1198] (1/4) Epoch 32, batch 400, loss[loss=0.2391, ctc_loss=0.1586, cr_loss=0.4025, over 17233.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1288, cr_loss=0.3452, over 2914267.60 frames. ], batch size: 55, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:50:10,726 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.23 vs. limit=15.0
2024-09-24 18:51:06,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=565679.3333333334, ans=0.025
2024-09-24 18:51:11,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=565679.3333333334, ans=0.025
2024-09-24 18:51:17,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=565726.0, ans=0.125
2024-09-24 18:51:18,654 INFO [train.py:1198] (1/4) Epoch 32, batch 450, loss[loss=0.2146, ctc_loss=0.1405, cr_loss=0.3708, over 16705.00 frames. ], tot_loss[loss=0.1977, ctc_loss=0.1288, cr_loss=0.3449, over 3013977.77 frames. ], batch size: 61, lr: 3.75e-03, grad_scale: 32.0
2024-09-24 18:51:50,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=565819.3333333334, ans=0.0
2024-09-24 18:52:10,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=565866.0, ans=0.125
2024-09-24 18:52:25,961 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.290e+02 1.389e+02 1.522e+02 2.571e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-24 18:52:28,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=565912.6666666666, ans=0.125
2024-09-24 18:52:33,467 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.96 vs. limit=22.5
2024-09-24 18:52:38,789 INFO [train.py:1198] (1/4) Epoch 32, batch 500, loss[loss=0.2209, ctc_loss=0.1469, cr_loss=0.3696, over 17046.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1285, cr_loss=0.3445, over 3089558.47 frames. ], batch size: 52, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 18:52:39,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=565959.3333333334, ans=0.125
2024-09-24 18:52:58,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=566006.0, ans=0.1
2024-09-24 18:53:25,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=566052.6666666666, ans=0.125
2024-09-24 18:53:47,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=566146.0, ans=0.0
2024-09-24 18:54:01,200 INFO [train.py:1198] (1/4) Epoch 32, batch 550, loss[loss=0.1782, ctc_loss=0.1174, cr_loss=0.3042, over 17082.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.128, cr_loss=0.3436, over 3156527.35 frames. ], batch size: 43, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 18:54:41,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=566286.0, ans=0.1
2024-09-24 18:54:46,246 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.36 vs. limit=22.5
2024-09-24 18:55:11,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=566379.3333333334, ans=15.0
2024-09-24 18:55:15,290 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.38 vs. limit=15.0
2024-09-24 18:55:16,236 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.263e+02 1.353e+02 1.425e+02 2.013e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-24 18:55:26,801 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.84 vs. limit=15.0
2024-09-24 18:55:29,023 INFO [train.py:1198] (1/4) Epoch 32, batch 600, loss[loss=0.2204, ctc_loss=0.1444, cr_loss=0.3801, over 16729.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1288, cr_loss=0.3453, over 3197811.12 frames. ], batch size: 61, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 18:55:42,873 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.51 vs. limit=10.0
2024-09-24 18:55:58,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=566472.6666666666, ans=0.07
2024-09-24 18:55:59,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=566519.3333333334, ans=0.0
2024-09-24 18:56:01,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=566519.3333333334, ans=0.125
2024-09-24 18:56:02,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=566519.3333333334, ans=0.2
2024-09-24 18:56:05,021 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.85 vs. limit=6.0
2024-09-24 18:56:29,030 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:56:38,569 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 18:56:49,244 INFO [train.py:1198] (1/4) Epoch 32, batch 650, loss[loss=0.2106, ctc_loss=0.1373, cr_loss=0.3662, over 17349.00 frames. ], tot_loss[loss=0.198, ctc_loss=0.1289, cr_loss=0.3456, over 3235266.30 frames. ], batch size: 48, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 18:57:09,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=566706.0, ans=0.1
2024-09-24 18:57:13,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=566706.0, ans=0.0
2024-09-24 18:58:00,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=566846.0, ans=10.0
2024-09-24 18:58:01,567 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.265e+02 1.321e+02 1.437e+02 2.053e+02, threshold=2.643e+02, percent-clipped=0.0
2024-09-24 18:58:05,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=566846.0, ans=0.0
2024-09-24 18:58:06,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=566846.0, ans=0.125
2024-09-24 18:58:13,019 INFO [train.py:1198] (1/4) Epoch 32, batch 700, loss[loss=0.1544, ctc_loss=0.09481, cr_loss=0.298, over 17034.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1283, cr_loss=0.3447, over 3270709.10 frames. ], batch size: 39, lr: 3.74e-03, grad_scale: 16.0
2024-09-24 18:58:17,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=566892.6666666666, ans=0.125
2024-09-24 18:58:30,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=566939.3333333334, ans=0.125
2024-09-24 18:58:37,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=566939.3333333334, ans=0.1
2024-09-24 18:58:37,450 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.86 vs. limit=10.0
2024-09-24 18:58:51,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=566986.0, ans=0.1
2024-09-24 18:58:55,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=566986.0, ans=0.125
2024-09-24 18:59:20,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=567079.3333333334, ans=0.125
2024-09-24 18:59:23,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=567079.3333333334, ans=0.125
2024-09-24 18:59:30,505 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.00 vs. limit=6.0
2024-09-24 18:59:33,160 INFO [train.py:1198] (1/4) Epoch 32, batch 750, loss[loss=0.2047, ctc_loss=0.1325, cr_loss=0.3612, over 17015.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.128, cr_loss=0.3447, over 3295319.51 frames. ], batch size: 39, lr: 3.74e-03, grad_scale: 16.0
2024-09-24 18:59:55,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=567172.6666666666, ans=0.0
2024-09-24 19:00:22,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=567219.3333333334, ans=0.125
2024-09-24 19:00:43,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=567312.6666666666, ans=0.1
2024-09-24 19:00:44,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=567312.6666666666, ans=0.125
2024-09-24 19:00:49,278 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.286e+02 1.352e+02 1.497e+02 2.163e+02, threshold=2.705e+02, percent-clipped=0.0
2024-09-24 19:01:00,448 INFO [train.py:1198] (1/4) Epoch 32, batch 800, loss[loss=0.1706, ctc_loss=0.1095, cr_loss=0.3055, over 17026.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1276, cr_loss=0.3432, over 3321812.62 frames. ], batch size: 39, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 19:01:08,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=567359.3333333334, ans=0.125
2024-09-24 19:01:14,960 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 19:01:30,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=567452.6666666666, ans=0.125
2024-09-24 19:01:35,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=567452.6666666666, ans=0.125
2024-09-24 19:01:50,598 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.36 vs. limit=15.0
2024-09-24 19:01:56,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=567499.3333333334, ans=0.025
2024-09-24 19:02:12,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=567546.0, ans=0.125
2024-09-24 19:02:20,178 INFO [train.py:1198] (1/4) Epoch 32, batch 850, loss[loss=0.1963, ctc_loss=0.1285, cr_loss=0.3388, over 17209.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1279, cr_loss=0.3431, over 3318913.37 frames. ], batch size: 50, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 19:02:46,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=567639.3333333334, ans=0.0
2024-09-24 19:03:25,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=567779.3333333334, ans=0.025
2024-09-24 19:03:25,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=567779.3333333334, ans=0.07
2024-09-24 19:03:31,999 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.300e+02 1.382e+02 1.535e+02 3.730e+02, threshold=2.764e+02, percent-clipped=1.0
2024-09-24 19:03:42,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=567826.0, ans=0.2
2024-09-24 19:03:43,406 INFO [train.py:1198] (1/4) Epoch 32, batch 900, loss[loss=0.2086, ctc_loss=0.1376, cr_loss=0.3553, over 17281.00 frames. ], tot_loss[loss=0.1979, ctc_loss=0.1288, cr_loss=0.3455, over 3334618.43 frames. ], batch size: 46, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 19:03:47,276 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.17 vs. limit=6.0
2024-09-24 19:03:56,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=567826.0, ans=0.0
2024-09-24 19:04:06,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=567872.6666666666, ans=0.0
2024-09-24 19:04:33,011 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.61 vs. limit=15.0
2024-09-24 19:04:58,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=568012.6666666666, ans=0.2
2024-09-24 19:05:09,169 INFO [train.py:1198] (1/4) Epoch 32, batch 950, loss[loss=0.1908, ctc_loss=0.1247, cr_loss=0.3303, over 16863.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1287, cr_loss=0.3454, over 3343958.37 frames. ], batch size: 58, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 19:05:10,133 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.55 vs. limit=12.0
2024-09-24 19:05:29,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=568106.0, ans=0.1
2024-09-24 19:05:37,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=568106.0, ans=0.05
2024-09-24 19:05:55,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=568152.6666666666, ans=0.0
2024-09-24 19:06:04,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=568199.3333333334, ans=0.0
2024-09-24 19:06:11,731 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.96 vs. limit=15.0
2024-09-24 19:06:13,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.31 vs. limit=15.0
2024-09-24 19:06:20,503 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.287e+02 1.382e+02 1.480e+02 1.903e+02, threshold=2.765e+02, percent-clipped=0.0
2024-09-24 19:06:31,838 INFO [train.py:1198] (1/4) Epoch 32, batch 1000, loss[loss=0.1876, ctc_loss=0.1225, cr_loss=0.3257, over 17162.00 frames. ], tot_loss[loss=0.1977, ctc_loss=0.1287, cr_loss=0.3454, over 3348495.73 frames. ], batch size: 45, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 19:07:25,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=568432.6666666666, ans=0.125
2024-09-24 19:07:52,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=568479.3333333334, ans=0.125
2024-09-24 19:07:54,840 INFO [train.py:1198] (1/4) Epoch 32, batch 1050, loss[loss=0.1785, ctc_loss=0.1147, cr_loss=0.3191, over 17303.00 frames. ], tot_loss[loss=0.198, ctc_loss=0.1289, cr_loss=0.3457, over 3345125.26 frames. ], batch size: 46, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 19:07:56,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=568526.0, ans=0.125
2024-09-24 19:08:06,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=568526.0, ans=0.025
2024-09-24 19:08:14,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=568572.6666666666, ans=0.0
2024-09-24 19:08:25,859 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.93 vs. limit=15.0
2024-09-24 19:08:33,930 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.67 vs. limit=15.0
2024-09-24 19:08:34,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=568619.3333333334, ans=0.125
2024-09-24 19:08:41,176 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=1.360e-02
2024-09-24 19:08:55,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=568666.0, ans=0.2
2024-09-24 19:08:55,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.max_positive, batch_count=568666.0, ans=0.95
2024-09-24 19:09:03,547 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.269e+02 1.394e+02 1.503e+02 2.012e+02, threshold=2.787e+02, percent-clipped=0.0
2024-09-24 19:09:03,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=568712.6666666666, ans=0.125
2024-09-24 19:09:15,037 INFO [train.py:1198] (1/4) Epoch 32, batch 1100, loss[loss=0.2062, ctc_loss=0.1346, cr_loss=0.3578, over 17319.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1284, cr_loss=0.3449, over 3350614.46 frames. ], batch size: 51, lr: 3.74e-03, grad_scale: 32.0
2024-09-24 19:09:16,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=568759.3333333334, ans=0.125
2024-09-24 19:09:18,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=568759.3333333334, ans=0.025
2024-09-24 19:09:21,543 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 19:09:31,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=568806.0, ans=0.1
2024-09-24 19:10:30,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=568946.0, ans=0.1
2024-09-24 19:10:33,410 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.78 vs. limit=15.0
2024-09-24 19:10:35,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=568946.0, ans=0.0
2024-09-24 19:10:37,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=568946.0, ans=10.0
2024-09-24 19:10:41,724 INFO [train.py:1198] (1/4) Epoch 32, batch 1150, loss[loss=0.2194, ctc_loss=0.1425, cr_loss=0.3846, over 17194.00 frames. ], tot_loss[loss=0.1973, ctc_loss=0.1283, cr_loss=0.345, over 3347197.18 frames. ], batch size: 55, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:10:48,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=568992.6666666666, ans=0.125
2024-09-24 19:10:50,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=568992.6666666666, ans=0.5
2024-09-24 19:11:23,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=569086.0, ans=0.5
2024-09-24 19:11:29,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=569132.6666666666, ans=0.0
2024-09-24 19:11:34,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=569132.6666666666, ans=0.0
2024-09-24 19:11:44,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=569179.3333333334, ans=0.125
2024-09-24 19:11:50,711 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.271e+02 1.362e+02 1.461e+02 3.149e+02, threshold=2.725e+02, percent-clipped=1.0
2024-09-24 19:11:55,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=569179.3333333334, ans=0.1
2024-09-24 19:12:01,961 INFO [train.py:1198] (1/4) Epoch 32, batch 1200, loss[loss=0.1776, ctc_loss=0.1126, cr_loss=0.3248, over 16284.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1287, cr_loss=0.3457, over 3350550.37 frames. ], batch size: 36, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:12:13,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=569226.0, ans=0.025
2024-09-24 19:12:26,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=569272.6666666666, ans=0.1
2024-09-24 19:12:37,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=569319.3333333334, ans=0.1
2024-09-24 19:12:49,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=569319.3333333334, ans=0.125
2024-09-24 19:12:57,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=569366.0, ans=0.125
2024-09-24 19:13:07,431 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.04 vs. limit=6.0
2024-09-24 19:13:10,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=569412.6666666666, ans=0.125
2024-09-24 19:13:10,343 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 19:13:24,570 INFO [train.py:1198] (1/4) Epoch 32, batch 1250, loss[loss=0.1804, ctc_loss=0.1166, cr_loss=0.3191, over 17264.00 frames. ], tot_loss[loss=0.1983, ctc_loss=0.1291, cr_loss=0.3459, over 3349831.31 frames. ], batch size: 44, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:13:24,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=569459.3333333334, ans=0.0
2024-09-24 19:13:26,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=569459.3333333334, ans=0.0
2024-09-24 19:13:26,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=569459.3333333334, ans=0.1
2024-09-24 19:13:34,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=569459.3333333334, ans=0.2
2024-09-24 19:13:37,686 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=569459.3333333334, ans=0.125
2024-09-24 19:13:44,595 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.58 vs. limit=15.0
2024-09-24 19:13:48,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=569506.0, ans=0.125
2024-09-24 19:13:50,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=569506.0, ans=0.125
2024-09-24 19:13:54,690 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.66 vs. limit=15.0
2024-09-24 19:14:26,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=569599.3333333334, ans=0.1
2024-09-24 19:14:31,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=569646.0, ans=0.1
2024-09-24 19:14:34,071 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.275e+02 1.363e+02 1.497e+02 1.949e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-24 19:14:44,773 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=8.07 vs. limit=12.0
2024-09-24 19:14:50,538 INFO [train.py:1198] (1/4) Epoch 32, batch 1300, loss[loss=0.1801, ctc_loss=0.1181, cr_loss=0.31, over 17229.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.1281, cr_loss=0.3444, over 3355240.60 frames. ], batch size: 44, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:14:52,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=569692.6666666666, ans=0.0
2024-09-24 19:15:07,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=569739.3333333334, ans=0.125
2024-09-24 19:15:10,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=569739.3333333334, ans=0.0
2024-09-24 19:15:33,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=569786.0, ans=0.2
2024-09-24 19:15:42,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=569832.6666666666, ans=0.1
2024-09-24 19:15:57,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=569879.3333333334, ans=0.0
2024-09-24 19:16:12,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=569926.0, ans=0.0
2024-09-24 19:16:13,218 INFO [train.py:1198] (1/4) Epoch 32, batch 1350, loss[loss=0.2131, ctc_loss=0.139, cr_loss=0.3702, over 17030.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1278, cr_loss=0.3436, over 3364998.34 frames. ], batch size: 51, lr: 3.73e-03, grad_scale: 16.0
2024-09-24 19:16:23,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=569926.0, ans=0.125
2024-09-24 19:16:42,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=569972.6666666666, ans=0.0
2024-09-24 19:16:46,129 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.51 vs. limit=15.0
2024-09-24 19:16:54,416 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.01 vs. limit=22.5
2024-09-24 19:16:55,970 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.76 vs. limit=8.0
2024-09-24 19:17:14,935 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.07 vs. limit=12.0
2024-09-24 19:17:23,671 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.251e+02 1.359e+02 1.461e+02 2.063e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-24 19:17:27,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=570112.6666666666, ans=0.025
2024-09-24 19:17:33,282 INFO [train.py:1198] (1/4) Epoch 32, batch 1400, loss[loss=0.1931, ctc_loss=0.1248, cr_loss=0.3414, over 17058.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1278, cr_loss=0.3438, over 3365300.74 frames. ], batch size: 46, lr: 3.73e-03, grad_scale: 16.0
2024-09-24 19:17:36,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=570159.3333333334, ans=0.0
2024-09-24 19:17:55,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=570206.0, ans=0.125
2024-09-24 19:18:24,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=570299.3333333334, ans=0.125
2024-09-24 19:18:24,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=570299.3333333334, ans=0.125
2024-09-24 19:18:47,716 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.95 vs. limit=15.0
2024-09-24 19:18:49,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=570346.0, ans=0.04949747468305833
2024-09-24 19:18:56,763 INFO [train.py:1198] (1/4) Epoch 32, batch 1450, loss[loss=0.2017, ctc_loss=0.1336, cr_loss=0.3407, over 15916.00 frames. ], tot_loss[loss=0.1976, ctc_loss=0.1286, cr_loss=0.3452, over 3362449.60 frames. ], batch size: 74, lr: 3.73e-03, grad_scale: 16.0
2024-09-24 19:19:09,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=570392.6666666666, ans=0.0
2024-09-24 19:19:37,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=570486.0, ans=0.125
2024-09-24 19:19:48,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=570532.6666666666, ans=0.0
2024-09-24 19:19:52,375 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=13.45 vs. limit=15.0
2024-09-24 19:20:06,916 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.64 vs. limit=15.0
2024-09-24 19:20:07,144 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.21 vs. limit=15.0
2024-09-24 19:20:12,693 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.251e+02 1.341e+02 1.423e+02 1.681e+02, threshold=2.682e+02, percent-clipped=0.0
2024-09-24 19:20:21,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=570579.3333333334, ans=0.1
2024-09-24 19:20:23,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=570626.0, ans=0.2
2024-09-24 19:20:24,787 INFO [train.py:1198] (1/4) Epoch 32, batch 1500, loss[loss=0.2193, ctc_loss=0.1413, cr_loss=0.3904, over 17215.00 frames. ], tot_loss[loss=0.1988, ctc_loss=0.1294, cr_loss=0.3466, over 3348301.52 frames. ], batch size: 47, lr: 3.73e-03, grad_scale: 16.0
2024-09-24 19:20:52,628 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.63 vs. limit=6.0
2024-09-24 19:20:52,654 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.27 vs. limit=12.0
2024-09-24 19:21:22,826 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.15 vs. limit=15.0
2024-09-24 19:21:44,251 INFO [train.py:1198] (1/4) Epoch 32, batch 1550, loss[loss=0.164, ctc_loss=0.1033, cr_loss=0.3034, over 16303.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1297, cr_loss=0.347, over 3345438.92 frames. ], batch size: 36, lr: 3.73e-03, grad_scale: 16.0
2024-09-24 19:21:57,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=570859.3333333334, ans=0.1
2024-09-24 19:22:28,926 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.36 vs. limit=22.5
2024-09-24 19:22:51,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=571046.0, ans=0.1
2024-09-24 19:22:56,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=571046.0, ans=0.1
2024-09-24 19:22:57,880 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.269e+02 1.362e+02 1.460e+02 2.798e+02, threshold=2.724e+02, percent-clipped=1.0
2024-09-24 19:23:06,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=571092.6666666666, ans=0.0
2024-09-24 19:23:07,637 INFO [train.py:1198] (1/4) Epoch 32, batch 1600, loss[loss=0.1815, ctc_loss=0.1168, cr_loss=0.3232, over 17088.00 frames. ], tot_loss[loss=0.1986, ctc_loss=0.1293, cr_loss=0.3468, over 3355691.40 frames. ], batch size: 43, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:23:41,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=571186.0, ans=0.125
2024-09-24 19:24:01,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=571232.6666666666, ans=0.125
2024-09-24 19:24:02,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=571232.6666666666, ans=0.1
2024-09-24 19:24:23,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=571279.3333333334, ans=0.5
2024-09-24 19:24:23,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=571279.3333333334, ans=0.125
2024-09-24 19:24:27,902 INFO [train.py:1198] (1/4) Epoch 32, batch 1650, loss[loss=0.1766, ctc_loss=0.1136, cr_loss=0.3149, over 17207.00 frames. ], tot_loss[loss=0.1992, ctc_loss=0.1297, cr_loss=0.3475, over 3360748.07 frames. ], batch size: 41, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:24:49,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=571372.6666666666, ans=0.1
2024-09-24 19:25:00,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=571372.6666666666, ans=0.1
2024-09-24 19:25:02,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.whiten.whitening_limit, batch_count=571372.6666666666, ans=12.0
2024-09-24 19:25:02,966 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=14.09 vs. limit=15.0
2024-09-24 19:25:04,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=571419.3333333334, ans=0.125
2024-09-24 19:25:25,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=571466.0, ans=0.1
2024-09-24 19:25:33,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=571466.0, ans=0.0
2024-09-24 19:25:46,145 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.279e+02 1.364e+02 1.457e+02 2.649e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-24 19:25:55,639 INFO [train.py:1198] (1/4) Epoch 32, batch 1700, loss[loss=0.1957, ctc_loss=0.1273, cr_loss=0.3421, over 17012.00 frames. ], tot_loss[loss=0.199, ctc_loss=0.1296, cr_loss=0.3471, over 3357395.32 frames. ], batch size: 44, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:26:15,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=571606.0, ans=0.2
2024-09-24 19:26:29,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=571652.6666666666, ans=0.0
2024-09-24 19:26:39,500 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.55 vs. limit=22.5
2024-09-24 19:26:56,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=571699.3333333334, ans=0.125
2024-09-24 19:27:06,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.min_positive, batch_count=571746.0, ans=0.05
2024-09-24 19:27:15,587 INFO [train.py:1198] (1/4) Epoch 32, batch 1750, loss[loss=0.2065, ctc_loss=0.1343, cr_loss=0.3614, over 17307.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1299, cr_loss=0.3474, over 3352227.18 frames. ], batch size: 51, lr: 3.73e-03, grad_scale: 32.0
2024-09-24 19:27:35,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=571839.3333333334, ans=0.125
2024-09-24 19:27:58,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=571886.0, ans=0.125
2024-09-24 19:28:21,043 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.69 vs. limit=15.0
2024-09-24 19:28:22,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=571979.3333333334, ans=0.0
2024-09-24 19:28:28,286 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.257e+02 1.347e+02 1.444e+02 1.830e+02, threshold=2.693e+02, percent-clipped=0.0
2024-09-24 19:28:37,947 INFO [train.py:1198] (1/4) Epoch 32, batch 1800, loss[loss=0.189, ctc_loss=0.1204, cr_loss=0.3431, over 17209.00 frames. ], tot_loss[loss=0.1992, ctc_loss=0.1296, cr_loss=0.3477, over 3361265.85 frames. ], batch size: 47, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:28:39,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=572026.0, ans=0.1
2024-09-24 19:28:57,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=572072.6666666666, ans=0.0
2024-09-24 19:29:04,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=572072.6666666666, ans=0.1
2024-09-24 19:29:05,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=572072.6666666666, ans=0.125
2024-09-24 19:29:07,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=572072.6666666666, ans=0.025
2024-09-24 19:29:11,232 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.66 vs. limit=15.0
2024-09-24 19:29:28,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=572166.0, ans=0.04949747468305833
2024-09-24 19:29:29,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=572166.0, ans=0.125
2024-09-24 19:29:46,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.96 vs. limit=12.0
2024-09-24 19:29:55,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=572212.6666666666, ans=0.125
2024-09-24 19:30:02,991 INFO [train.py:1198] (1/4) Epoch 32, batch 1850, loss[loss=0.1537, ctc_loss=0.09676, cr_loss=0.2849, over 17081.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1299, cr_loss=0.3478, over 3359280.63 frames. ], batch size: 43, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:30:04,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=572259.3333333334, ans=0.0
2024-09-24 19:30:04,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=572259.3333333334, ans=0.025
2024-09-24 19:30:09,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=572259.3333333334, ans=0.125
2024-09-24 19:30:44,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=572352.6666666666, ans=0.1
2024-09-24 19:30:52,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=572399.3333333334, ans=0.125
2024-09-24 19:30:52,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=572399.3333333334, ans=0.0
2024-09-24 19:30:58,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=572399.3333333334, ans=0.0
2024-09-24 19:30:59,089 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=14.59 vs. limit=22.5
2024-09-24 19:31:17,413 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.271e+02 1.345e+02 1.449e+02 2.207e+02, threshold=2.691e+02, percent-clipped=0.0
2024-09-24 19:31:25,383 INFO [train.py:1198] (1/4) Epoch 32, batch 1900, loss[loss=0.209, ctc_loss=0.1362, cr_loss=0.3641, over 16987.00 frames. ], tot_loss[loss=0.1992, ctc_loss=0.1296, cr_loss=0.3477, over 3356950.31 frames. ], batch size: 53, lr: 3.72e-03, grad_scale: 16.0
2024-09-24 19:31:33,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=572492.6666666666, ans=0.0
2024-09-24 19:31:40,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=572539.3333333334, ans=0.0
2024-09-24 19:32:36,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=572679.3333333334, ans=0.125
2024-09-24 19:32:47,861 INFO [train.py:1198] (1/4) Epoch 32, batch 1950, loss[loss=0.198, ctc_loss=0.1294, cr_loss=0.3428, over 17226.00 frames. ], tot_loss[loss=0.1992, ctc_loss=0.1297, cr_loss=0.3475, over 3356640.12 frames. ], batch size: 50, lr: 3.72e-03, grad_scale: 16.0
2024-09-24 19:33:12,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=572772.6666666666, ans=0.125
2024-09-24 19:33:23,536 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.49 vs. limit=22.5
2024-09-24 19:33:36,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=572866.0, ans=0.0
2024-09-24 19:33:51,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys.whitening_limit, batch_count=572912.6666666666, ans=6.0
2024-09-24 19:34:00,277 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.299e+02 1.379e+02 1.509e+02 1.987e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-24 19:34:08,190 INFO [train.py:1198] (1/4) Epoch 32, batch 2000, loss[loss=0.2178, ctc_loss=0.1435, cr_loss=0.3719, over 16985.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1294, cr_loss=0.3467, over 3355008.38 frames. ], batch size: 53, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:35:01,389 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.84 vs. limit=6.0
2024-09-24 19:35:12,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=573099.3333333334, ans=0.05
2024-09-24 19:35:24,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=573146.0, ans=0.0
2024-09-24 19:35:33,440 INFO [train.py:1198] (1/4) Epoch 32, batch 2050, loss[loss=0.217, ctc_loss=0.1403, cr_loss=0.3834, over 17302.00 frames. ], tot_loss[loss=0.2002, ctc_loss=0.1304, cr_loss=0.3486, over 3351514.86 frames. ], batch size: 46, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:35:33,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=573192.6666666666, ans=0.125
2024-09-24 19:36:17,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=573286.0, ans=0.125
2024-09-24 19:36:41,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=573379.3333333334, ans=0.125
2024-09-24 19:36:45,681 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.270e+02 1.344e+02 1.458e+02 2.417e+02, threshold=2.689e+02, percent-clipped=0.0
2024-09-24 19:36:53,624 INFO [train.py:1198] (1/4) Epoch 32, batch 2100, loss[loss=0.1949, ctc_loss=0.1234, cr_loss=0.3575, over 17245.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1307, cr_loss=0.3496, over 3355982.60 frames. ], batch size: 44, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:37:42,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=573566.0, ans=0.1
2024-09-24 19:38:15,960 INFO [train.py:1198] (1/4) Epoch 32, batch 2150, loss[loss=0.1748, ctc_loss=0.1144, cr_loss=0.3017, over 17020.00 frames. ], tot_loss[loss=0.1988, ctc_loss=0.1293, cr_loss=0.3478, over 3364373.63 frames. ], batch size: 51, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:38:36,871 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 19:38:36,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=573706.0, ans=0.2
2024-09-24 19:38:44,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=573706.0, ans=0.2
2024-09-24 19:38:48,901 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.44 vs. limit=8.0
2024-09-24 19:38:55,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=573752.6666666666, ans=0.0
2024-09-24 19:39:33,826 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.018e+02 1.275e+02 1.342e+02 1.446e+02 2.934e+02, threshold=2.683e+02, percent-clipped=1.0
2024-09-24 19:39:40,379 INFO [train.py:1198] (1/4) Epoch 32, batch 2200, loss[loss=0.1799, ctc_loss=0.1156, cr_loss=0.3213, over 17168.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.129, cr_loss=0.3471, over 3362529.78 frames. ], batch size: 45, lr: 3.72e-03, grad_scale: 16.0
2024-09-24 19:39:47,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=573892.6666666666, ans=0.025
2024-09-24 19:40:18,812 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.53 vs. limit=15.0
2024-09-24 19:40:23,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=573986.0, ans=0.125
2024-09-24 19:40:48,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten.whitening_limit, batch_count=574079.3333333334, ans=15.0
2024-09-24 19:41:02,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=574126.0, ans=0.125
2024-09-24 19:41:03,500 INFO [train.py:1198] (1/4) Epoch 32, batch 2250, loss[loss=0.2002, ctc_loss=0.1313, cr_loss=0.3443, over 17017.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1274, cr_loss=0.3442, over 3362059.47 frames. ], batch size: 51, lr: 3.72e-03, grad_scale: 16.0
2024-09-24 19:41:16,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=574126.0, ans=0.125
2024-09-24 19:41:40,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=574219.3333333334, ans=0.125
2024-09-24 19:41:41,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=574219.3333333334, ans=0.2
2024-09-24 19:42:16,564 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.263e+02 1.326e+02 1.428e+02 1.960e+02, threshold=2.652e+02, percent-clipped=0.0
2024-09-24 19:42:22,902 INFO [train.py:1198] (1/4) Epoch 32, batch 2300, loss[loss=0.2072, ctc_loss=0.1389, cr_loss=0.3413, over 17297.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1276, cr_loss=0.3442, over 3364709.41 frames. ], batch size: 51, lr: 3.72e-03, grad_scale: 16.0
2024-09-24 19:42:48,787 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.46 vs. limit=15.0
2024-09-24 19:42:56,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=574452.6666666666, ans=0.2
2024-09-24 19:43:09,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.85 vs. limit=10.0
2024-09-24 19:43:17,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.max_abs, batch_count=574499.3333333334, ans=10.0
2024-09-24 19:43:18,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=574499.3333333334, ans=0.125
2024-09-24 19:43:41,705 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.92 vs. limit=15.0
2024-09-24 19:43:41,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=574546.0, ans=22.5
2024-09-24 19:43:45,787 INFO [train.py:1198] (1/4) Epoch 32, batch 2350, loss[loss=0.1732, ctc_loss=0.1105, cr_loss=0.314, over 17192.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1275, cr_loss=0.3443, over 3358844.56 frames. ], batch size: 41, lr: 3.72e-03, grad_scale: 16.0
2024-09-24 19:44:00,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=574639.3333333334, ans=0.0
2024-09-24 19:44:00,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=574639.3333333334, ans=0.125
2024-09-24 19:44:10,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=574639.3333333334, ans=0.125
2024-09-24 19:44:16,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=574686.0, ans=0.125
2024-09-24 19:44:25,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=574686.0, ans=0.1
2024-09-24 19:44:32,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=574686.0, ans=0.0
2024-09-24 19:44:33,538 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.02 vs. limit=15.0
2024-09-24 19:45:04,495 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.259e+02 1.339e+02 1.484e+02 2.121e+02, threshold=2.678e+02, percent-clipped=0.0
2024-09-24 19:45:13,561 INFO [train.py:1198] (1/4) Epoch 32, batch 2400, loss[loss=0.1857, ctc_loss=0.1191, cr_loss=0.3332, over 17202.00 frames. ], tot_loss[loss=0.1973, ctc_loss=0.1282, cr_loss=0.3457, over 3341403.03 frames. ], batch size: 41, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:45:28,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=574872.6666666666, ans=0.0
2024-09-24 19:45:47,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=574919.3333333334, ans=0.125
2024-09-24 19:45:53,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=574919.3333333334, ans=0.125
2024-09-24 19:45:56,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=574919.3333333334, ans=0.125
2024-09-24 19:46:24,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=575012.6666666666, ans=0.025
2024-09-24 19:46:26,503 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.28 vs. limit=10.0
2024-09-24 19:46:30,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=575012.6666666666, ans=0.2
2024-09-24 19:46:33,599 INFO [train.py:1198] (1/4) Epoch 32, batch 2450, loss[loss=0.2009, ctc_loss=0.1327, cr_loss=0.341, over 17213.00 frames. ], tot_loss[loss=0.1966, ctc_loss=0.1276, cr_loss=0.3445, over 3349629.83 frames. ], batch size: 47, lr: 3.72e-03, grad_scale: 32.0
2024-09-24 19:46:46,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=575059.3333333334, ans=0.125
2024-09-24 19:47:10,766 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 19:47:37,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=575199.3333333334, ans=0.1
2024-09-24 19:47:48,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=575246.0, ans=0.025
2024-09-24 19:47:49,729 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.250e+02 1.318e+02 1.420e+02 2.123e+02, threshold=2.637e+02, percent-clipped=0.0
2024-09-24 19:47:56,339 INFO [train.py:1198] (1/4) Epoch 32, batch 2500, loss[loss=0.172, ctc_loss=0.1108, cr_loss=0.3062, over 16301.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1274, cr_loss=0.3437, over 3351817.40 frames. ], batch size: 36, lr: 3.71e-03, grad_scale: 32.0
2024-09-24 19:48:05,112 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.94 vs. limit=12.0
2024-09-24 19:48:07,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=575292.6666666666, ans=0.0
2024-09-24 19:48:09,593 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 19:48:32,696 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.60 vs. limit=12.0
2024-09-24 19:49:05,900 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.96 vs. limit=22.5
2024-09-24 19:49:09,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=575479.3333333334, ans=0.125
2024-09-24 19:49:18,663 INFO [train.py:1198] (1/4) Epoch 32, batch 2550, loss[loss=0.1582, ctc_loss=0.1009, cr_loss=0.2864, over 17144.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1271, cr_loss=0.3436, over 3361727.89 frames. ], batch size: 40, lr: 3.71e-03, grad_scale: 32.0
2024-09-24 19:49:36,789 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.84 vs. limit=6.0
2024-09-24 19:49:37,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=575572.6666666666, ans=0.1
2024-09-24 19:49:45,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=575572.6666666666, ans=0.125
2024-09-24 19:49:52,425 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.86 vs. limit=15.0
2024-09-24 19:49:53,621 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=575619.3333333334, ans=0.0
2024-09-24 19:50:00,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=575619.3333333334, ans=0.2
2024-09-24 19:50:06,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=575619.3333333334, ans=0.0
2024-09-24 19:50:39,009 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.044e+02 1.280e+02 1.336e+02 1.446e+02 2.286e+02, threshold=2.672e+02, percent-clipped=0.0
2024-09-24 19:50:43,839 INFO [train.py:1198] (1/4) Epoch 32, batch 2600, loss[loss=0.1801, ctc_loss=0.114, cr_loss=0.3306, over 17260.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1272, cr_loss=0.3438, over 3364102.78 frames. ], batch size: 44, lr: 3.71e-03, grad_scale: 16.0
2024-09-24 19:51:38,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=575899.3333333334, ans=0.1
2024-09-24 19:51:42,243 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.81 vs. limit=10.0
2024-09-24 19:51:54,806 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.13 vs. limit=12.0
2024-09-24 19:52:03,680 INFO [train.py:1198] (1/4) Epoch 32, batch 2650, loss[loss=0.2102, ctc_loss=0.1372, cr_loss=0.3649, over 17149.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1274, cr_loss=0.344, over 3372361.56 frames. ], batch size: 45, lr: 3.71e-03, grad_scale: 16.0
2024-09-24 19:52:54,298 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=16.44 vs. limit=22.5
2024-09-24 19:53:01,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=576132.6666666666, ans=0.1
2024-09-24 19:53:16,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=576179.3333333334, ans=0.125
2024-09-24 19:53:18,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=576179.3333333334, ans=0.1
2024-09-24 19:53:22,461 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.282e+02 1.356e+02 1.452e+02 2.905e+02, threshold=2.711e+02, percent-clipped=3.0
2024-09-24 19:53:25,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=576226.0, ans=0.025
2024-09-24 19:53:27,413 INFO [train.py:1198] (1/4) Epoch 32, batch 2700, loss[loss=0.1843, ctc_loss=0.1187, cr_loss=0.3278, over 17293.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.128, cr_loss=0.3447, over 3368598.32 frames. ], batch size: 49, lr: 3.71e-03, grad_scale: 16.0
2024-09-24 19:53:29,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=576226.0, ans=0.125
2024-09-24 19:53:38,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=576226.0, ans=0.1
2024-09-24 19:53:47,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=576272.6666666666, ans=0.0
2024-09-24 19:54:17,972 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.98 vs. limit=15.0
2024-09-24 19:54:22,354 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.79 vs. limit=6.0
2024-09-24 19:54:26,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=576366.0, ans=0.1
2024-09-24 19:54:50,197 INFO [train.py:1198] (1/4) Epoch 32, batch 2750, loss[loss=0.1685, ctc_loss=0.1091, cr_loss=0.2972, over 17051.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1272, cr_loss=0.3434, over 3370907.20 frames. ], batch size: 39, lr: 3.71e-03, grad_scale: 16.0
2024-09-24 19:55:21,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=576506.0, ans=0.09899494936611666
2024-09-24 19:56:08,200 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.294e+02 1.403e+02 1.500e+02 1.947e+02, threshold=2.805e+02, percent-clipped=0.0
2024-09-24 19:56:08,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=576646.0, ans=0.015
2024-09-24 19:56:11,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=576692.6666666666, ans=0.1
2024-09-24 19:56:13,016 INFO [train.py:1198] (1/4) Epoch 32, batch 2800, loss[loss=0.1412, ctc_loss=0.0868, cr_loss=0.272, over 17097.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1269, cr_loss=0.3425, over 3371614.35 frames. ], batch size: 43, lr: 3.71e-03, grad_scale: 32.0
2024-09-24 19:56:13,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=576692.6666666666, ans=0.125
2024-09-24 19:56:52,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=576786.0, ans=0.1
2024-09-24 19:56:55,304 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 19:57:06,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=576832.6666666666, ans=0.125
2024-09-24 19:57:08,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=576832.6666666666, ans=0.125
2024-09-24 19:57:13,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=576832.6666666666, ans=0.125
2024-09-24 19:57:36,057 INFO [train.py:1198] (1/4) Epoch 32, batch 2850, loss[loss=0.193, ctc_loss=0.1259, cr_loss=0.3352, over 15912.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1266, cr_loss=0.3421, over 3368170.88 frames. ], batch size: 35, lr: 3.71e-03, grad_scale: 32.0
2024-09-24 19:57:38,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.61 vs. limit=22.5
2024-09-24 19:57:47,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=576926.0, ans=0.2
2024-09-24 19:57:58,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=576972.6666666666, ans=0.1
2024-09-24 19:57:58,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=576972.6666666666, ans=0.125
2024-09-24 19:58:17,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=577019.3333333334, ans=0.2
2024-09-24 19:58:33,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=577066.0, ans=0.1
2024-09-24 19:58:51,150 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.250e+02 1.340e+02 1.459e+02 3.976e+02, threshold=2.680e+02, percent-clipped=1.0
2024-09-24 19:58:53,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=577112.6666666666, ans=0.0
2024-09-24 19:58:55,897 INFO [train.py:1198] (1/4) Epoch 32, batch 2900, loss[loss=0.2227, ctc_loss=0.1458, cr_loss=0.3847, over 17225.00 frames. ], tot_loss[loss=0.1956, ctc_loss=0.127, cr_loss=0.3429, over 3370639.12 frames. ], batch size: 50, lr: 3.71e-03, grad_scale: 32.0
2024-09-24 19:58:57,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=577159.3333333334, ans=0.125
2024-09-24 19:59:02,954 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.60 vs. limit=15.0
2024-09-24 19:59:31,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=577252.6666666666, ans=0.125
2024-09-24 19:59:46,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=577252.6666666666, ans=0.07
2024-09-24 19:59:48,055 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 20:00:02,573 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.82 vs. limit=15.0
2024-09-24 20:00:14,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=577346.0, ans=0.0
2024-09-24 20:00:23,209 INFO [train.py:1198] (1/4) Epoch 32, batch 2950, loss[loss=0.1737, ctc_loss=0.1095, cr_loss=0.3211, over 17188.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1274, cr_loss=0.3439, over 3370606.28 frames. ], batch size: 41, lr: 3.71e-03, grad_scale: 32.0
2024-09-24 20:00:25,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=577392.6666666666, ans=0.2
2024-09-24 20:00:39,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=577439.3333333334, ans=0.1
2024-09-24 20:00:57,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=577486.0, ans=0.125
2024-09-24 20:01:06,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=577486.0, ans=0.125
2024-09-24 20:01:33,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=577579.3333333334, ans=0.125
2024-09-24 20:01:36,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=577579.3333333334, ans=0.025
2024-09-24 20:01:39,276 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.244e+02 1.340e+02 1.434e+02 1.849e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-24 20:01:42,521 INFO [train.py:1198] (1/4) Epoch 32, batch 3000, loss[loss=0.187, ctc_loss=0.1211, cr_loss=0.3294, over 16943.00 frames. ], tot_loss[loss=0.1971, ctc_loss=0.1281, cr_loss=0.345, over 3370122.25 frames. ], batch size: 58, lr: 3.71e-03, grad_scale: 16.0
2024-09-24 20:01:42,521 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 20:01:51,724 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.5.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([3.5112, 5.2119, 4.9702, 5.2021], device='cuda:1')
2024-09-24 20:01:52,672 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.5.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([2.4193, 4.0327, 3.5045, 4.4410], device='cuda:1')
2024-09-24 20:01:57,949 INFO [train.py:1230] (1/4) Epoch 32, validation: loss=0.03608, ctc_loss=0.03608, cr_loss=9.027e-15, over 944034.00 frames. 
2024-09-24 20:01:57,949 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 20:02:18,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=577672.6666666666, ans=0.125
2024-09-24 20:02:23,839 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.41 vs. limit=15.0
2024-09-24 20:02:55,591 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.83 vs. limit=15.0
2024-09-24 20:03:12,635 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.01 vs. limit=6.0
2024-09-24 20:03:19,565 INFO [train.py:1198] (1/4) Epoch 32, batch 3050, loss[loss=0.1675, ctc_loss=0.1058, cr_loss=0.3083, over 17253.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1278, cr_loss=0.3443, over 3367900.22 frames. ], batch size: 44, lr: 3.71e-03, grad_scale: 16.0
2024-09-24 20:03:26,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=577859.3333333334, ans=0.125
2024-09-24 20:03:38,970 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.04 vs. limit=10.0
2024-09-24 20:03:44,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=577906.0, ans=0.125
2024-09-24 20:04:34,696 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.228e+02 1.314e+02 1.465e+02 2.452e+02, threshold=2.627e+02, percent-clipped=0.0
2024-09-24 20:04:37,214 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.45 vs. limit=22.5
2024-09-24 20:04:37,879 INFO [train.py:1198] (1/4) Epoch 32, batch 3100, loss[loss=0.1734, ctc_loss=0.1131, cr_loss=0.3013, over 17027.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1279, cr_loss=0.3449, over 3372461.43 frames. ], batch size: 44, lr: 3.71e-03, grad_scale: 16.0
2024-09-24 20:04:59,595 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=8.64 vs. limit=22.5
2024-09-24 20:05:00,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=578139.3333333334, ans=0.125
2024-09-24 20:05:05,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=578139.3333333334, ans=0.0
2024-09-24 20:05:25,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=578232.6666666666, ans=0.1
2024-09-24 20:05:25,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=578232.6666666666, ans=0.125
2024-09-24 20:05:25,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=578232.6666666666, ans=0.1
2024-09-24 20:05:47,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=578279.3333333334, ans=0.0
2024-09-24 20:05:56,195 INFO [train.py:1198] (1/4) Epoch 32, batch 3150, loss[loss=0.1986, ctc_loss=0.1304, cr_loss=0.3411, over 16992.00 frames. ], tot_loss[loss=0.1975, ctc_loss=0.1283, cr_loss=0.3458, over 3376447.67 frames. ], batch size: 56, lr: 3.70e-03, grad_scale: 16.0
2024-09-24 20:06:06,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=578326.0, ans=0.125
2024-09-24 20:06:16,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=578372.6666666666, ans=0.0
2024-09-24 20:07:12,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=578512.6666666666, ans=0.1
2024-09-24 20:07:15,658 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.274e+02 1.414e+02 1.514e+02 2.013e+02, threshold=2.828e+02, percent-clipped=0.0
2024-09-24 20:07:18,910 INFO [train.py:1198] (1/4) Epoch 32, batch 3200, loss[loss=0.1777, ctc_loss=0.1099, cr_loss=0.3388, over 17021.00 frames. ], tot_loss[loss=0.1977, ctc_loss=0.1284, cr_loss=0.3462, over 3373591.49 frames. ], batch size: 44, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:07:53,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=578652.6666666666, ans=0.125
2024-09-24 20:08:05,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=578652.6666666666, ans=0.125
2024-09-24 20:08:19,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=578699.3333333334, ans=0.125
2024-09-24 20:08:30,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=578746.0, ans=15.0
2024-09-24 20:08:37,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=578792.6666666666, ans=0.125
2024-09-24 20:08:39,227 INFO [train.py:1198] (1/4) Epoch 32, batch 3250, loss[loss=0.17, ctc_loss=0.1074, cr_loss=0.3131, over 17087.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1278, cr_loss=0.3445, over 3373696.42 frames. ], batch size: 43, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:08:40,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=578792.6666666666, ans=0.0
2024-09-24 20:09:03,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=578839.3333333334, ans=0.0
2024-09-24 20:09:12,517 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.68 vs. limit=6.0
2024-09-24 20:09:25,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=578886.0, ans=0.125
2024-09-24 20:09:26,649 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=12.34 vs. limit=15.0
2024-09-24 20:09:56,536 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.279e+02 1.356e+02 1.447e+02 1.834e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-24 20:09:59,775 INFO [train.py:1198] (1/4) Epoch 32, batch 3300, loss[loss=0.1976, ctc_loss=0.1272, cr_loss=0.3516, over 17041.00 frames. ], tot_loss[loss=0.1971, ctc_loss=0.1281, cr_loss=0.3449, over 3369228.79 frames. ], batch size: 56, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:10:19,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.55 vs. limit=8.0
2024-09-24 20:10:28,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=579072.6666666666, ans=0.125
2024-09-24 20:10:42,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=579119.3333333334, ans=0.0
2024-09-24 20:11:05,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=579212.6666666666, ans=0.0
2024-09-24 20:11:13,176 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 20:11:17,459 INFO [train.py:1198] (1/4) Epoch 32, batch 3350, loss[loss=0.2119, ctc_loss=0.1395, cr_loss=0.3622, over 17307.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1282, cr_loss=0.3447, over 3363505.04 frames. ], batch size: 46, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:11:36,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=579306.0, ans=10.0
2024-09-24 20:12:02,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=579399.3333333334, ans=0.125
2024-09-24 20:12:07,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=579399.3333333334, ans=0.0
2024-09-24 20:12:23,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=579446.0, ans=0.125
2024-09-24 20:12:23,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=579446.0, ans=0.2
2024-09-24 20:12:32,489 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.280e+02 1.354e+02 1.467e+02 2.390e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-24 20:12:35,626 INFO [train.py:1198] (1/4) Epoch 32, batch 3400, loss[loss=0.2137, ctc_loss=0.1424, cr_loss=0.3563, over 17345.00 frames. ], tot_loss[loss=0.1971, ctc_loss=0.1282, cr_loss=0.3445, over 3365558.84 frames. ], batch size: 48, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:12:57,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=579539.3333333334, ans=0.125
2024-09-24 20:13:01,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=579539.3333333334, ans=0.125
2024-09-24 20:13:12,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=579586.0, ans=0.125
2024-09-24 20:13:13,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=579586.0, ans=0.0
2024-09-24 20:13:21,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=579632.6666666666, ans=0.1
2024-09-24 20:13:54,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=579726.0, ans=0.125
2024-09-24 20:13:55,923 INFO [train.py:1198] (1/4) Epoch 32, batch 3450, loss[loss=0.1885, ctc_loss=0.1238, cr_loss=0.3235, over 17351.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1278, cr_loss=0.3433, over 3356134.01 frames. ], batch size: 48, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:14:00,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=579726.0, ans=0.2
2024-09-24 20:14:16,914 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.44 vs. limit=12.0
2024-09-24 20:14:24,129 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 20:14:38,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=579819.3333333334, ans=0.1
2024-09-24 20:14:48,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=579866.0, ans=0.025
2024-09-24 20:14:55,417 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 20:15:03,025 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 20:15:10,635 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.299e+02 1.417e+02 1.544e+02 2.266e+02, threshold=2.834e+02, percent-clipped=0.0
2024-09-24 20:15:13,759 INFO [train.py:1198] (1/4) Epoch 32, batch 3500, loss[loss=0.2491, ctc_loss=0.1697, cr_loss=0.3973, over 15984.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1281, cr_loss=0.3433, over 3345828.19 frames. ], batch size: 74, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:15:55,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=580052.6666666666, ans=0.0
2024-09-24 20:16:03,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten.whitening_limit, batch_count=580099.3333333334, ans=15.0
2024-09-24 20:16:29,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=580146.0, ans=0.125
2024-09-24 20:16:32,169 INFO [train.py:1198] (1/4) Epoch 32, batch 3550, loss[loss=0.218, ctc_loss=0.1398, cr_loss=0.3908, over 17299.00 frames. ], tot_loss[loss=0.198, ctc_loss=0.1289, cr_loss=0.3453, over 3341267.93 frames. ], batch size: 51, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:16:52,846 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.57 vs. limit=5.0
2024-09-24 20:16:56,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=580239.3333333334, ans=0.125
2024-09-24 20:16:59,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=580239.3333333334, ans=0.1
2024-09-24 20:17:20,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=580286.0, ans=0.1
2024-09-24 20:17:22,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=580332.6666666666, ans=0.125
2024-09-24 20:17:45,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=580379.3333333334, ans=0.2
2024-09-24 20:17:48,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=580379.3333333334, ans=0.125
2024-09-24 20:17:51,751 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.264e+02 1.345e+02 1.464e+02 2.321e+02, threshold=2.690e+02, percent-clipped=0.0
2024-09-24 20:17:54,936 INFO [train.py:1198] (1/4) Epoch 32, batch 3600, loss[loss=0.2116, ctc_loss=0.1388, cr_loss=0.3643, over 16586.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1282, cr_loss=0.3448, over 3344906.71 frames. ], batch size: 66, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:18:08,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=8.14 vs. limit=12.0
2024-09-24 20:18:14,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=580472.6666666666, ans=0.1
2024-09-24 20:18:28,923 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.87 vs. limit=15.0
2024-09-24 20:18:29,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=580519.3333333334, ans=0.125
2024-09-24 20:18:52,719 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.84 vs. limit=6.0
2024-09-24 20:19:09,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=580612.6666666666, ans=0.125
2024-09-24 20:19:15,282 INFO [train.py:1198] (1/4) Epoch 32, batch 3650, loss[loss=0.209, ctc_loss=0.1397, cr_loss=0.3465, over 16878.00 frames. ], tot_loss[loss=0.1991, ctc_loss=0.1296, cr_loss=0.3473, over 3332627.66 frames. ], batch size: 58, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:19:21,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=580659.3333333334, ans=0.025
2024-09-24 20:19:31,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=580706.0, ans=0.125
2024-09-24 20:19:37,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=580706.0, ans=0.0
2024-09-24 20:19:37,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten.whitening_limit, batch_count=580706.0, ans=15.0
2024-09-24 20:19:42,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=580706.0, ans=0.1
2024-09-24 20:19:53,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=580752.6666666666, ans=0.125
2024-09-24 20:19:54,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=580752.6666666666, ans=0.015
2024-09-24 20:20:11,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=580799.3333333334, ans=0.1
2024-09-24 20:20:16,909 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.80 vs. limit=15.0
2024-09-24 20:20:22,750 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.81 vs. limit=12.0
2024-09-24 20:20:30,685 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.254e+02 1.367e+02 1.508e+02 1.700e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-24 20:20:33,725 INFO [train.py:1198] (1/4) Epoch 32, batch 3700, loss[loss=0.2017, ctc_loss=0.1317, cr_loss=0.3498, over 16387.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1304, cr_loss=0.3486, over 3337574.12 frames. ], batch size: 66, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:20:50,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.93 vs. limit=6.0
2024-09-24 20:21:01,539 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.81 vs. limit=6.0
2024-09-24 20:21:51,884 INFO [train.py:1198] (1/4) Epoch 32, batch 3750, loss[loss=0.2089, ctc_loss=0.1388, cr_loss=0.3503, over 16000.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.13, cr_loss=0.3474, over 3327453.64 frames. ], batch size: 74, lr: 3.70e-03, grad_scale: 32.0
2024-09-24 20:22:04,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=581126.0, ans=0.125
2024-09-24 20:22:04,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=581126.0, ans=0.1
2024-09-24 20:22:51,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=581266.0, ans=0.0
2024-09-24 20:22:54,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=581312.6666666666, ans=0.0
2024-09-24 20:22:56,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=581312.6666666666, ans=0.125
2024-09-24 20:23:06,879 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.283e+02 1.375e+02 1.482e+02 1.854e+02, threshold=2.749e+02, percent-clipped=0.0
2024-09-24 20:23:10,821 INFO [train.py:1198] (1/4) Epoch 32, batch 3800, loss[loss=0.2333, ctc_loss=0.1556, cr_loss=0.3887, over 15137.00 frames. ], tot_loss[loss=0.1999, ctc_loss=0.1304, cr_loss=0.3478, over 3315165.46 frames. ], batch size: 89, lr: 3.69e-03, grad_scale: 32.0
2024-09-24 20:23:15,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=581359.3333333334, ans=0.0
2024-09-24 20:23:24,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=581406.0, ans=0.125
2024-09-24 20:23:29,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=581406.0, ans=0.025
2024-09-24 20:23:32,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=581406.0, ans=0.07
2024-09-24 20:24:03,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=581499.3333333334, ans=0.0
2024-09-24 20:24:04,500 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.73 vs. limit=6.0
2024-09-24 20:24:07,457 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=9.38 vs. limit=22.5
2024-09-24 20:24:20,960 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.85 vs. limit=22.5
2024-09-24 20:24:28,214 INFO [train.py:1198] (1/4) Epoch 32, batch 3850, loss[loss=0.2155, ctc_loss=0.1426, cr_loss=0.3647, over 16920.00 frames. ], tot_loss[loss=0.2019, ctc_loss=0.1319, cr_loss=0.3498, over 3277247.76 frames. ], batch size: 58, lr: 3.69e-03, grad_scale: 32.0
2024-09-24 20:24:51,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=581639.3333333334, ans=0.0
2024-09-24 20:24:53,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=16.10 vs. limit=22.5
2024-09-24 20:24:59,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=581686.0, ans=0.125
2024-09-24 20:25:03,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=581686.0, ans=0.09899494936611666
2024-09-24 20:25:23,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=581732.6666666666, ans=0.0
2024-09-24 20:25:33,239 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.74 vs. limit=15.0
2024-09-24 20:25:33,495 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.61 vs. limit=22.5
2024-09-24 20:25:36,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.31 vs. limit=6.0
2024-09-24 20:26:29,988 INFO [train.py:1198] (1/4) Epoch 33, batch 0, loss[loss=0.1666, ctc_loss=0.1045, cr_loss=0.3104, over 16678.00 frames. ], tot_loss[loss=0.1666, ctc_loss=0.1045, cr_loss=0.3104, over 16678.00 frames. ], batch size: 37, lr: 3.64e-03, grad_scale: 32.0
2024-09-24 20:26:29,989 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 20:26:46,729 INFO [train.py:1230] (1/4) Epoch 33, validation: loss=0.03608, ctc_loss=0.03608, cr_loss=9.001e-15, over 944034.00 frames. 
2024-09-24 20:26:46,730 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 20:26:52,642 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.455e+02 1.559e+02 1.655e+02 2.375e+02, threshold=3.119e+02, percent-clipped=0.0
2024-09-24 20:26:57,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=581807.3333333334, ans=0.035
2024-09-24 20:27:20,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=581900.6666666666, ans=0.125
2024-09-24 20:27:30,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=581900.6666666666, ans=0.125
2024-09-24 20:27:36,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=581947.3333333334, ans=0.0
2024-09-24 20:27:42,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=581947.3333333334, ans=0.1
2024-09-24 20:27:52,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=581994.0, ans=0.2
2024-09-24 20:27:55,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2.whitening_limit, batch_count=581994.0, ans=15.0
2024-09-24 20:28:09,435 INFO [train.py:1198] (1/4) Epoch 33, batch 50, loss[loss=0.1888, ctc_loss=0.1223, cr_loss=0.3326, over 17286.00 frames. ], tot_loss[loss=0.2011, ctc_loss=0.131, cr_loss=0.3505, over 753838.80 frames. ], batch size: 46, lr: 3.64e-03, grad_scale: 32.0
2024-09-24 20:28:28,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=582087.3333333334, ans=0.125
2024-09-24 20:28:47,788 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.73 vs. limit=22.5
2024-09-24 20:29:13,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=582180.6666666666, ans=0.0
2024-09-24 20:29:19,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=582227.3333333334, ans=0.125
2024-09-24 20:29:31,539 INFO [train.py:1198] (1/4) Epoch 33, batch 100, loss[loss=0.2054, ctc_loss=0.133, cr_loss=0.362, over 17067.00 frames. ], tot_loss[loss=0.2005, ctc_loss=0.1306, cr_loss=0.3499, over 1324065.97 frames. ], batch size: 46, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:29:34,671 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.274e+02 1.348e+02 1.462e+02 2.671e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-24 20:29:39,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=582274.0, ans=0.0
2024-09-24 20:29:56,673 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.78 vs. limit=15.0
2024-09-24 20:29:59,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=582320.6666666666, ans=0.125
2024-09-24 20:30:48,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=582460.6666666666, ans=0.0
2024-09-24 20:30:54,480 INFO [train.py:1198] (1/4) Epoch 33, batch 150, loss[loss=0.2079, ctc_loss=0.1378, cr_loss=0.3505, over 17125.00 frames. ], tot_loss[loss=0.2007, ctc_loss=0.1308, cr_loss=0.3495, over 1767150.13 frames. ], batch size: 48, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:31:01,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=582507.3333333334, ans=0.1
2024-09-24 20:31:04,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=582507.3333333334, ans=0.2
2024-09-24 20:31:06,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=582507.3333333334, ans=0.0
2024-09-24 20:31:07,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=582507.3333333334, ans=0.125
2024-09-24 20:32:07,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=582694.0, ans=0.125
2024-09-24 20:32:09,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=582694.0, ans=0.125
2024-09-24 20:32:13,142 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.41 vs. limit=22.5
2024-09-24 20:32:20,260 INFO [train.py:1198] (1/4) Epoch 33, batch 200, loss[loss=0.2193, ctc_loss=0.1439, cr_loss=0.3769, over 17283.00 frames. ], tot_loss[loss=0.1994, ctc_loss=0.1299, cr_loss=0.3477, over 2108134.31 frames. ], batch size: 44, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:32:21,236 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.80 vs. limit=15.0
2024-09-24 20:32:23,423 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.261e+02 1.350e+02 1.462e+02 5.443e+02, threshold=2.700e+02, percent-clipped=2.0
2024-09-24 20:32:23,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=582740.6666666666, ans=0.1
2024-09-24 20:32:23,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=582740.6666666666, ans=0.125
2024-09-24 20:32:25,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=582740.6666666666, ans=0.0
2024-09-24 20:32:30,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=582740.6666666666, ans=0.0
2024-09-24 20:32:32,413 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.27 vs. limit=15.0
2024-09-24 20:32:47,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=582787.3333333334, ans=0.0
2024-09-24 20:32:52,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=582834.0, ans=0.125
2024-09-24 20:32:54,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=582834.0, ans=0.1
2024-09-24 20:33:08,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=582880.6666666666, ans=0.0
2024-09-24 20:33:22,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=582927.3333333334, ans=0.1
2024-09-24 20:33:42,545 INFO [train.py:1198] (1/4) Epoch 33, batch 250, loss[loss=0.184, ctc_loss=0.1174, cr_loss=0.333, over 17146.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1285, cr_loss=0.3465, over 2393370.75 frames. ], batch size: 45, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:33:46,006 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-24 20:34:01,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=583020.6666666666, ans=0.07
2024-09-24 20:34:01,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=583020.6666666666, ans=10.0
2024-09-24 20:34:56,314 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.17 vs. limit=15.0
2024-09-24 20:35:01,849 INFO [train.py:1198] (1/4) Epoch 33, batch 300, loss[loss=0.1514, ctc_loss=0.09618, cr_loss=0.2763, over 17274.00 frames. ], tot_loss[loss=0.1976, ctc_loss=0.1284, cr_loss=0.3463, over 2609013.47 frames. ], batch size: 42, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:35:05,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.274e+02 1.352e+02 1.473e+02 1.783e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-24 20:35:13,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=583207.3333333334, ans=0.2
2024-09-24 20:35:45,897 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.07 vs. limit=22.5
2024-09-24 20:35:57,350 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.90 vs. limit=15.0
2024-09-24 20:36:04,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=583347.3333333334, ans=0.125
2024-09-24 20:36:07,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=583394.0, ans=0.2
2024-09-24 20:36:14,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=583394.0, ans=0.125
2024-09-24 20:36:15,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=583394.0, ans=0.05
2024-09-24 20:36:17,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=583394.0, ans=0.07
2024-09-24 20:36:25,026 INFO [train.py:1198] (1/4) Epoch 33, batch 350, loss[loss=0.2128, ctc_loss=0.1385, cr_loss=0.3714, over 17043.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1283, cr_loss=0.3455, over 2772156.43 frames. ], batch size: 52, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:36:43,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=583440.6666666666, ans=0.125
2024-09-24 20:36:56,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=583487.3333333334, ans=0.0
2024-09-24 20:36:56,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=583487.3333333334, ans=0.125
2024-09-24 20:36:56,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=13.92 vs. limit=15.0
2024-09-24 20:37:12,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=583534.0, ans=0.025
2024-09-24 20:37:50,367 INFO [train.py:1198] (1/4) Epoch 33, batch 400, loss[loss=0.1674, ctc_loss=0.1075, cr_loss=0.2996, over 17082.00 frames. ], tot_loss[loss=0.1963, ctc_loss=0.1277, cr_loss=0.3435, over 2896150.04 frames. ], batch size: 43, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:37:53,545 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.274e+02 1.352e+02 1.470e+02 2.470e+02, threshold=2.705e+02, percent-clipped=0.0
2024-09-24 20:38:00,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=583674.0, ans=0.0
2024-09-24 20:38:56,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=583860.6666666666, ans=0.0
2024-09-24 20:39:02,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.88 vs. limit=15.0
2024-09-24 20:39:12,683 INFO [train.py:1198] (1/4) Epoch 33, batch 450, loss[loss=0.2219, ctc_loss=0.1458, cr_loss=0.3805, over 17037.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1275, cr_loss=0.3435, over 3003490.84 frames. ], batch size: 56, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:39:16,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten.whitening_limit, batch_count=583907.3333333334, ans=15.0
2024-09-24 20:39:29,322 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.73 vs. limit=22.5
2024-09-24 20:40:14,559 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.64 vs. limit=15.0
2024-09-24 20:40:33,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=584140.6666666666, ans=0.125
2024-09-24 20:40:35,212 INFO [train.py:1198] (1/4) Epoch 33, batch 500, loss[loss=0.2123, ctc_loss=0.1397, cr_loss=0.3631, over 16921.00 frames. ], tot_loss[loss=0.1982, ctc_loss=0.1288, cr_loss=0.347, over 3082462.15 frames. ], batch size: 58, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:40:38,396 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.260e+02 1.369e+02 1.441e+02 1.904e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-24 20:40:48,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=584140.6666666666, ans=0.1
2024-09-24 20:41:40,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=584280.6666666666, ans=0.125
2024-09-24 20:42:00,692 INFO [train.py:1198] (1/4) Epoch 33, batch 550, loss[loss=0.2109, ctc_loss=0.1364, cr_loss=0.3723, over 16991.00 frames. ], tot_loss[loss=0.1987, ctc_loss=0.1292, cr_loss=0.3475, over 3140903.17 frames. ], batch size: 53, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:42:29,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=584420.6666666666, ans=0.0
2024-09-24 20:42:31,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=584467.3333333334, ans=0.1
2024-09-24 20:42:39,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=584467.3333333334, ans=0.125
2024-09-24 20:42:45,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=584467.3333333334, ans=0.0
2024-09-24 20:42:45,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=584467.3333333334, ans=0.1
2024-09-24 20:42:52,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=584514.0, ans=0.07
2024-09-24 20:42:54,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.79 vs. limit=22.5
2024-09-24 20:42:58,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=584514.0, ans=0.025
2024-09-24 20:43:08,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=584560.6666666666, ans=0.2
2024-09-24 20:43:20,876 INFO [train.py:1198] (1/4) Epoch 33, batch 600, loss[loss=0.2094, ctc_loss=0.1345, cr_loss=0.3742, over 17088.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.1289, cr_loss=0.3474, over 3187384.06 frames. ], batch size: 46, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:43:26,765 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.269e+02 1.358e+02 1.468e+02 2.109e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-24 20:43:43,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=584654.0, ans=0.125
2024-09-24 20:44:14,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=584747.3333333334, ans=0.04949747468305833
2024-09-24 20:44:21,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=584747.3333333334, ans=0.0
2024-09-24 20:44:21,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=584747.3333333334, ans=0.125
2024-09-24 20:44:43,687 INFO [train.py:1198] (1/4) Epoch 33, batch 650, loss[loss=0.1746, ctc_loss=0.1079, cr_loss=0.3335, over 17089.00 frames. ], tot_loss[loss=0.1973, ctc_loss=0.1282, cr_loss=0.3454, over 3231321.56 frames. ], batch size: 40, lr: 3.63e-03, grad_scale: 32.0
2024-09-24 20:44:44,688 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.63 vs. limit=15.0
2024-09-24 20:44:45,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=584840.6666666666, ans=0.125
2024-09-24 20:44:47,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=584840.6666666666, ans=10.0
2024-09-24 20:45:41,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=584980.6666666666, ans=0.0
2024-09-24 20:45:47,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=584980.6666666666, ans=0.0
2024-09-24 20:45:47,736 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=7.73 vs. limit=15.0
2024-09-24 20:45:53,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=585027.3333333334, ans=0.0
2024-09-24 20:45:55,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=585027.3333333334, ans=0.0
2024-09-24 20:46:06,204 INFO [train.py:1198] (1/4) Epoch 33, batch 700, loss[loss=0.213, ctc_loss=0.141, cr_loss=0.36, over 16757.00 frames. ], tot_loss[loss=0.1982, ctc_loss=0.1288, cr_loss=0.3467, over 3256438.04 frames. ], batch size: 61, lr: 3.63e-03, grad_scale: 16.0
2024-09-24 20:46:10,908 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.247e+02 1.323e+02 1.439e+02 2.225e+02, threshold=2.645e+02, percent-clipped=0.0
2024-09-24 20:46:37,996 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.88 vs. limit=22.5
2024-09-24 20:46:43,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=585167.3333333334, ans=0.125
2024-09-24 20:46:51,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=585167.3333333334, ans=0.0
2024-09-24 20:46:56,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=585167.3333333334, ans=0.125
2024-09-24 20:46:58,955 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.19 vs. limit=22.5
2024-09-24 20:46:59,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=585214.0, ans=0.05
2024-09-24 20:47:01,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=585214.0, ans=0.125
2024-09-24 20:47:19,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=585260.6666666666, ans=0.125
2024-09-24 20:47:19,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=585260.6666666666, ans=0.0
2024-09-24 20:47:31,641 INFO [train.py:1198] (1/4) Epoch 33, batch 750, loss[loss=0.175, ctc_loss=0.1138, cr_loss=0.3057, over 17029.00 frames. ], tot_loss[loss=0.1976, ctc_loss=0.1284, cr_loss=0.346, over 3279257.15 frames. ], batch size: 44, lr: 3.63e-03, grad_scale: 16.0
2024-09-24 20:47:46,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=585354.0, ans=0.125
2024-09-24 20:48:47,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=585494.0, ans=0.1
2024-09-24 20:48:53,865 INFO [train.py:1198] (1/4) Epoch 33, batch 800, loss[loss=0.2042, ctc_loss=0.1338, cr_loss=0.3518, over 17236.00 frames. ], tot_loss[loss=0.1973, ctc_loss=0.1282, cr_loss=0.3457, over 3303464.15 frames. ], batch size: 55, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:48:58,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.263e+02 1.335e+02 1.404e+02 2.398e+02, threshold=2.669e+02, percent-clipped=0.0
2024-09-24 20:49:01,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=585540.6666666666, ans=0.125
2024-09-24 20:49:37,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=585634.0, ans=0.125
2024-09-24 20:49:37,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=585634.0, ans=0.125
2024-09-24 20:49:56,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=585727.3333333334, ans=0.125
2024-09-24 20:50:12,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=585774.0, ans=0.125
2024-09-24 20:50:14,051 INFO [train.py:1198] (1/4) Epoch 33, batch 850, loss[loss=0.1994, ctc_loss=0.1313, cr_loss=0.3403, over 17140.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1273, cr_loss=0.3445, over 3318512.77 frames. ], batch size: 48, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:50:27,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=585774.0, ans=0.2
2024-09-24 20:50:28,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=585820.6666666666, ans=0.2
2024-09-24 20:51:22,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=585960.6666666666, ans=0.125
2024-09-24 20:51:37,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=586007.3333333334, ans=0.015
2024-09-24 20:51:39,163 INFO [train.py:1198] (1/4) Epoch 33, batch 900, loss[loss=0.2277, ctc_loss=0.149, cr_loss=0.3934, over 16098.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1283, cr_loss=0.3456, over 3313948.29 frames. ], batch size: 74, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:51:44,052 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.262e+02 1.327e+02 1.436e+02 1.975e+02, threshold=2.653e+02, percent-clipped=0.0
2024-09-24 20:52:31,152 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.48 vs. limit=15.0
2024-09-24 20:52:31,451 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.27 vs. limit=15.0
2024-09-24 20:52:38,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=586147.3333333334, ans=0.125
2024-09-24 20:52:58,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=586240.6666666666, ans=0.2
2024-09-24 20:52:59,225 INFO [train.py:1198] (1/4) Epoch 33, batch 950, loss[loss=0.2221, ctc_loss=0.1468, cr_loss=0.3764, over 16487.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1285, cr_loss=0.3464, over 3323023.12 frames. ], batch size: 66, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:53:05,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=586240.6666666666, ans=0.0
2024-09-24 20:53:29,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=586287.3333333334, ans=0.1
2024-09-24 20:53:49,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=586380.6666666666, ans=0.125
2024-09-24 20:53:54,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=586380.6666666666, ans=0.1
2024-09-24 20:54:19,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=586474.0, ans=0.125
2024-09-24 20:54:21,205 INFO [train.py:1198] (1/4) Epoch 33, batch 1000, loss[loss=0.1807, ctc_loss=0.1157, cr_loss=0.3248, over 17162.00 frames. ], tot_loss[loss=0.1981, ctc_loss=0.1288, cr_loss=0.3466, over 3337726.93 frames. ], batch size: 41, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:54:26,138 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.300e+02 1.389e+02 1.469e+02 2.721e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-24 20:54:42,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=586520.6666666666, ans=0.125
2024-09-24 20:54:53,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=586567.3333333334, ans=0.035
2024-09-24 20:54:57,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=586567.3333333334, ans=0.125
2024-09-24 20:55:05,750 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.82 vs. limit=15.0
2024-09-24 20:55:44,467 INFO [train.py:1198] (1/4) Epoch 33, batch 1050, loss[loss=0.1942, ctc_loss=0.1226, cr_loss=0.3578, over 17303.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.129, cr_loss=0.3472, over 3345337.55 frames. ], batch size: 46, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:56:25,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=586800.6666666666, ans=0.125
2024-09-24 20:57:09,127 INFO [train.py:1198] (1/4) Epoch 33, batch 1100, loss[loss=0.1679, ctc_loss=0.1067, cr_loss=0.306, over 17205.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.1279, cr_loss=0.3456, over 3355215.33 frames. ], batch size: 41, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:57:10,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=586940.6666666666, ans=0.0
2024-09-24 20:57:11,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=586940.6666666666, ans=0.0
2024-09-24 20:57:13,959 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.249e+02 1.328e+02 1.446e+02 1.774e+02, threshold=2.656e+02, percent-clipped=0.0
2024-09-24 20:57:15,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=586940.6666666666, ans=0.125
2024-09-24 20:57:22,638 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.64 vs. limit=15.0
2024-09-24 20:57:32,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=586987.3333333334, ans=0.0
2024-09-24 20:58:01,463 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.02 vs. limit=6.0
2024-09-24 20:58:07,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=587080.6666666666, ans=0.125
2024-09-24 20:58:21,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=587127.3333333334, ans=0.07
2024-09-24 20:58:22,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=587127.3333333334, ans=0.2
2024-09-24 20:58:32,050 INFO [train.py:1198] (1/4) Epoch 33, batch 1150, loss[loss=0.2114, ctc_loss=0.1379, cr_loss=0.3675, over 17115.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1277, cr_loss=0.3448, over 3353562.35 frames. ], batch size: 49, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:58:41,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=587174.0, ans=0.0
2024-09-24 20:58:46,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=587220.6666666666, ans=0.1
2024-09-24 20:58:48,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=587220.6666666666, ans=0.0
2024-09-24 20:58:48,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=587220.6666666666, ans=0.1
2024-09-24 20:59:00,158 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.02 vs. limit=6.0
2024-09-24 20:59:08,943 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 20:59:16,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=587267.3333333334, ans=0.125
2024-09-24 20:59:48,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=587360.6666666666, ans=0.1
2024-09-24 20:59:51,717 INFO [train.py:1198] (1/4) Epoch 33, batch 1200, loss[loss=0.1537, ctc_loss=0.09649, cr_loss=0.2862, over 17100.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1276, cr_loss=0.3443, over 3349459.61 frames. ], batch size: 40, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 20:59:53,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=587407.3333333334, ans=0.125
2024-09-24 20:59:56,479 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.281e+02 1.363e+02 1.447e+02 2.223e+02, threshold=2.727e+02, percent-clipped=0.0
2024-09-24 21:00:03,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=587407.3333333334, ans=0.1
2024-09-24 21:00:18,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=587454.0, ans=0.0
2024-09-24 21:00:31,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=587500.6666666666, ans=0.125
2024-09-24 21:00:40,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=587547.3333333334, ans=0.125
2024-09-24 21:00:40,874 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.52 vs. limit=22.5
2024-09-24 21:01:13,623 INFO [train.py:1198] (1/4) Epoch 33, batch 1250, loss[loss=0.261, ctc_loss=0.1834, cr_loss=0.3876, over 11637.00 frames. ], tot_loss[loss=0.1975, ctc_loss=0.1284, cr_loss=0.3458, over 3349195.75 frames. ], batch size: 123, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 21:01:18,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=11.95 vs. limit=15.0
2024-09-24 21:01:47,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=587687.3333333334, ans=0.125
2024-09-24 21:01:52,859 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=11.35 vs. limit=15.0
2024-09-24 21:02:00,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=587734.0, ans=0.09899494936611666
2024-09-24 21:02:02,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=587734.0, ans=0.2
2024-09-24 21:02:08,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten.whitening_limit, batch_count=587780.6666666666, ans=15.0
2024-09-24 21:02:08,895 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.78 vs. limit=15.0
2024-09-24 21:02:24,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=587827.3333333334, ans=0.2
2024-09-24 21:02:38,675 INFO [train.py:1198] (1/4) Epoch 33, batch 1300, loss[loss=0.2478, ctc_loss=0.1633, cr_loss=0.4225, over 17136.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1283, cr_loss=0.3455, over 3350208.38 frames. ], batch size: 48, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 21:02:43,488 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.271e+02 1.359e+02 1.460e+02 2.870e+02, threshold=2.719e+02, percent-clipped=1.0
2024-09-24 21:02:56,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=587920.6666666666, ans=0.0
2024-09-24 21:03:15,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten.whitening_limit, batch_count=587967.3333333334, ans=15.0
2024-09-24 21:03:37,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=588014.0, ans=0.125
2024-09-24 21:04:00,913 INFO [train.py:1198] (1/4) Epoch 33, batch 1350, loss[loss=0.1932, ctc_loss=0.1272, cr_loss=0.3299, over 17143.00 frames. ], tot_loss[loss=0.1976, ctc_loss=0.1284, cr_loss=0.3459, over 3346785.54 frames. ], batch size: 45, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 21:04:21,146 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.38 vs. limit=12.0
2024-09-24 21:04:27,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=588154.0, ans=0.125
2024-09-24 21:05:15,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=588294.0, ans=0.5
2024-09-24 21:05:20,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=588340.6666666666, ans=0.125
2024-09-24 21:05:21,445 INFO [train.py:1198] (1/4) Epoch 33, batch 1400, loss[loss=0.2022, ctc_loss=0.1318, cr_loss=0.352, over 16977.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1277, cr_loss=0.344, over 3348323.62 frames. ], batch size: 56, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 21:05:26,392 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.268e+02 1.325e+02 1.452e+02 1.895e+02, threshold=2.651e+02, percent-clipped=0.0
2024-09-24 21:05:26,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=588340.6666666666, ans=0.07
2024-09-24 21:05:33,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=588340.6666666666, ans=0.1
2024-09-24 21:05:56,730 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.14 vs. limit=15.0
2024-09-24 21:05:59,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=588434.0, ans=0.0
2024-09-24 21:06:04,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=588434.0, ans=0.0
2024-09-24 21:06:44,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=588527.3333333334, ans=0.1
2024-09-24 21:06:49,513 INFO [train.py:1198] (1/4) Epoch 33, batch 1450, loss[loss=0.2094, ctc_loss=0.1379, cr_loss=0.3575, over 17157.00 frames. ], tot_loss[loss=0.1975, ctc_loss=0.1284, cr_loss=0.3454, over 3343079.38 frames. ], batch size: 48, lr: 3.62e-03, grad_scale: 32.0
2024-09-24 21:06:56,707 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.31 vs. limit=15.0
2024-09-24 21:07:28,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=588667.3333333334, ans=0.0
2024-09-24 21:07:29,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=588667.3333333334, ans=0.1
2024-09-24 21:07:34,652 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 21:07:41,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=588714.0, ans=0.125
2024-09-24 21:07:42,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=588714.0, ans=0.0
2024-09-24 21:07:46,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten.whitening_limit, batch_count=588714.0, ans=22.5
2024-09-24 21:07:53,864 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=588760.6666666666, ans=0.125
2024-09-24 21:08:12,481 INFO [train.py:1198] (1/4) Epoch 33, batch 1500, loss[loss=0.164, ctc_loss=0.1055, cr_loss=0.2928, over 16756.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1276, cr_loss=0.3443, over 3353632.53 frames. ], batch size: 37, lr: 3.61e-03, grad_scale: 32.0
2024-09-24 21:08:17,327 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.288e+02 1.351e+02 1.447e+02 2.720e+02, threshold=2.702e+02, percent-clipped=1.0
2024-09-24 21:08:19,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=588807.3333333334, ans=0.125
2024-09-24 21:08:49,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=588900.6666666666, ans=0.0
2024-09-24 21:09:07,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=588947.3333333334, ans=0.125
2024-09-24 21:09:25,453 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.81 vs. limit=15.0
2024-09-24 21:09:29,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=588994.0, ans=0.2
2024-09-24 21:09:32,673 INFO [train.py:1198] (1/4) Epoch 33, batch 1550, loss[loss=0.2166, ctc_loss=0.1424, cr_loss=0.3713, over 16487.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1283, cr_loss=0.3455, over 3355097.71 frames. ], batch size: 66, lr: 3.61e-03, grad_scale: 32.0
2024-09-24 21:09:39,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=589040.6666666666, ans=0.1
2024-09-24 21:09:45,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=589040.6666666666, ans=0.0
2024-09-24 21:09:50,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=589087.3333333334, ans=0.025
2024-09-24 21:10:01,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=589087.3333333334, ans=0.0
2024-09-24 21:10:03,474 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=589134.0, ans=0.125
2024-09-24 21:10:24,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=589180.6666666666, ans=0.125
2024-09-24 21:10:30,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=589180.6666666666, ans=0.025
2024-09-24 21:10:52,931 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.90 vs. limit=15.0
2024-09-24 21:10:55,334 INFO [train.py:1198] (1/4) Epoch 33, batch 1600, loss[loss=0.2015, ctc_loss=0.1314, cr_loss=0.3504, over 16852.00 frames. ], tot_loss[loss=0.1977, ctc_loss=0.1284, cr_loss=0.3465, over 3362264.11 frames. ], batch size: 58, lr: 3.61e-03, grad_scale: 32.0
2024-09-24 21:11:00,334 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.243e+02 1.338e+02 1.455e+02 2.020e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-24 21:11:10,687 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.70 vs. limit=15.0
2024-09-24 21:11:14,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=589320.6666666666, ans=0.125
2024-09-24 21:11:32,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=589367.3333333334, ans=0.0
2024-09-24 21:12:02,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=589460.6666666666, ans=0.1
2024-09-24 21:12:11,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=589460.6666666666, ans=0.2
2024-09-24 21:12:14,239 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=2.571e-03
2024-09-24 21:12:20,303 INFO [train.py:1198] (1/4) Epoch 33, batch 1650, loss[loss=0.2145, ctc_loss=0.1387, cr_loss=0.379, over 17042.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.1289, cr_loss=0.3475, over 3367430.39 frames. ], batch size: 56, lr: 3.61e-03, grad_scale: 16.0
2024-09-24 21:12:25,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=589507.3333333334, ans=0.125
2024-09-24 21:13:41,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=589740.6666666666, ans=0.0
2024-09-24 21:13:43,006 INFO [train.py:1198] (1/4) Epoch 33, batch 1700, loss[loss=0.2094, ctc_loss=0.1349, cr_loss=0.3724, over 17105.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1285, cr_loss=0.3462, over 3364858.94 frames. ], batch size: 49, lr: 3.61e-03, grad_scale: 16.0
2024-09-24 21:13:43,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten.whitening_limit, batch_count=589740.6666666666, ans=22.5
2024-09-24 21:13:44,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=589740.6666666666, ans=0.125
2024-09-24 21:13:49,388 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.272e+02 1.348e+02 1.489e+02 2.581e+02, threshold=2.697e+02, percent-clipped=0.0
2024-09-24 21:15:01,144 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=14.34 vs. limit=15.0
2024-09-24 21:15:02,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=589974.0, ans=0.125
2024-09-24 21:15:03,862 INFO [train.py:1198] (1/4) Epoch 33, batch 1750, loss[loss=0.2106, ctc_loss=0.1385, cr_loss=0.3605, over 17005.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1285, cr_loss=0.3464, over 3359097.96 frames. ], batch size: 53, lr: 3.61e-03, grad_scale: 16.0
2024-09-24 21:15:44,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=590067.3333333334, ans=0.125
2024-09-24 21:16:05,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=590114.0, ans=0.0
2024-09-24 21:16:20,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=590160.6666666666, ans=0.125
2024-09-24 21:16:31,141 INFO [train.py:1198] (1/4) Epoch 33, batch 1800, loss[loss=0.1928, ctc_loss=0.1226, cr_loss=0.351, over 17154.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.1279, cr_loss=0.3454, over 3359952.31 frames. ], batch size: 48, lr: 3.61e-03, grad_scale: 16.0
2024-09-24 21:16:37,338 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.268e+02 1.341e+02 1.421e+02 2.295e+02, threshold=2.681e+02, percent-clipped=0.0
2024-09-24 21:16:40,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=590207.3333333334, ans=0.125
2024-09-24 21:16:57,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=590254.0, ans=0.0
2024-09-24 21:17:44,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=590394.0, ans=0.05
2024-09-24 21:17:50,250 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.93 vs. limit=6.0
2024-09-24 21:17:50,740 INFO [train.py:1198] (1/4) Epoch 33, batch 1850, loss[loss=0.1968, ctc_loss=0.1281, cr_loss=0.3438, over 17069.00 frames. ], tot_loss[loss=0.1973, ctc_loss=0.1282, cr_loss=0.3457, over 3356436.29 frames. ], batch size: 46, lr: 3.61e-03, grad_scale: 16.0
2024-09-24 21:18:00,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=590440.6666666666, ans=0.125
2024-09-24 21:18:25,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=590534.0, ans=10.0
2024-09-24 21:19:13,089 INFO [train.py:1198] (1/4) Epoch 33, batch 1900, loss[loss=0.2108, ctc_loss=0.1384, cr_loss=0.362, over 17209.00 frames. ], tot_loss[loss=0.1973, ctc_loss=0.1282, cr_loss=0.3454, over 3352670.19 frames. ], batch size: 55, lr: 3.61e-03, grad_scale: 16.0
2024-09-24 21:19:18,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=5.64 vs. limit=15.0
2024-09-24 21:19:19,404 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.258e+02 1.359e+02 1.471e+02 2.658e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-24 21:19:32,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.min_positive, batch_count=590720.6666666666, ans=0.025
2024-09-24 21:19:33,051 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.72 vs. limit=10.0
2024-09-24 21:19:33,166 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.46 vs. limit=22.5
2024-09-24 21:19:37,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=590720.6666666666, ans=0.125
2024-09-24 21:20:14,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=590814.0, ans=0.0
2024-09-24 21:20:33,375 INFO [train.py:1198] (1/4) Epoch 33, batch 1950, loss[loss=0.1967, ctc_loss=0.1261, cr_loss=0.3533, over 17293.00 frames. ], tot_loss[loss=0.1986, ctc_loss=0.1292, cr_loss=0.3471, over 3352622.72 frames. ], batch size: 49, lr: 3.61e-03, grad_scale: 8.0
2024-09-24 21:21:00,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=590954.0, ans=0.0
2024-09-24 21:21:18,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=591000.6666666666, ans=0.1
2024-09-24 21:21:40,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=591047.3333333334, ans=0.1
2024-09-24 21:21:50,354 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.90 vs. limit=15.0
2024-09-24 21:21:56,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=591094.0, ans=0.125
2024-09-24 21:21:57,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=591094.0, ans=0.0
2024-09-24 21:22:00,618 INFO [train.py:1198] (1/4) Epoch 33, batch 2000, loss[loss=0.1936, ctc_loss=0.1227, cr_loss=0.3545, over 17300.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1286, cr_loss=0.3459, over 3346882.49 frames. ], batch size: 51, lr: 3.61e-03, grad_scale: 16.0
2024-09-24 21:22:08,625 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.300e+02 1.359e+02 1.463e+02 1.672e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-24 21:22:28,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=591187.3333333334, ans=0.0
2024-09-24 21:22:35,240 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.35 vs. limit=15.0
2024-09-24 21:22:35,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=591234.0, ans=0.125
2024-09-24 21:22:47,389 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 21:23:10,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=591327.3333333334, ans=0.0
2024-09-24 21:23:22,761 INFO [train.py:1198] (1/4) Epoch 33, batch 2050, loss[loss=0.2101, ctc_loss=0.1354, cr_loss=0.3737, over 17139.00 frames. ], tot_loss[loss=0.1976, ctc_loss=0.1285, cr_loss=0.3455, over 3336153.70 frames. ], batch size: 48, lr: 3.61e-03, grad_scale: 8.0
2024-09-24 21:23:39,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=591420.6666666666, ans=0.2
2024-09-24 21:23:47,846 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.60 vs. limit=22.5
2024-09-24 21:24:17,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=591514.0, ans=0.1
2024-09-24 21:24:27,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=591560.6666666666, ans=0.0
2024-09-24 21:24:38,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=591560.6666666666, ans=0.125
2024-09-24 21:24:42,770 INFO [train.py:1198] (1/4) Epoch 33, batch 2100, loss[loss=0.2282, ctc_loss=0.151, cr_loss=0.386, over 16465.00 frames. ], tot_loss[loss=0.1979, ctc_loss=0.1288, cr_loss=0.3459, over 3328304.63 frames. ], batch size: 66, lr: 3.61e-03, grad_scale: 8.0
2024-09-24 21:24:48,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.04 vs. limit=6.0
2024-09-24 21:24:52,136 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.31 vs. limit=5.0
2024-09-24 21:24:52,446 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.286e+02 1.370e+02 1.480e+02 2.165e+02, threshold=2.740e+02, percent-clipped=0.0
2024-09-24 21:25:04,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=591654.0, ans=0.125
2024-09-24 21:25:07,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_abs, batch_count=591654.0, ans=0.5
2024-09-24 21:25:08,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.69 vs. limit=15.0
2024-09-24 21:25:28,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=591700.6666666666, ans=0.0
2024-09-24 21:25:31,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=591747.3333333334, ans=0.0
2024-09-24 21:25:42,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=591747.3333333334, ans=0.2
2024-09-24 21:26:05,919 INFO [train.py:1198] (1/4) Epoch 33, batch 2150, loss[loss=0.2575, ctc_loss=0.1726, cr_loss=0.4245, over 14972.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1281, cr_loss=0.3456, over 3339294.90 frames. ], batch size: 89, lr: 3.61e-03, grad_scale: 8.0
2024-09-24 21:26:16,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=591840.6666666666, ans=0.125
2024-09-24 21:26:43,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=591934.0, ans=0.125
2024-09-24 21:26:48,924 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.44 vs. limit=22.5
2024-09-24 21:26:50,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=591934.0, ans=0.125
2024-09-24 21:27:02,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_positive, batch_count=591980.6666666666, ans=0.05
2024-09-24 21:27:17,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=592027.3333333334, ans=0.125
2024-09-24 21:27:24,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=592027.3333333334, ans=0.0
2024-09-24 21:27:32,259 INFO [train.py:1198] (1/4) Epoch 33, batch 2200, loss[loss=0.1931, ctc_loss=0.1223, cr_loss=0.354, over 17169.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1283, cr_loss=0.3453, over 3328834.24 frames. ], batch size: 45, lr: 3.60e-03, grad_scale: 8.0
2024-09-24 21:27:38,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=592074.0, ans=0.125
2024-09-24 21:27:41,951 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.250e+02 1.335e+02 1.414e+02 1.894e+02, threshold=2.669e+02, percent-clipped=0.0
2024-09-24 21:28:01,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.32 vs. limit=8.0
2024-09-24 21:28:16,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=592167.3333333334, ans=0.0
2024-09-24 21:28:54,690 INFO [train.py:1198] (1/4) Epoch 33, batch 2250, loss[loss=0.2073, ctc_loss=0.1332, cr_loss=0.3702, over 17287.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1278, cr_loss=0.3444, over 3334936.17 frames. ], batch size: 49, lr: 3.60e-03, grad_scale: 8.0
2024-09-24 21:29:14,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=592354.0, ans=0.125
2024-09-24 21:29:27,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=592400.6666666666, ans=0.125
2024-09-24 21:30:14,785 INFO [train.py:1198] (1/4) Epoch 33, batch 2300, loss[loss=0.2207, ctc_loss=0.1463, cr_loss=0.3718, over 16568.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1279, cr_loss=0.344, over 3327028.73 frames. ], batch size: 66, lr: 3.60e-03, grad_scale: 8.0
2024-09-24 21:30:19,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=592540.6666666666, ans=0.0
2024-09-24 21:30:23,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=592540.6666666666, ans=0.125
2024-09-24 21:30:24,326 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.313e+02 1.398e+02 1.523e+02 4.380e+02, threshold=2.797e+02, percent-clipped=1.0
2024-09-24 21:30:30,183 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=11.37 vs. limit=15.0
2024-09-24 21:30:32,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=592587.3333333334, ans=0.1
2024-09-24 21:30:36,959 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 21:30:49,262 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.70 vs. limit=6.0
2024-09-24 21:30:50,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=592634.0, ans=0.125
2024-09-24 21:31:07,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=592680.6666666666, ans=0.0
2024-09-24 21:31:39,107 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.65 vs. limit=22.5
2024-09-24 21:31:43,090 INFO [train.py:1198] (1/4) Epoch 33, batch 2350, loss[loss=0.2248, ctc_loss=0.1524, cr_loss=0.3616, over 11585.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1267, cr_loss=0.3412, over 3335272.59 frames. ], batch size: 123, lr: 3.60e-03, grad_scale: 8.0
2024-09-24 21:32:01,565 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.89 vs. limit=15.0
2024-09-24 21:32:04,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=592820.6666666666, ans=0.1
2024-09-24 21:32:15,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=592867.3333333334, ans=0.1
2024-09-24 21:32:23,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=592867.3333333334, ans=0.0
2024-09-24 21:33:05,026 INFO [train.py:1198] (1/4) Epoch 33, batch 2400, loss[loss=0.2097, ctc_loss=0.136, cr_loss=0.3681, over 17152.00 frames. ], tot_loss[loss=0.1956, ctc_loss=0.1271, cr_loss=0.3423, over 3344356.99 frames. ], batch size: 48, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:33:06,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=593007.3333333334, ans=0.125
2024-09-24 21:33:14,695 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.266e+02 1.359e+02 1.474e+02 2.316e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-24 21:33:30,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=593054.0, ans=0.0
2024-09-24 21:33:56,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=593147.3333333334, ans=0.0
2024-09-24 21:34:18,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=593194.0, ans=0.2
2024-09-24 21:34:24,926 INFO [train.py:1198] (1/4) Epoch 33, batch 2450, loss[loss=0.2262, ctc_loss=0.1483, cr_loss=0.389, over 17017.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1277, cr_loss=0.3435, over 3351409.82 frames. ], batch size: 52, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:35:11,448 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 21:35:35,602 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.66 vs. limit=15.0
2024-09-24 21:35:48,033 INFO [train.py:1198] (1/4) Epoch 33, batch 2500, loss[loss=0.1663, ctc_loss=0.1041, cr_loss=0.3113, over 16953.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1272, cr_loss=0.3432, over 3355477.97 frames. ], batch size: 42, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:36:00,299 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.280e+02 1.356e+02 1.432e+02 1.776e+02, threshold=2.712e+02, percent-clipped=0.0
2024-09-24 21:36:14,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=593520.6666666666, ans=0.025
2024-09-24 21:36:19,613 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.20 vs. limit=15.0
2024-09-24 21:36:40,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=593614.0, ans=0.125
2024-09-24 21:37:13,486 INFO [train.py:1198] (1/4) Epoch 33, batch 2550, loss[loss=0.1776, ctc_loss=0.1135, cr_loss=0.3204, over 17088.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1268, cr_loss=0.3437, over 3362075.22 frames. ], batch size: 43, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:37:19,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.42 vs. limit=8.0
2024-09-24 21:37:23,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=593707.3333333334, ans=0.125
2024-09-24 21:37:26,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=593707.3333333334, ans=0.125
2024-09-24 21:37:58,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=593800.6666666666, ans=0.1
2024-09-24 21:38:35,207 INFO [train.py:1198] (1/4) Epoch 33, batch 2600, loss[loss=0.1928, ctc_loss=0.1275, cr_loss=0.3264, over 17256.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1273, cr_loss=0.3436, over 3350852.76 frames. ], batch size: 44, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:38:37,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=593940.6666666666, ans=0.0
2024-09-24 21:38:44,724 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.263e+02 1.356e+02 1.473e+02 2.021e+02, threshold=2.712e+02, percent-clipped=0.0
2024-09-24 21:39:10,828 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.26 vs. limit=12.0
2024-09-24 21:39:29,912 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.01 vs. limit=15.0
2024-09-24 21:39:54,548 INFO [train.py:1198] (1/4) Epoch 33, batch 2650, loss[loss=0.2302, ctc_loss=0.1539, cr_loss=0.3818, over 16516.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1266, cr_loss=0.342, over 3351498.29 frames. ], batch size: 66, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:39:57,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=594174.0, ans=0.125
2024-09-24 21:40:07,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=594174.0, ans=0.125
2024-09-24 21:40:14,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=594220.6666666666, ans=0.0
2024-09-24 21:40:23,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=594220.6666666666, ans=0.125
2024-09-24 21:40:23,969 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.54 vs. limit=15.0
2024-09-24 21:41:22,173 INFO [train.py:1198] (1/4) Epoch 33, batch 2700, loss[loss=0.2222, ctc_loss=0.1474, cr_loss=0.3737, over 15242.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1273, cr_loss=0.3435, over 3352901.56 frames. ], batch size: 89, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:41:30,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=594407.3333333334, ans=0.1
2024-09-24 21:41:31,683 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.245e+02 1.329e+02 1.418e+02 2.018e+02, threshold=2.657e+02, percent-clipped=0.0
2024-09-24 21:41:44,609 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 21:42:04,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=594500.6666666666, ans=0.0
2024-09-24 21:42:44,878 INFO [train.py:1198] (1/4) Epoch 33, batch 2750, loss[loss=0.1897, ctc_loss=0.1232, cr_loss=0.3325, over 17255.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1273, cr_loss=0.3435, over 3351558.84 frames. ], batch size: 44, lr: 3.60e-03, grad_scale: 16.0
2024-09-24 21:43:02,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=594687.3333333334, ans=0.2
2024-09-24 21:43:36,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=594780.6666666666, ans=0.125
2024-09-24 21:44:04,858 INFO [train.py:1198] (1/4) Epoch 33, batch 2800, loss[loss=0.1942, ctc_loss=0.1262, cr_loss=0.3399, over 17019.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.1263, cr_loss=0.3421, over 3357034.65 frames. ], batch size: 51, lr: 3.60e-03, grad_scale: 32.0
2024-09-24 21:44:05,709 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.10 vs. limit=15.0
2024-09-24 21:44:13,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=594874.0, ans=0.125
2024-09-24 21:44:14,277 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.285e+02 1.356e+02 1.483e+02 1.883e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-24 21:44:20,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_abs, batch_count=594920.6666666666, ans=0.5
2024-09-24 21:44:34,605 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.16 vs. limit=15.0
2024-09-24 21:44:40,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=594967.3333333334, ans=0.125
2024-09-24 21:45:10,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=595060.6666666666, ans=0.1
2024-09-24 21:45:24,680 INFO [train.py:1198] (1/4) Epoch 33, batch 2850, loss[loss=0.2092, ctc_loss=0.139, cr_loss=0.351, over 17226.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1267, cr_loss=0.3431, over 3363288.81 frames. ], batch size: 55, lr: 3.60e-03, grad_scale: 32.0
2024-09-24 21:45:25,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.80 vs. limit=15.0
2024-09-24 21:45:32,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=595107.3333333334, ans=0.07
2024-09-24 21:45:40,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=595107.3333333334, ans=0.0
2024-09-24 21:45:49,264 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 21:46:16,276 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.89 vs. limit=6.0
2024-09-24 21:46:17,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=595200.6666666666, ans=0.1
2024-09-24 21:46:22,671 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.90 vs. limit=12.0
2024-09-24 21:46:36,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=595294.0, ans=0.0
2024-09-24 21:46:52,041 INFO [train.py:1198] (1/4) Epoch 33, batch 2900, loss[loss=0.2501, ctc_loss=0.1672, cr_loss=0.4144, over 16656.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1273, cr_loss=0.3444, over 3364648.33 frames. ], batch size: 66, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:46:58,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=595340.6666666666, ans=0.2
2024-09-24 21:47:00,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=595340.6666666666, ans=0.0
2024-09-24 21:47:01,426 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=8.85 vs. limit=15.0
2024-09-24 21:47:01,747 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.278e+02 1.378e+02 1.507e+02 2.269e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-24 21:47:07,807 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=10.22 vs. limit=10.0
2024-09-24 21:47:10,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=595387.3333333334, ans=0.2
2024-09-24 21:47:13,754 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.82 vs. limit=15.0
2024-09-24 21:47:32,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=595434.0, ans=0.125
2024-09-24 21:47:44,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=595480.6666666666, ans=0.0
2024-09-24 21:48:07,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=595527.3333333334, ans=0.1
2024-09-24 21:48:10,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=595527.3333333334, ans=0.0
2024-09-24 21:48:15,323 INFO [train.py:1198] (1/4) Epoch 33, batch 2950, loss[loss=0.2456, ctc_loss=0.1684, cr_loss=0.386, over 11819.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1271, cr_loss=0.3445, over 3362767.56 frames. ], batch size: 123, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:48:20,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=595574.0, ans=0.1
2024-09-24 21:48:26,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=595574.0, ans=0.125
2024-09-24 21:48:41,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=595620.6666666666, ans=0.125
2024-09-24 21:48:59,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=595667.3333333334, ans=0.0
2024-09-24 21:49:08,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=595714.0, ans=0.2
2024-09-24 21:49:34,546 INFO [train.py:1198] (1/4) Epoch 33, batch 3000, loss[loss=0.1506, ctc_loss=0.09561, cr_loss=0.2751, over 16957.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1275, cr_loss=0.3445, over 3350387.57 frames. ], batch size: 42, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:49:34,547 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 21:49:50,278 INFO [train.py:1230] (1/4) Epoch 33, validation: loss=0.03597, ctc_loss=0.03597, cr_loss=9.382e-15, over 944034.00 frames. 
2024-09-24 21:49:50,278 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 21:49:59,653 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.292e+02 1.353e+02 1.495e+02 2.152e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-24 21:50:25,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=9.65 vs. limit=10.0
2024-09-24 21:50:29,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=595900.6666666666, ans=0.125
2024-09-24 21:50:58,522 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.57 vs. limit=15.0
2024-09-24 21:51:08,635 INFO [train.py:1198] (1/4) Epoch 33, batch 3050, loss[loss=0.1908, ctc_loss=0.1233, cr_loss=0.3378, over 17304.00 frames. ], tot_loss[loss=0.1963, ctc_loss=0.1274, cr_loss=0.3446, over 3349322.21 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:51:08,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=596040.6666666666, ans=0.0
2024-09-24 21:51:18,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=596040.6666666666, ans=0.0
2024-09-24 21:51:26,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=596087.3333333334, ans=0.025
2024-09-24 21:51:32,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=596087.3333333334, ans=0.125
2024-09-24 21:51:53,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=596134.0, ans=0.125
2024-09-24 21:52:01,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=596180.6666666666, ans=0.2
2024-09-24 21:52:06,692 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.56 vs. limit=15.0
2024-09-24 21:52:18,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=596227.3333333334, ans=10.0
2024-09-24 21:52:27,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=596227.3333333334, ans=0.0
2024-09-24 21:52:28,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=596227.3333333334, ans=0.0
2024-09-24 21:52:34,217 INFO [train.py:1198] (1/4) Epoch 33, batch 3100, loss[loss=0.207, ctc_loss=0.1337, cr_loss=0.3669, over 17301.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1278, cr_loss=0.3452, over 3355822.71 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:52:36,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=596274.0, ans=0.0
2024-09-24 21:52:43,658 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.278e+02 1.346e+02 1.474e+02 1.974e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-24 21:53:53,433 INFO [train.py:1198] (1/4) Epoch 33, batch 3150, loss[loss=0.2395, ctc_loss=0.1592, cr_loss=0.4012, over 17006.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1278, cr_loss=0.3456, over 3359064.67 frames. ], batch size: 53, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:54:03,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=596507.3333333334, ans=0.025
2024-09-24 21:54:11,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=596554.0, ans=0.05
2024-09-24 21:54:13,737 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.70 vs. limit=6.0
2024-09-24 21:54:19,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=596554.0, ans=0.125
2024-09-24 21:54:23,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=596600.6666666666, ans=0.125
2024-09-24 21:54:45,835 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=5.59 vs. limit=15.0
2024-09-24 21:54:54,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=596694.0, ans=0.125
2024-09-24 21:55:08,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=596694.0, ans=0.125
2024-09-24 21:55:11,608 INFO [train.py:1198] (1/4) Epoch 33, batch 3200, loss[loss=0.1806, ctc_loss=0.1157, cr_loss=0.3244, over 17264.00 frames. ], tot_loss[loss=0.1966, ctc_loss=0.1277, cr_loss=0.3449, over 3364203.25 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:55:13,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=596740.6666666666, ans=0.5
2024-09-24 21:55:22,508 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.266e+02 1.354e+02 1.465e+02 1.945e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-24 21:55:27,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=596787.3333333334, ans=0.125
2024-09-24 21:55:38,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=596787.3333333334, ans=0.125
2024-09-24 21:56:14,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=596927.3333333334, ans=0.0
2024-09-24 21:56:17,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=596927.3333333334, ans=0.1
2024-09-24 21:56:17,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=596927.3333333334, ans=0.125
2024-09-24 21:56:31,858 INFO [train.py:1198] (1/4) Epoch 33, batch 3250, loss[loss=0.227, ctc_loss=0.1501, cr_loss=0.3844, over 16604.00 frames. ], tot_loss[loss=0.1963, ctc_loss=0.1274, cr_loss=0.3447, over 3370897.11 frames. ], batch size: 66, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:56:33,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=596974.0, ans=0.0
2024-09-24 21:56:33,935 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 21:56:40,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=596974.0, ans=0.125
2024-09-24 21:57:30,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=597114.0, ans=0.125
2024-09-24 21:57:50,337 INFO [train.py:1198] (1/4) Epoch 33, batch 3300, loss[loss=0.1739, ctc_loss=0.1092, cr_loss=0.3238, over 17053.00 frames. ], tot_loss[loss=0.1966, ctc_loss=0.1276, cr_loss=0.3454, over 3373137.53 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:57:57,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=597207.3333333334, ans=0.125
2024-09-24 21:58:01,413 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.317e+02 1.415e+02 1.516e+02 3.233e+02, threshold=2.830e+02, percent-clipped=1.0
2024-09-24 21:58:04,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=597254.0, ans=0.1
2024-09-24 21:58:23,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=597300.6666666666, ans=0.05
2024-09-24 21:59:11,412 INFO [train.py:1198] (1/4) Epoch 33, batch 3350, loss[loss=0.2013, ctc_loss=0.1314, cr_loss=0.3497, over 17035.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3445, over 3369992.96 frames. ], batch size: 52, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 21:59:13,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=597440.6666666666, ans=0.0
2024-09-24 21:59:22,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=597440.6666666666, ans=0.0
2024-09-24 21:59:36,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=597487.3333333334, ans=0.125
2024-09-24 22:00:29,915 INFO [train.py:1198] (1/4) Epoch 33, batch 3400, loss[loss=0.2142, ctc_loss=0.1424, cr_loss=0.359, over 14903.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1269, cr_loss=0.3438, over 3365219.93 frames. ], batch size: 89, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 22:00:31,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=597674.0, ans=0.025
2024-09-24 22:00:37,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=597674.0, ans=0.0
2024-09-24 22:00:40,768 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.249e+02 1.331e+02 1.438e+02 2.060e+02, threshold=2.663e+02, percent-clipped=0.0
2024-09-24 22:01:05,014 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.86 vs. limit=15.0
2024-09-24 22:01:12,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=597767.3333333334, ans=0.0
2024-09-24 22:01:38,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=597860.6666666666, ans=0.2
2024-09-24 22:01:47,156 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.47 vs. limit=15.0
2024-09-24 22:01:47,178 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.81 vs. limit=10.0
2024-09-24 22:01:48,346 INFO [train.py:1198] (1/4) Epoch 33, batch 3450, loss[loss=0.2374, ctc_loss=0.1571, cr_loss=0.4017, over 16474.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1272, cr_loss=0.3445, over 3363397.48 frames. ], batch size: 66, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 22:01:48,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=597907.3333333334, ans=0.1
2024-09-24 22:01:52,874 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.07 vs. limit=22.5
2024-09-24 22:01:53,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=597907.3333333334, ans=0.125
2024-09-24 22:01:56,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=597907.3333333334, ans=0.0
2024-09-24 22:02:12,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=597954.0, ans=0.125
2024-09-24 22:02:13,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=597954.0, ans=0.2
2024-09-24 22:02:24,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=598000.6666666666, ans=0.125
2024-09-24 22:02:28,106 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.83 vs. limit=15.0
2024-09-24 22:02:40,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=598047.3333333334, ans=0.1
2024-09-24 22:02:46,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=598047.3333333334, ans=0.2
2024-09-24 22:02:59,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.84 vs. limit=6.0
2024-09-24 22:03:04,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=598094.0, ans=0.2
2024-09-24 22:03:12,223 INFO [train.py:1198] (1/4) Epoch 33, batch 3500, loss[loss=0.1911, ctc_loss=0.122, cr_loss=0.3459, over 17297.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1278, cr_loss=0.3454, over 3367224.06 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 22:03:15,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=598140.6666666666, ans=0.125
2024-09-24 22:03:23,015 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.259e+02 1.352e+02 1.466e+02 4.097e+02, threshold=2.703e+02, percent-clipped=1.0
2024-09-24 22:03:23,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=598140.6666666666, ans=0.04949747468305833
2024-09-24 22:03:24,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=598140.6666666666, ans=0.125
2024-09-24 22:04:04,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=598280.6666666666, ans=0.125
2024-09-24 22:04:13,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=598327.3333333334, ans=0.025
2024-09-24 22:04:24,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=598327.3333333334, ans=0.125
2024-09-24 22:04:27,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=598327.3333333334, ans=0.0
2024-09-24 22:04:30,420 INFO [train.py:1198] (1/4) Epoch 33, batch 3550, loss[loss=0.157, ctc_loss=0.1007, cr_loss=0.2814, over 17204.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.1279, cr_loss=0.3454, over 3365741.02 frames. ], batch size: 41, lr: 3.59e-03, grad_scale: 32.0
2024-09-24 22:04:35,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=598374.0, ans=0.0
2024-09-24 22:04:38,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=598374.0, ans=0.125
2024-09-24 22:04:41,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=598374.0, ans=0.125
2024-09-24 22:04:58,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=598420.6666666666, ans=0.1
2024-09-24 22:05:44,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=598560.6666666666, ans=0.035
2024-09-24 22:05:48,756 INFO [train.py:1198] (1/4) Epoch 33, batch 3600, loss[loss=0.2231, ctc_loss=0.1468, cr_loss=0.3812, over 16568.00 frames. ], tot_loss[loss=0.1979, ctc_loss=0.1286, cr_loss=0.3466, over 3358557.12 frames. ], batch size: 66, lr: 3.58e-03, grad_scale: 32.0
2024-09-24 22:05:59,708 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.266e+02 1.339e+02 1.451e+02 1.959e+02, threshold=2.678e+02, percent-clipped=0.0
2024-09-24 22:06:01,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=598607.3333333334, ans=0.0
2024-09-24 22:06:30,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=598700.6666666666, ans=0.0
2024-09-24 22:06:42,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=598747.3333333334, ans=0.125
2024-09-24 22:07:08,976 INFO [train.py:1198] (1/4) Epoch 33, batch 3650, loss[loss=0.1955, ctc_loss=0.1253, cr_loss=0.351, over 17311.00 frames. ], tot_loss[loss=0.1976, ctc_loss=0.1283, cr_loss=0.3464, over 3358670.06 frames. ], batch size: 46, lr: 3.58e-03, grad_scale: 32.0
2024-09-24 22:07:10,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=598840.6666666666, ans=0.1
2024-09-24 22:07:11,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.89 vs. limit=22.5
2024-09-24 22:07:20,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=598840.6666666666, ans=0.0
2024-09-24 22:07:34,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=598887.3333333334, ans=0.1
2024-09-24 22:07:45,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=598934.0, ans=0.125
2024-09-24 22:07:47,509 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.96 vs. limit=15.0
2024-09-24 22:08:06,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=598980.6666666666, ans=0.0
2024-09-24 22:08:11,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=599027.3333333334, ans=0.1
2024-09-24 22:08:27,969 INFO [train.py:1198] (1/4) Epoch 33, batch 3700, loss[loss=0.1862, ctc_loss=0.1204, cr_loss=0.3292, over 17033.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1277, cr_loss=0.3453, over 3359888.07 frames. ], batch size: 52, lr: 3.58e-03, grad_scale: 32.0
2024-09-24 22:08:39,001 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.282e+02 1.323e+02 1.479e+02 2.496e+02, threshold=2.646e+02, percent-clipped=0.0
2024-09-24 22:08:39,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=599074.0, ans=0.125
2024-09-24 22:09:22,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=599214.0, ans=0.1
2024-09-24 22:09:29,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=599260.6666666666, ans=0.125
2024-09-24 22:09:45,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.66 vs. limit=6.0
2024-09-24 22:09:45,897 INFO [train.py:1198] (1/4) Epoch 33, batch 3750, loss[loss=0.2095, ctc_loss=0.1359, cr_loss=0.3679, over 17029.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1279, cr_loss=0.3454, over 3347914.00 frames. ], batch size: 56, lr: 3.58e-03, grad_scale: 32.0
2024-09-24 22:09:51,230 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.99 vs. limit=15.0
2024-09-24 22:10:03,917 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.90 vs. limit=15.0
2024-09-24 22:11:03,989 INFO [train.py:1198] (1/4) Epoch 33, batch 3800, loss[loss=0.2319, ctc_loss=0.1544, cr_loss=0.3878, over 15165.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.129, cr_loss=0.3468, over 3323801.33 frames. ], batch size: 89, lr: 3.58e-03, grad_scale: 32.0
2024-09-24 22:11:08,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=599540.6666666666, ans=0.2
2024-09-24 22:11:14,753 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.301e+02 1.426e+02 1.535e+02 2.880e+02, threshold=2.852e+02, percent-clipped=1.0
2024-09-24 22:11:19,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=599587.3333333334, ans=0.125
2024-09-24 22:11:22,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=599587.3333333334, ans=0.0
2024-09-24 22:11:40,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=599634.0, ans=0.09899494936611666
2024-09-24 22:11:46,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=599634.0, ans=0.125
2024-09-24 22:11:54,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=599680.6666666666, ans=0.0
2024-09-24 22:12:09,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=599727.3333333334, ans=0.125
2024-09-24 22:12:13,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=599727.3333333334, ans=0.0
2024-09-24 22:12:18,990 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.22 vs. limit=15.0
2024-09-24 22:12:22,914 INFO [train.py:1198] (1/4) Epoch 33, batch 3850, loss[loss=0.2386, ctc_loss=0.164, cr_loss=0.3726, over 11753.00 frames. ], tot_loss[loss=0.2006, ctc_loss=0.1307, cr_loss=0.3496, over 3290290.40 frames. ], batch size: 123, lr: 3.58e-03, grad_scale: 16.0
2024-09-24 22:12:29,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=599774.0, ans=0.125
2024-09-24 22:12:32,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=599774.0, ans=0.0
2024-09-24 22:12:40,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=599820.6666666666, ans=0.2
2024-09-24 22:12:41,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=599820.6666666666, ans=0.0
2024-09-24 22:12:45,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=599820.6666666666, ans=0.1
2024-09-24 22:12:46,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=599820.6666666666, ans=0.1
2024-09-24 22:14:24,101 INFO [train.py:1198] (1/4) Epoch 34, batch 0, loss[loss=0.2021, ctc_loss=0.1289, cr_loss=0.3663, over 16874.00 frames. ], tot_loss[loss=0.2021, ctc_loss=0.1289, cr_loss=0.3663, over 16874.00 frames. ], batch size: 58, lr: 3.53e-03, grad_scale: 32.0
2024-09-24 22:14:24,101 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 22:14:39,385 INFO [train.py:1230] (1/4) Epoch 34, validation: loss=0.03567, ctc_loss=0.03567, cr_loss=1.032e-14, over 944034.00 frames. 
2024-09-24 22:14:39,386 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 22:14:43,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=599988.6666666666, ans=0.0
2024-09-24 22:14:44,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=599988.6666666666, ans=0.1
2024-09-24 22:14:58,491 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.377e+02 1.527e+02 1.748e+02 2.707e+02, threshold=3.055e+02, percent-clipped=0.0
2024-09-24 22:15:21,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=600082.0, ans=0.1
2024-09-24 22:15:59,188 INFO [train.py:1198] (1/4) Epoch 34, batch 50, loss[loss=0.1926, ctc_loss=0.1256, cr_loss=0.335, over 17057.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1278, cr_loss=0.3456, over 756465.68 frames. ], batch size: 46, lr: 3.53e-03, grad_scale: 32.0
2024-09-24 22:16:02,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=600222.0, ans=0.0
2024-09-24 22:16:06,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.64 vs. limit=15.0
2024-09-24 22:16:16,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=600268.6666666666, ans=0.125
2024-09-24 22:16:20,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=600268.6666666666, ans=0.2
2024-09-24 22:17:00,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=600362.0, ans=0.125
2024-09-24 22:17:06,000 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.73 vs. limit=12.0
2024-09-24 22:17:08,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=600408.6666666666, ans=0.125
2024-09-24 22:17:11,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=600408.6666666666, ans=0.125
2024-09-24 22:17:24,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=600408.6666666666, ans=0.1
2024-09-24 22:17:29,796 INFO [train.py:1198] (1/4) Epoch 34, batch 100, loss[loss=0.1847, ctc_loss=0.1198, cr_loss=0.3242, over 17305.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1275, cr_loss=0.3448, over 1326198.54 frames. ], batch size: 49, lr: 3.53e-03, grad_scale: 32.0
2024-09-24 22:17:49,214 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.286e+02 1.376e+02 1.498e+02 2.035e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-24 22:17:52,764 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 22:18:23,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=600595.3333333334, ans=0.125
2024-09-24 22:18:37,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer_ff3.min_abs, batch_count=600642.0, ans=0.2
2024-09-24 22:18:49,900 INFO [train.py:1198] (1/4) Epoch 34, batch 150, loss[loss=0.1656, ctc_loss=0.1049, cr_loss=0.3034, over 17134.00 frames. ], tot_loss[loss=0.1982, ctc_loss=0.1288, cr_loss=0.3468, over 1765678.73 frames. ], batch size: 40, lr: 3.52e-03, grad_scale: 32.0
2024-09-24 22:18:54,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=600688.6666666666, ans=0.0
2024-09-24 22:19:01,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=600688.6666666666, ans=0.125
2024-09-24 22:19:13,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=600735.3333333334, ans=0.125
2024-09-24 22:19:27,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=600782.0, ans=0.125
2024-09-24 22:19:38,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=600828.6666666666, ans=0.125
2024-09-24 22:19:50,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=600828.6666666666, ans=0.125
2024-09-24 22:20:07,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=600922.0, ans=0.125
2024-09-24 22:20:09,193 INFO [train.py:1198] (1/4) Epoch 34, batch 200, loss[loss=0.2009, ctc_loss=0.1305, cr_loss=0.3522, over 17300.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1281, cr_loss=0.3465, over 2128629.84 frames. ], batch size: 49, lr: 3.52e-03, grad_scale: 32.0
2024-09-24 22:20:14,219 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=600922.0, ans=0.125
2024-09-24 22:20:17,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=600922.0, ans=0.125
2024-09-24 22:20:17,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=600922.0, ans=0.125
2024-09-24 22:20:30,037 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.259e+02 1.333e+02 1.425e+02 2.058e+02, threshold=2.665e+02, percent-clipped=0.0
2024-09-24 22:20:41,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=601015.3333333334, ans=0.125
2024-09-24 22:20:59,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=601062.0, ans=0.125
2024-09-24 22:21:00,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=601062.0, ans=0.5
2024-09-24 22:21:07,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=601062.0, ans=0.025
2024-09-24 22:21:08,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.50 vs. limit=12.0
2024-09-24 22:21:10,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=601062.0, ans=0.0
2024-09-24 22:21:18,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=601108.6666666666, ans=0.125
2024-09-24 22:21:32,286 INFO [train.py:1198] (1/4) Epoch 34, batch 250, loss[loss=0.2207, ctc_loss=0.1432, cr_loss=0.3873, over 16945.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1269, cr_loss=0.3443, over 2404664.95 frames. ], batch size: 58, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:22:18,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=601248.6666666666, ans=0.125
2024-09-24 22:22:28,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=601295.3333333334, ans=0.125
2024-09-24 22:22:29,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=601295.3333333334, ans=0.125
2024-09-24 22:23:00,622 INFO [train.py:1198] (1/4) Epoch 34, batch 300, loss[loss=0.1968, ctc_loss=0.1284, cr_loss=0.3421, over 17247.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3447, over 2623471.61 frames. ], batch size: 44, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:23:13,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=601388.6666666666, ans=0.0
2024-09-24 22:23:18,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=601435.3333333334, ans=0.1
2024-09-24 22:23:21,391 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.266e+02 1.387e+02 1.525e+02 2.483e+02, threshold=2.773e+02, percent-clipped=0.0
2024-09-24 22:23:34,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=601482.0, ans=0.125
2024-09-24 22:23:38,655 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.70 vs. limit=12.0
2024-09-24 22:23:53,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=601528.6666666666, ans=0.0
2024-09-24 22:23:56,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=601528.6666666666, ans=0.125
2024-09-24 22:24:17,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=601575.3333333334, ans=0.125
2024-09-24 22:24:20,383 INFO [train.py:1198] (1/4) Epoch 34, batch 350, loss[loss=0.2049, ctc_loss=0.1344, cr_loss=0.3522, over 17111.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.3438, over 2782954.11 frames. ], batch size: 43, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:24:20,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=601622.0, ans=0.2
2024-09-24 22:24:38,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=601668.6666666666, ans=0.0
2024-09-24 22:24:54,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=601715.3333333334, ans=0.125
2024-09-24 22:25:39,904 INFO [train.py:1198] (1/4) Epoch 34, batch 400, loss[loss=0.2177, ctc_loss=0.1438, cr_loss=0.3695, over 17009.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1269, cr_loss=0.344, over 2912064.31 frames. ], batch size: 51, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:25:45,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=601855.3333333334, ans=0.1
2024-09-24 22:25:56,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=601902.0, ans=0.5
2024-09-24 22:26:02,214 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.275e+02 1.410e+02 1.501e+02 2.362e+02, threshold=2.821e+02, percent-clipped=0.0
2024-09-24 22:26:12,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=601948.6666666666, ans=0.125
2024-09-24 22:26:29,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=601995.3333333334, ans=0.1
2024-09-24 22:27:03,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.88 vs. limit=6.0
2024-09-24 22:27:05,426 INFO [train.py:1198] (1/4) Epoch 34, batch 450, loss[loss=0.1768, ctc_loss=0.1122, cr_loss=0.323, over 17013.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1271, cr_loss=0.3445, over 3011527.99 frames. ], batch size: 44, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:27:33,869 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=5.89 vs. limit=15.0
2024-09-24 22:27:44,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=602182.0, ans=0.125
2024-09-24 22:27:58,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=602228.6666666666, ans=0.2
2024-09-24 22:28:01,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=602228.6666666666, ans=0.125
2024-09-24 22:28:02,563 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.11 vs. limit=15.0
2024-09-24 22:28:15,237 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.99 vs. limit=15.0
2024-09-24 22:28:30,683 INFO [train.py:1198] (1/4) Epoch 34, batch 500, loss[loss=0.2043, ctc_loss=0.1325, cr_loss=0.3589, over 17309.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1271, cr_loss=0.3448, over 3090401.19 frames. ], batch size: 49, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:28:33,180 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.74 vs. limit=22.5
2024-09-24 22:28:52,903 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.302e+02 1.372e+02 1.474e+02 2.066e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-24 22:29:13,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=602415.3333333334, ans=0.125
2024-09-24 22:29:28,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=602462.0, ans=0.0
2024-09-24 22:29:32,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=602508.6666666666, ans=0.125
2024-09-24 22:29:33,578 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.91 vs. limit=12.0
2024-09-24 22:29:35,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=602508.6666666666, ans=0.05
2024-09-24 22:29:40,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=602508.6666666666, ans=0.125
2024-09-24 22:29:49,895 INFO [train.py:1198] (1/4) Epoch 34, batch 550, loss[loss=0.2177, ctc_loss=0.1445, cr_loss=0.3661, over 17220.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1277, cr_loss=0.3456, over 3146109.16 frames. ], batch size: 50, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:30:10,509 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.93 vs. limit=15.0
2024-09-24 22:30:24,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=602648.6666666666, ans=0.1
2024-09-24 22:30:51,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=602695.3333333334, ans=0.125
2024-09-24 22:31:08,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=602788.6666666666, ans=0.0
2024-09-24 22:31:09,783 INFO [train.py:1198] (1/4) Epoch 34, batch 600, loss[loss=0.2035, ctc_loss=0.1296, cr_loss=0.3699, over 17136.00 frames. ], tot_loss[loss=0.198, ctc_loss=0.1285, cr_loss=0.3473, over 3188487.67 frames. ], batch size: 48, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:31:11,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=602788.6666666666, ans=0.125
2024-09-24 22:31:11,967 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.23 vs. limit=6.0
2024-09-24 22:31:22,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=602788.6666666666, ans=0.025
2024-09-24 22:31:34,416 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.259e+02 1.339e+02 1.429e+02 3.195e+02, threshold=2.679e+02, percent-clipped=1.0
2024-09-24 22:31:39,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=602835.3333333334, ans=0.2
2024-09-24 22:32:39,991 INFO [train.py:1198] (1/4) Epoch 34, batch 650, loss[loss=0.2068, ctc_loss=0.134, cr_loss=0.3637, over 16897.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1284, cr_loss=0.347, over 3228198.94 frames. ], batch size: 58, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:33:03,283 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.36 vs. limit=15.0
2024-09-24 22:33:25,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=603115.3333333334, ans=0.1
2024-09-24 22:33:36,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=603162.0, ans=0.0
2024-09-24 22:33:44,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=603208.6666666666, ans=0.125
2024-09-24 22:34:00,563 INFO [train.py:1198] (1/4) Epoch 34, batch 700, loss[loss=0.1907, ctc_loss=0.1216, cr_loss=0.3455, over 17240.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1279, cr_loss=0.3462, over 3262088.24 frames. ], batch size: 44, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:34:23,177 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.279e+02 1.394e+02 1.540e+02 2.036e+02, threshold=2.788e+02, percent-clipped=0.0
2024-09-24 22:35:21,219 INFO [train.py:1198] (1/4) Epoch 34, batch 750, loss[loss=0.2357, ctc_loss=0.1594, cr_loss=0.3815, over 15943.00 frames. ], tot_loss[loss=0.1977, ctc_loss=0.1284, cr_loss=0.3466, over 3285640.20 frames. ], batch size: 74, lr: 3.52e-03, grad_scale: 16.0
2024-09-24 22:35:37,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=603535.3333333334, ans=0.2
2024-09-24 22:36:12,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=603628.6666666666, ans=0.2
2024-09-24 22:36:18,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=603628.6666666666, ans=0.125
2024-09-24 22:36:24,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=603628.6666666666, ans=0.125
2024-09-24 22:36:36,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=603675.3333333334, ans=0.0
2024-09-24 22:36:43,571 INFO [train.py:1198] (1/4) Epoch 34, batch 800, loss[loss=0.1868, ctc_loss=0.1236, cr_loss=0.3162, over 17311.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1277, cr_loss=0.3459, over 3308501.54 frames. ], batch size: 49, lr: 3.52e-03, grad_scale: 32.0
2024-09-24 22:36:56,930 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=9.58 vs. limit=15.0
2024-09-24 22:37:08,640 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.269e+02 1.369e+02 1.482e+02 1.915e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-24 22:37:18,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=603768.6666666666, ans=0.0
2024-09-24 22:37:33,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=603815.3333333334, ans=0.0
2024-09-24 22:37:44,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=603862.0, ans=0.1
2024-09-24 22:38:01,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=603908.6666666666, ans=0.2
2024-09-24 22:38:04,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=603908.6666666666, ans=0.0
2024-09-24 22:38:12,078 INFO [train.py:1198] (1/4) Epoch 34, batch 850, loss[loss=0.2131, ctc_loss=0.1417, cr_loss=0.3566, over 17314.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1271, cr_loss=0.345, over 3320232.77 frames. ], batch size: 49, lr: 3.52e-03, grad_scale: 32.0
2024-09-24 22:38:15,638 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 22:38:21,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=603955.3333333334, ans=0.125
2024-09-24 22:38:21,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=603955.3333333334, ans=0.5
2024-09-24 22:38:47,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=604048.6666666666, ans=0.2
2024-09-24 22:38:52,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.30 vs. limit=15.0
2024-09-24 22:39:09,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=604095.3333333334, ans=0.2
2024-09-24 22:39:14,291 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=604142.0, ans=0.1
2024-09-24 22:39:27,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=604142.0, ans=0.0
2024-09-24 22:39:31,803 INFO [train.py:1198] (1/4) Epoch 34, batch 900, loss[loss=0.2061, ctc_loss=0.1316, cr_loss=0.3726, over 17288.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1278, cr_loss=0.3467, over 3341858.45 frames. ], batch size: 51, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:39:54,076 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.293e+02 1.407e+02 1.509e+02 3.747e+02, threshold=2.814e+02, percent-clipped=1.0
2024-09-24 22:40:10,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=604282.0, ans=0.125
2024-09-24 22:40:11,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.29 vs. limit=15.0
2024-09-24 22:40:23,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=604328.6666666666, ans=0.0
2024-09-24 22:40:26,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=604328.6666666666, ans=0.2
2024-09-24 22:40:30,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=604328.6666666666, ans=0.125
2024-09-24 22:40:38,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=604375.3333333334, ans=0.0
2024-09-24 22:40:51,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=604422.0, ans=0.125
2024-09-24 22:40:52,440 INFO [train.py:1198] (1/4) Epoch 34, batch 950, loss[loss=0.1753, ctc_loss=0.1148, cr_loss=0.3024, over 17121.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1283, cr_loss=0.3473, over 3345306.29 frames. ], batch size: 43, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:41:07,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=604468.6666666666, ans=0.0
2024-09-24 22:42:23,021 INFO [train.py:1198] (1/4) Epoch 34, batch 1000, loss[loss=0.2095, ctc_loss=0.1398, cr_loss=0.3488, over 17007.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1275, cr_loss=0.3461, over 3357128.82 frames. ], batch size: 44, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:42:32,049 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.52 vs. limit=22.5
2024-09-24 22:42:45,412 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.270e+02 1.349e+02 1.472e+02 1.904e+02, threshold=2.698e+02, percent-clipped=0.0
2024-09-24 22:43:14,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=604795.3333333334, ans=0.0
2024-09-24 22:43:21,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=604795.3333333334, ans=0.2
2024-09-24 22:43:26,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=604842.0, ans=0.125
2024-09-24 22:43:30,251 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.73 vs. limit=22.5
2024-09-24 22:43:40,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=604842.0, ans=0.1
2024-09-24 22:43:43,609 INFO [train.py:1198] (1/4) Epoch 34, batch 1050, loss[loss=0.202, ctc_loss=0.1318, cr_loss=0.3513, over 16997.00 frames. ], tot_loss[loss=0.1971, ctc_loss=0.1279, cr_loss=0.3462, over 3354106.44 frames. ], batch size: 53, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:43:45,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=604888.6666666666, ans=0.125
2024-09-24 22:43:48,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=604888.6666666666, ans=0.025
2024-09-24 22:44:15,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=604982.0, ans=0.5
2024-09-24 22:44:25,810 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.56 vs. limit=22.5
2024-09-24 22:44:27,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten.whitening_limit, batch_count=604982.0, ans=22.5
2024-09-24 22:44:42,110 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=7.56 vs. limit=15.0
2024-09-24 22:44:44,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=605028.6666666666, ans=0.125
2024-09-24 22:45:03,652 INFO [train.py:1198] (1/4) Epoch 34, batch 1100, loss[loss=0.2178, ctc_loss=0.1456, cr_loss=0.3608, over 16093.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1276, cr_loss=0.3462, over 3359442.12 frames. ], batch size: 74, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:45:26,104 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.266e+02 1.354e+02 1.438e+02 1.919e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-24 22:45:44,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=605215.3333333334, ans=0.0
2024-09-24 22:46:01,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=605262.0, ans=0.125
2024-09-24 22:46:23,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=605355.3333333334, ans=0.2
2024-09-24 22:46:26,898 INFO [train.py:1198] (1/4) Epoch 34, batch 1150, loss[loss=0.2076, ctc_loss=0.1358, cr_loss=0.3595, over 17225.00 frames. ], tot_loss[loss=0.1971, ctc_loss=0.1278, cr_loss=0.3462, over 3356250.75 frames. ], batch size: 50, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:46:33,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=605355.3333333334, ans=0.125
2024-09-24 22:46:53,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.82 vs. limit=6.0
2024-09-24 22:46:53,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=605402.0, ans=0.125
2024-09-24 22:47:07,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=605448.6666666666, ans=0.125
2024-09-24 22:47:15,553 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=10.69 vs. limit=22.5
2024-09-24 22:47:47,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.29 vs. limit=15.0
2024-09-24 22:47:48,587 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.70 vs. limit=15.0
2024-09-24 22:47:54,034 INFO [train.py:1198] (1/4) Epoch 34, batch 1200, loss[loss=0.1959, ctc_loss=0.1244, cr_loss=0.3574, over 17286.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1277, cr_loss=0.346, over 3354250.59 frames. ], batch size: 46, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:47:59,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=605588.6666666666, ans=0.125
2024-09-24 22:48:07,980 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.28 vs. limit=22.5
2024-09-24 22:48:16,699 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.285e+02 1.360e+02 1.428e+02 2.074e+02, threshold=2.720e+02, percent-clipped=0.0
2024-09-24 22:48:37,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=605682.0, ans=0.125
2024-09-24 22:48:55,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=605728.6666666666, ans=0.125
2024-09-24 22:49:08,845 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=5.70 vs. limit=12.0
2024-09-24 22:49:14,480 INFO [train.py:1198] (1/4) Epoch 34, batch 1250, loss[loss=0.1887, ctc_loss=0.1236, cr_loss=0.3256, over 17058.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1271, cr_loss=0.3447, over 3355182.85 frames. ], batch size: 46, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:49:14,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=605822.0, ans=0.125
2024-09-24 22:49:14,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=605822.0, ans=0.2
2024-09-24 22:49:38,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=605868.6666666666, ans=0.125
2024-09-24 22:49:40,975 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.95 vs. limit=15.0
2024-09-24 22:49:48,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.74 vs. limit=15.0
2024-09-24 22:50:07,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=605962.0, ans=0.125
2024-09-24 22:50:08,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=605962.0, ans=0.0
2024-09-24 22:50:31,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=606008.6666666666, ans=0.0
2024-09-24 22:50:34,624 INFO [train.py:1198] (1/4) Epoch 34, batch 1300, loss[loss=0.1919, ctc_loss=0.1235, cr_loss=0.3417, over 17090.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1273, cr_loss=0.3459, over 3363065.92 frames. ], batch size: 49, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:50:38,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=606055.3333333334, ans=0.125
2024-09-24 22:50:57,098 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.261e+02 1.315e+02 1.424e+02 2.011e+02, threshold=2.630e+02, percent-clipped=0.0
2024-09-24 22:50:59,730 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.63 vs. limit=6.0
2024-09-24 22:51:06,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=606148.6666666666, ans=0.0
2024-09-24 22:51:19,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=606148.6666666666, ans=0.125
2024-09-24 22:51:59,708 INFO [train.py:1198] (1/4) Epoch 34, batch 1350, loss[loss=0.2088, ctc_loss=0.1369, cr_loss=0.3594, over 17291.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.127, cr_loss=0.3451, over 3357177.42 frames. ], batch size: 49, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:52:03,582 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.54 vs. limit=15.0
2024-09-24 22:52:12,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=606288.6666666666, ans=0.125
2024-09-24 22:52:24,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=606335.3333333334, ans=0.125
2024-09-24 22:52:46,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=606382.0, ans=0.025
2024-09-24 22:52:51,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=606428.6666666666, ans=0.0
2024-09-24 22:52:54,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=606428.6666666666, ans=0.1
2024-09-24 22:53:24,834 INFO [train.py:1198] (1/4) Epoch 34, batch 1400, loss[loss=0.2146, ctc_loss=0.1396, cr_loss=0.3751, over 17292.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1277, cr_loss=0.3457, over 3356552.39 frames. ], batch size: 49, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:53:32,218 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=5.59 vs. limit=15.0
2024-09-24 22:53:36,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=606522.0, ans=0.125
2024-09-24 22:53:45,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=606568.6666666666, ans=0.0
2024-09-24 22:53:47,246 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.283e+02 1.396e+02 1.529e+02 1.918e+02, threshold=2.793e+02, percent-clipped=0.0
2024-09-24 22:54:21,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=606662.0, ans=0.0
2024-09-24 22:54:28,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_abs, batch_count=606708.6666666666, ans=0.5
2024-09-24 22:54:35,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=606708.6666666666, ans=0.125
2024-09-24 22:54:44,663 INFO [train.py:1198] (1/4) Epoch 34, batch 1450, loss[loss=0.2212, ctc_loss=0.1441, cr_loss=0.3856, over 17234.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.1281, cr_loss=0.3466, over 3351707.49 frames. ], batch size: 50, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:55:09,021 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 22:55:56,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer_ff3.min_abs, batch_count=606942.0, ans=0.2
2024-09-24 22:56:04,598 INFO [train.py:1198] (1/4) Epoch 34, batch 1500, loss[loss=0.1778, ctc_loss=0.1142, cr_loss=0.3179, over 17112.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1277, cr_loss=0.3459, over 3349785.44 frames. ], batch size: 40, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:56:29,448 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.258e+02 1.350e+02 1.442e+02 1.821e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-24 22:56:29,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=607035.3333333334, ans=0.0
2024-09-24 22:56:33,068 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 22:56:56,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=607128.6666666666, ans=0.125
2024-09-24 22:57:04,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=607128.6666666666, ans=0.125
2024-09-24 22:57:04,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=607128.6666666666, ans=0.0
2024-09-24 22:57:20,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=607175.3333333334, ans=0.07
2024-09-24 22:57:24,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=607175.3333333334, ans=0.125
2024-09-24 22:57:30,999 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=5.29 vs. limit=15.0
2024-09-24 22:57:34,685 INFO [train.py:1198] (1/4) Epoch 34, batch 1550, loss[loss=0.1927, ctc_loss=0.124, cr_loss=0.3434, over 17022.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1266, cr_loss=0.3437, over 3350592.26 frames. ], batch size: 51, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:57:42,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=607222.0, ans=0.0
2024-09-24 22:58:00,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=607268.6666666666, ans=0.125
2024-09-24 22:58:11,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=607315.3333333334, ans=0.125
2024-09-24 22:58:26,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=607362.0, ans=0.125
2024-09-24 22:58:27,956 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.39 vs. limit=12.0
2024-09-24 22:58:54,425 INFO [train.py:1198] (1/4) Epoch 34, batch 1600, loss[loss=0.1923, ctc_loss=0.1209, cr_loss=0.3572, over 17121.00 frames. ], tot_loss[loss=0.1952, ctc_loss=0.1265, cr_loss=0.3435, over 3354386.24 frames. ], batch size: 40, lr: 3.51e-03, grad_scale: 32.0
2024-09-24 22:59:03,530 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.93 vs. limit=10.0
2024-09-24 22:59:05,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=607455.3333333334, ans=0.125
2024-09-24 22:59:12,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=607502.0, ans=0.125
2024-09-24 22:59:16,752 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.263e+02 1.329e+02 1.418e+02 2.097e+02, threshold=2.657e+02, percent-clipped=0.0
2024-09-24 22:59:25,981 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.48 vs. limit=6.0
2024-09-24 22:59:27,141 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.32 vs. limit=15.0
2024-09-24 23:00:14,909 INFO [train.py:1198] (1/4) Epoch 34, batch 1650, loss[loss=0.1893, ctc_loss=0.1211, cr_loss=0.3411, over 17065.00 frames. ], tot_loss[loss=0.1946, ctc_loss=0.1261, cr_loss=0.3426, over 3360960.10 frames. ], batch size: 46, lr: 3.50e-03, grad_scale: 16.0
2024-09-24 23:00:29,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=607735.3333333334, ans=0.125
2024-09-24 23:00:31,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=607735.3333333334, ans=0.0
2024-09-24 23:00:31,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.34 vs. limit=15.0
2024-09-24 23:00:49,325 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.86 vs. limit=10.0
2024-09-24 23:01:16,192 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.58 vs. limit=15.0
2024-09-24 23:01:24,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=607875.3333333334, ans=0.125
2024-09-24 23:01:29,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=607875.3333333334, ans=0.0
2024-09-24 23:01:36,804 INFO [train.py:1198] (1/4) Epoch 34, batch 1700, loss[loss=0.223, ctc_loss=0.1494, cr_loss=0.3677, over 16731.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.3439, over 3346878.26 frames. ], batch size: 61, lr: 3.50e-03, grad_scale: 16.0
2024-09-24 23:01:44,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=607922.0, ans=0.125
2024-09-24 23:02:05,581 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.278e+02 1.343e+02 1.428e+02 2.095e+02, threshold=2.685e+02, percent-clipped=0.0
2024-09-24 23:02:11,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=607968.6666666666, ans=0.125
2024-09-24 23:02:23,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=608015.3333333334, ans=0.0
2024-09-24 23:02:26,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=608015.3333333334, ans=0.0
2024-09-24 23:02:32,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=608062.0, ans=0.0
2024-09-24 23:02:37,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=608062.0, ans=0.0
2024-09-24 23:02:51,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=608108.6666666666, ans=0.1
2024-09-24 23:03:04,326 INFO [train.py:1198] (1/4) Epoch 34, batch 1750, loss[loss=0.2014, ctc_loss=0.1317, cr_loss=0.3488, over 16132.00 frames. ], tot_loss[loss=0.1948, ctc_loss=0.1262, cr_loss=0.3428, over 3357238.61 frames. ], batch size: 74, lr: 3.50e-03, grad_scale: 16.0
2024-09-24 23:03:11,349 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.24 vs. limit=12.0
2024-09-24 23:03:55,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=608295.3333333334, ans=0.0
2024-09-24 23:04:13,980 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.68 vs. limit=15.0
2024-09-24 23:04:24,156 INFO [train.py:1198] (1/4) Epoch 34, batch 1800, loss[loss=0.2136, ctc_loss=0.1398, cr_loss=0.3693, over 16975.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1269, cr_loss=0.3446, over 3351617.23 frames. ], batch size: 53, lr: 3.50e-03, grad_scale: 16.0
2024-09-24 23:04:48,241 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.244e+02 1.332e+02 1.441e+02 2.602e+02, threshold=2.663e+02, percent-clipped=0.0
2024-09-24 23:05:33,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=608575.3333333334, ans=0.0
2024-09-24 23:05:44,350 INFO [train.py:1198] (1/4) Epoch 34, batch 1850, loss[loss=0.1676, ctc_loss=0.1079, cr_loss=0.2986, over 17014.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1265, cr_loss=0.3436, over 3355713.53 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 16.0
2024-09-24 23:05:52,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=608622.0, ans=0.2
2024-09-24 23:06:44,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=608762.0, ans=0.125
2024-09-24 23:06:54,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=608808.6666666666, ans=0.125
2024-09-24 23:06:59,971 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.89 vs. limit=15.0
2024-09-24 23:07:15,203 INFO [train.py:1198] (1/4) Epoch 34, batch 1900, loss[loss=0.2307, ctc_loss=0.1508, cr_loss=0.3997, over 15077.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1269, cr_loss=0.3449, over 3357735.38 frames. ], batch size: 89, lr: 3.50e-03, grad_scale: 16.0
2024-09-24 23:07:26,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=608855.3333333334, ans=0.125
2024-09-24 23:07:38,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2.whitening_limit, batch_count=608902.0, ans=15.0
2024-09-24 23:07:39,345 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.279e+02 1.348e+02 1.451e+02 1.794e+02, threshold=2.695e+02, percent-clipped=0.0
2024-09-24 23:07:41,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=608902.0, ans=0.125
2024-09-24 23:07:57,740 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.13 vs. limit=15.0
2024-09-24 23:08:10,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=608995.3333333334, ans=0.1
2024-09-24 23:08:11,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=608995.3333333334, ans=0.0
2024-09-24 23:08:18,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.83 vs. limit=10.0
2024-09-24 23:08:24,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer_na.min_abs, batch_count=609042.0, ans=0.02
2024-09-24 23:08:35,254 INFO [train.py:1198] (1/4) Epoch 34, batch 1950, loss[loss=0.2172, ctc_loss=0.1434, cr_loss=0.369, over 17005.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.3443, over 3355404.87 frames. ], batch size: 53, lr: 3.50e-03, grad_scale: 16.0
2024-09-24 23:08:50,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=609135.3333333334, ans=0.0
2024-09-24 23:09:14,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=609182.0, ans=0.1
2024-09-24 23:09:20,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer_ff2.min_abs, batch_count=609182.0, ans=0.1
2024-09-24 23:09:40,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=609275.3333333334, ans=0.2
2024-09-24 23:09:51,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=609275.3333333334, ans=0.0
2024-09-24 23:09:56,093 INFO [train.py:1198] (1/4) Epoch 34, batch 2000, loss[loss=0.1989, ctc_loss=0.1284, cr_loss=0.3521, over 17004.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.127, cr_loss=0.3442, over 3346791.85 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 32.0
2024-09-24 23:10:07,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=609322.0, ans=0.07
2024-09-24 23:10:19,956 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.293e+02 1.367e+02 1.523e+02 2.152e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-24 23:10:25,582 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.64 vs. limit=12.0
2024-09-24 23:10:49,775 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:10:50,041 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.20 vs. limit=15.0
2024-09-24 23:11:10,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=609508.6666666666, ans=0.0
2024-09-24 23:11:18,784 INFO [train.py:1198] (1/4) Epoch 34, batch 2050, loss[loss=0.1855, ctc_loss=0.1214, cr_loss=0.3204, over 16083.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1269, cr_loss=0.3444, over 3342781.50 frames. ], batch size: 74, lr: 3.50e-03, grad_scale: 32.0
2024-09-24 23:11:45,857 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.85 vs. limit=15.0
2024-09-24 23:12:45,744 INFO [train.py:1198] (1/4) Epoch 34, batch 2100, loss[loss=0.1718, ctc_loss=0.1097, cr_loss=0.3105, over 17256.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1278, cr_loss=0.3457, over 3352454.30 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 32.0
2024-09-24 23:13:10,054 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.285e+02 1.366e+02 1.478e+02 2.142e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-24 23:14:01,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=609975.3333333334, ans=0.1
2024-09-24 23:14:06,284 INFO [train.py:1198] (1/4) Epoch 34, batch 2150, loss[loss=0.174, ctc_loss=0.1146, cr_loss=0.2971, over 17238.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1277, cr_loss=0.3453, over 3354857.42 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 32.0
2024-09-24 23:14:14,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=610022.0, ans=0.09899494936611666
2024-09-24 23:14:24,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=610068.6666666666, ans=0.1
2024-09-24 23:14:36,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=610115.3333333334, ans=0.125
2024-09-24 23:15:12,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=610208.6666666666, ans=0.07
2024-09-24 23:15:12,692 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=21.66 vs. limit=22.5
2024-09-24 23:15:26,283 INFO [train.py:1198] (1/4) Epoch 34, batch 2200, loss[loss=0.2323, ctc_loss=0.1542, cr_loss=0.3905, over 16409.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1277, cr_loss=0.3457, over 3362289.05 frames. ], batch size: 66, lr: 3.50e-03, grad_scale: 32.0
2024-09-24 23:15:36,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=610255.3333333334, ans=0.09899494936611666
2024-09-24 23:15:42,806 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:15:50,348 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.250e+02 1.358e+02 1.486e+02 2.433e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-24 23:16:13,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=610395.3333333334, ans=0.125
2024-09-24 23:16:31,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=610442.0, ans=0.125
2024-09-24 23:16:43,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=610442.0, ans=10.0
2024-09-24 23:16:43,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=610442.0, ans=0.1
2024-09-24 23:16:50,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.38 vs. limit=12.0
2024-09-24 23:16:51,472 INFO [train.py:1198] (1/4) Epoch 34, batch 2250, loss[loss=0.2018, ctc_loss=0.1309, cr_loss=0.3541, over 17304.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1274, cr_loss=0.3452, over 3363145.31 frames. ], batch size: 46, lr: 3.50e-03, grad_scale: 32.0
2024-09-24 23:17:00,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=610488.6666666666, ans=0.1
2024-09-24 23:17:10,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=610535.3333333334, ans=0.2
2024-09-24 23:17:45,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=610628.6666666666, ans=0.125
2024-09-24 23:17:47,804 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.91 vs. limit=15.0
2024-09-24 23:18:14,097 INFO [train.py:1198] (1/4) Epoch 34, batch 2300, loss[loss=0.2186, ctc_loss=0.1473, cr_loss=0.3563, over 16431.00 frames. ], tot_loss[loss=0.1967, ctc_loss=0.1276, cr_loss=0.3456, over 3370820.84 frames. ], batch size: 66, lr: 3.50e-03, grad_scale: 32.0
2024-09-24 23:18:20,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=610722.0, ans=0.0
2024-09-24 23:18:38,051 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.297e+02 1.390e+02 1.515e+02 2.091e+02, threshold=2.780e+02, percent-clipped=0.0
2024-09-24 23:19:07,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=610862.0, ans=0.025
2024-09-24 23:19:12,126 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=18.65 vs. limit=22.5
2024-09-24 23:19:12,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=610862.0, ans=15.0
2024-09-24 23:19:22,367 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=12.05 vs. limit=15.0
2024-09-24 23:19:34,056 INFO [train.py:1198] (1/4) Epoch 34, batch 2350, loss[loss=0.2189, ctc_loss=0.1433, cr_loss=0.3781, over 17020.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1268, cr_loss=0.3435, over 3372010.85 frames. ], batch size: 51, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:19:46,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=610955.3333333334, ans=0.1
2024-09-24 23:20:02,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=611002.0, ans=0.125
2024-09-24 23:20:44,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=611142.0, ans=0.125
2024-09-24 23:20:53,277 INFO [train.py:1198] (1/4) Epoch 34, batch 2400, loss[loss=0.2116, ctc_loss=0.1434, cr_loss=0.3408, over 11452.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1268, cr_loss=0.3435, over 3358187.51 frames. ], batch size: 123, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:21:19,778 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.247e+02 1.313e+02 1.426e+02 1.860e+02, threshold=2.625e+02, percent-clipped=0.0
2024-09-24 23:21:37,517 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.26 vs. limit=15.0
2024-09-24 23:21:40,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=611282.0, ans=0.1
2024-09-24 23:21:43,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=611282.0, ans=0.1
2024-09-24 23:21:56,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=611328.6666666666, ans=0.0
2024-09-24 23:22:04,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=611328.6666666666, ans=0.0
2024-09-24 23:22:08,219 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:22:23,849 INFO [train.py:1198] (1/4) Epoch 34, batch 2450, loss[loss=0.2342, ctc_loss=0.1608, cr_loss=0.3668, over 11791.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1268, cr_loss=0.3435, over 3352709.42 frames. ], batch size: 124, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:22:25,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=611422.0, ans=0.125
2024-09-24 23:22:26,474 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.28 vs. limit=22.5
2024-09-24 23:22:35,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=611422.0, ans=0.125
2024-09-24 23:22:35,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=611422.0, ans=0.0
2024-09-24 23:22:39,283 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.19 vs. limit=15.0
2024-09-24 23:22:41,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=611468.6666666666, ans=0.125
2024-09-24 23:22:59,621 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=16.41 vs. limit=22.5
2024-09-24 23:23:03,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=611515.3333333334, ans=0.0
2024-09-24 23:23:07,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=611515.3333333334, ans=0.0
2024-09-24 23:23:12,365 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:23:36,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=611608.6666666666, ans=0.125
2024-09-24 23:23:43,933 INFO [train.py:1198] (1/4) Epoch 34, batch 2500, loss[loss=0.157, ctc_loss=0.1004, cr_loss=0.2827, over 17267.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1272, cr_loss=0.3436, over 3352258.93 frames. ], batch size: 44, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:23:45,933 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:24:08,084 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.282e+02 1.373e+02 1.484e+02 1.977e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-24 23:24:18,430 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.35 vs. limit=15.0
2024-09-24 23:24:27,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=611748.6666666666, ans=0.0
2024-09-24 23:24:29,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.min_positive, batch_count=611748.6666666666, ans=0.05
2024-09-24 23:24:40,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=611795.3333333334, ans=0.125
2024-09-24 23:24:42,427 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.41 vs. limit=6.0
2024-09-24 23:24:48,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=611842.0, ans=0.09899494936611666
2024-09-24 23:25:04,064 INFO [train.py:1198] (1/4) Epoch 34, batch 2550, loss[loss=0.2021, ctc_loss=0.1332, cr_loss=0.3448, over 16982.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1274, cr_loss=0.3443, over 3356832.64 frames. ], batch size: 53, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:25:05,789 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:25:10,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=611888.6666666666, ans=0.0
2024-09-24 23:25:56,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=612028.6666666666, ans=0.0
2024-09-24 23:26:04,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=612028.6666666666, ans=0.125
2024-09-24 23:26:23,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=612075.3333333334, ans=0.125
2024-09-24 23:26:29,169 INFO [train.py:1198] (1/4) Epoch 34, batch 2600, loss[loss=0.1819, ctc_loss=0.1173, cr_loss=0.3229, over 16868.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1272, cr_loss=0.3439, over 3361803.54 frames. ], batch size: 58, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:26:42,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=612122.0, ans=0.0
2024-09-24 23:26:55,519 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.298e+02 1.376e+02 1.484e+02 2.452e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-24 23:27:00,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=612168.6666666666, ans=0.125
2024-09-24 23:27:02,455 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:27:10,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=612215.3333333334, ans=0.09899494936611666
2024-09-24 23:27:11,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer_ff2.min_abs, batch_count=612215.3333333334, ans=0.1
2024-09-24 23:27:16,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=612215.3333333334, ans=0.125
2024-09-24 23:27:37,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=612308.6666666666, ans=0.0
2024-09-24 23:27:38,310 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.81 vs. limit=10.0
2024-09-24 23:27:51,788 INFO [train.py:1198] (1/4) Epoch 34, batch 2650, loss[loss=0.1867, ctc_loss=0.1188, cr_loss=0.3397, over 17196.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1271, cr_loss=0.3439, over 3365610.40 frames. ], batch size: 41, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:28:09,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=612402.0, ans=0.0
2024-09-24 23:28:43,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=612495.3333333334, ans=0.125
2024-09-24 23:28:50,468 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.73 vs. limit=15.0
2024-09-24 23:28:52,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_abs, batch_count=612495.3333333334, ans=0.5
2024-09-24 23:28:53,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.32 vs. limit=15.0
2024-09-24 23:29:12,127 INFO [train.py:1198] (1/4) Epoch 34, batch 2700, loss[loss=0.2063, ctc_loss=0.1342, cr_loss=0.3605, over 16480.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1271, cr_loss=0.3438, over 3363097.51 frames. ], batch size: 66, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:29:17,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=612588.6666666666, ans=0.1
2024-09-24 23:29:24,034 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.82 vs. limit=22.5
2024-09-24 23:29:36,055 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.298e+02 1.372e+02 1.486e+02 2.496e+02, threshold=2.744e+02, percent-clipped=0.0
2024-09-24 23:30:07,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.46 vs. limit=22.5
2024-09-24 23:30:13,353 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.78 vs. limit=10.0
2024-09-24 23:30:30,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=612822.0, ans=0.025
2024-09-24 23:30:31,910 INFO [train.py:1198] (1/4) Epoch 34, batch 2750, loss[loss=0.2542, ctc_loss=0.169, cr_loss=0.4263, over 15019.00 frames. ], tot_loss[loss=0.1976, ctc_loss=0.1283, cr_loss=0.3466, over 3361434.96 frames. ], batch size: 89, lr: 3.49e-03, grad_scale: 16.0
2024-09-24 23:30:35,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=612822.0, ans=0.125
2024-09-24 23:30:35,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=612822.0, ans=0.125
2024-09-24 23:30:38,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=612822.0, ans=0.0
2024-09-24 23:31:08,982 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.71 vs. limit=15.0
2024-09-24 23:31:14,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=612915.3333333334, ans=0.125
2024-09-24 23:31:35,623 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.23 vs. limit=15.0
2024-09-24 23:32:02,048 INFO [train.py:1198] (1/4) Epoch 34, batch 2800, loss[loss=0.1925, ctc_loss=0.123, cr_loss=0.3477, over 16632.00 frames. ], tot_loss[loss=0.1975, ctc_loss=0.1281, cr_loss=0.3467, over 3367736.35 frames. ], batch size: 61, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:32:27,770 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.265e+02 1.343e+02 1.457e+02 2.911e+02, threshold=2.687e+02, percent-clipped=1.0
2024-09-24 23:32:42,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=613148.6666666666, ans=0.0
2024-09-24 23:32:47,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=613148.6666666666, ans=0.125
2024-09-24 23:32:48,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=613195.3333333334, ans=0.2
2024-09-24 23:32:50,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=613195.3333333334, ans=0.1
2024-09-24 23:33:15,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.64 vs. limit=15.0
2024-09-24 23:33:22,279 INFO [train.py:1198] (1/4) Epoch 34, batch 2850, loss[loss=0.21, ctc_loss=0.1333, cr_loss=0.3835, over 17005.00 frames. ], tot_loss[loss=0.1974, ctc_loss=0.128, cr_loss=0.3469, over 3365864.63 frames. ], batch size: 44, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:33:29,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=613288.6666666666, ans=0.1
2024-09-24 23:33:30,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=613288.6666666666, ans=0.0
2024-09-24 23:33:35,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=613288.6666666666, ans=0.05
2024-09-24 23:33:51,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=613335.3333333334, ans=0.125
2024-09-24 23:34:02,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=613382.0, ans=0.0
2024-09-24 23:34:18,630 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=613428.6666666666, ans=0.125
2024-09-24 23:34:24,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=613475.3333333334, ans=0.1
2024-09-24 23:34:28,830 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.17 vs. limit=15.0
2024-09-24 23:34:42,222 INFO [train.py:1198] (1/4) Epoch 34, batch 2900, loss[loss=0.2026, ctc_loss=0.1303, cr_loss=0.3615, over 17303.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1276, cr_loss=0.3464, over 3369122.01 frames. ], batch size: 51, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:35:07,635 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.262e+02 1.340e+02 1.438e+02 2.197e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-24 23:35:08,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=613568.6666666666, ans=0.025
2024-09-24 23:35:20,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=613615.3333333334, ans=0.0
2024-09-24 23:35:44,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=613708.6666666666, ans=0.0
2024-09-24 23:35:49,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=613708.6666666666, ans=0.2
2024-09-24 23:36:00,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=613708.6666666666, ans=0.0
2024-09-24 23:36:00,970 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.88 vs. limit=15.0
2024-09-24 23:36:01,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=613708.6666666666, ans=0.025
2024-09-24 23:36:04,802 INFO [train.py:1198] (1/4) Epoch 34, batch 2950, loss[loss=0.1874, ctc_loss=0.1218, cr_loss=0.3279, over 17298.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.1277, cr_loss=0.3465, over 3367545.40 frames. ], batch size: 49, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:36:28,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=613802.0, ans=0.0
2024-09-24 23:36:48,714 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.97 vs. limit=15.0
2024-09-24 23:37:12,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=613895.3333333334, ans=0.125
2024-09-24 23:37:27,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=613942.0, ans=0.09899494936611666
2024-09-24 23:37:32,109 INFO [train.py:1198] (1/4) Epoch 34, batch 3000, loss[loss=0.1663, ctc_loss=0.1053, cr_loss=0.3047, over 17288.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1283, cr_loss=0.3475, over 3361105.10 frames. ], batch size: 42, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:37:32,109 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-24 23:37:46,202 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.4.encoder.layers.2.self_attn_weights, attn_weights_entropy = tensor([3.7262, 2.9668, 3.2304, 3.3840], device='cuda:1')
2024-09-24 23:37:46,297 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.0.layers.0.self_attn_weights, attn_weights_entropy = tensor([5.0990, 4.8479, 4.4240, 4.6285], device='cuda:1')
2024-09-24 23:37:47,937 INFO [train.py:1230] (1/4) Epoch 34, validation: loss=0.03583, ctc_loss=0.03583, cr_loss=9.471e-15, over 944034.00 frames. 
2024-09-24 23:37:47,938 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-24 23:38:05,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=614035.3333333334, ans=0.2
2024-09-24 23:38:12,760 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.290e+02 1.384e+02 1.469e+02 2.229e+02, threshold=2.767e+02, percent-clipped=0.0
2024-09-24 23:38:33,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=614128.6666666666, ans=10.0
2024-09-24 23:39:00,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=614175.3333333334, ans=0.125
2024-09-24 23:39:00,697 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.59 vs. limit=15.0
2024-09-24 23:39:00,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=14.63 vs. limit=15.0
2024-09-24 23:39:05,968 INFO [train.py:1198] (1/4) Epoch 34, batch 3050, loss[loss=0.1666, ctc_loss=0.1053, cr_loss=0.3065, over 16388.00 frames. ], tot_loss[loss=0.197, ctc_loss=0.1277, cr_loss=0.3466, over 3360504.72 frames. ], batch size: 36, lr: 3.49e-03, grad_scale: 32.0
2024-09-24 23:39:22,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten.whitening_limit, batch_count=614268.6666666666, ans=15.0
2024-09-24 23:39:23,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=614268.6666666666, ans=0.2
2024-09-24 23:39:42,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=614315.3333333334, ans=0.125
2024-09-24 23:39:45,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=614315.3333333334, ans=0.0
2024-09-24 23:40:16,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=614408.6666666666, ans=0.125
2024-09-24 23:40:18,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=614408.6666666666, ans=0.125
2024-09-24 23:40:23,998 INFO [train.py:1198] (1/4) Epoch 34, batch 3100, loss[loss=0.2126, ctc_loss=0.1405, cr_loss=0.3604, over 16714.00 frames. ], tot_loss[loss=0.1969, ctc_loss=0.1277, cr_loss=0.3461, over 3361449.82 frames. ], batch size: 61, lr: 3.49e-03, grad_scale: 16.0
2024-09-24 23:40:30,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=614455.3333333334, ans=0.1
2024-09-24 23:40:35,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=614455.3333333334, ans=0.125
2024-09-24 23:40:50,475 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.258e+02 1.350e+02 1.404e+02 1.862e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-24 23:41:09,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=614595.3333333334, ans=10.0
2024-09-24 23:41:24,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=614642.0, ans=0.125
2024-09-24 23:41:34,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=614642.0, ans=0.1
2024-09-24 23:41:34,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=614642.0, ans=0.125
2024-09-24 23:41:41,890 INFO [train.py:1198] (1/4) Epoch 34, batch 3150, loss[loss=0.2375, ctc_loss=0.1555, cr_loss=0.4097, over 17222.00 frames. ], tot_loss[loss=0.1973, ctc_loss=0.1281, cr_loss=0.3464, over 3357632.30 frames. ], batch size: 50, lr: 3.48e-03, grad_scale: 16.0
2024-09-24 23:41:42,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=614688.6666666666, ans=0.1
2024-09-24 23:41:51,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=614688.6666666666, ans=0.125
2024-09-24 23:42:09,667 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.77 vs. limit=15.0
2024-09-24 23:42:19,024 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.76 vs. limit=15.0
2024-09-24 23:42:42,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=614828.6666666666, ans=0.0
2024-09-24 23:43:00,718 INFO [train.py:1198] (1/4) Epoch 34, batch 3200, loss[loss=0.2282, ctc_loss=0.1484, cr_loss=0.3993, over 17262.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1277, cr_loss=0.3458, over 3362750.18 frames. ], batch size: 55, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:43:11,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=614922.0, ans=0.025
2024-09-24 23:43:11,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=614922.0, ans=0.1
2024-09-24 23:43:15,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=614968.6666666666, ans=0.125
2024-09-24 23:43:27,299 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.284e+02 1.377e+02 1.475e+02 3.177e+02, threshold=2.753e+02, percent-clipped=2.0
2024-09-24 23:43:34,336 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.40 vs. limit=22.5
2024-09-24 23:43:47,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=615062.0, ans=0.125
2024-09-24 23:44:00,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=615062.0, ans=0.125
2024-09-24 23:44:19,049 INFO [train.py:1198] (1/4) Epoch 34, batch 3250, loss[loss=0.1828, ctc_loss=0.1189, cr_loss=0.3196, over 17013.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1266, cr_loss=0.3436, over 3371459.67 frames. ], batch size: 44, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:44:35,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=615202.0, ans=0.2
2024-09-24 23:44:40,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.00 vs. limit=15.0
2024-09-24 23:44:49,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=615248.6666666666, ans=10.0
2024-09-24 23:44:52,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=615248.6666666666, ans=0.125
2024-09-24 23:45:05,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=615248.6666666666, ans=0.0
2024-09-24 23:45:18,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=615295.3333333334, ans=0.07
2024-09-24 23:45:39,575 INFO [train.py:1198] (1/4) Epoch 34, batch 3300, loss[loss=0.1615, ctc_loss=0.1016, cr_loss=0.2995, over 17184.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3447, over 3359211.56 frames. ], batch size: 41, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:45:41,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=615388.6666666666, ans=0.0
2024-09-24 23:45:44,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=615388.6666666666, ans=0.125
2024-09-24 23:45:54,259 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.50 vs. limit=10.0
2024-09-24 23:46:10,484 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.276e+02 1.373e+02 1.543e+02 3.468e+02, threshold=2.745e+02, percent-clipped=1.0
2024-09-24 23:46:14,445 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.46 vs. limit=6.0
2024-09-24 23:46:36,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=615528.6666666666, ans=0.025
2024-09-24 23:46:54,715 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.40 vs. limit=22.5
2024-09-24 23:47:04,474 INFO [train.py:1198] (1/4) Epoch 34, batch 3350, loss[loss=0.1948, ctc_loss=0.1241, cr_loss=0.3533, over 17018.00 frames. ], tot_loss[loss=0.1952, ctc_loss=0.1265, cr_loss=0.3436, over 3357111.53 frames. ], batch size: 44, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:47:35,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=615715.3333333334, ans=0.0
2024-09-24 23:48:22,375 INFO [train.py:1198] (1/4) Epoch 34, batch 3400, loss[loss=0.1837, ctc_loss=0.1189, cr_loss=0.3241, over 17021.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1266, cr_loss=0.3436, over 3351446.98 frames. ], batch size: 44, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:48:43,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=8.84 vs. limit=10.0
2024-09-24 23:48:45,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=615902.0, ans=0.0
2024-09-24 23:48:46,565 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.44 vs. limit=15.0
2024-09-24 23:48:48,722 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.295e+02 1.404e+02 1.516e+02 2.292e+02, threshold=2.807e+02, percent-clipped=0.0
2024-09-24 23:48:52,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=615948.6666666666, ans=0.0
2024-09-24 23:49:01,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=615948.6666666666, ans=0.05
2024-09-24 23:49:12,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=615995.3333333334, ans=0.125
2024-09-24 23:49:22,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=615995.3333333334, ans=0.125
2024-09-24 23:49:42,038 INFO [train.py:1198] (1/4) Epoch 34, batch 3450, loss[loss=0.2114, ctc_loss=0.1371, cr_loss=0.3712, over 17031.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.344, over 3354496.68 frames. ], batch size: 52, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:49:44,530 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.63 vs. limit=22.5
2024-09-24 23:50:06,029 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.47 vs. limit=22.5
2024-09-24 23:50:13,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=616182.0, ans=0.2
2024-09-24 23:50:18,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=616182.0, ans=0.025
2024-09-24 23:51:00,312 INFO [train.py:1198] (1/4) Epoch 34, batch 3500, loss[loss=0.1854, ctc_loss=0.1212, cr_loss=0.3209, over 17017.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1271, cr_loss=0.344, over 3351318.16 frames. ], batch size: 44, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:51:26,844 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.270e+02 1.396e+02 1.524e+02 2.184e+02, threshold=2.793e+02, percent-clipped=0.0
2024-09-24 23:51:31,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=616415.3333333334, ans=0.125
2024-09-24 23:51:41,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=616415.3333333334, ans=0.0
2024-09-24 23:51:48,353 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.94 vs. limit=15.0
2024-09-24 23:51:59,307 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=12.52 vs. limit=22.5
2024-09-24 23:52:06,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=616508.6666666666, ans=0.0
2024-09-24 23:52:18,496 INFO [train.py:1198] (1/4) Epoch 34, batch 3550, loss[loss=0.1845, ctc_loss=0.1193, cr_loss=0.3259, over 17303.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1273, cr_loss=0.3445, over 3347415.39 frames. ], batch size: 46, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:52:28,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=616555.3333333334, ans=0.2
2024-09-24 23:52:31,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=616555.3333333334, ans=0.0
2024-09-24 23:52:32,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=616602.0, ans=0.025
2024-09-24 23:52:41,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=616602.0, ans=0.125
2024-09-24 23:52:46,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=616602.0, ans=0.125
2024-09-24 23:52:56,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.92 vs. limit=15.0
2024-09-24 23:53:02,514 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:53:10,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=616695.3333333334, ans=0.125
2024-09-24 23:53:35,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=616788.6666666666, ans=0.0
2024-09-24 23:53:36,433 INFO [train.py:1198] (1/4) Epoch 34, batch 3600, loss[loss=0.1766, ctc_loss=0.1119, cr_loss=0.3231, over 17298.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1271, cr_loss=0.3441, over 3355344.84 frames. ], batch size: 46, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:53:54,572 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.65 vs. limit=15.0
2024-09-24 23:54:03,052 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.287e+02 1.340e+02 1.449e+02 1.947e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-24 23:54:09,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=616882.0, ans=0.2
2024-09-24 23:54:16,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten.whitening_limit, batch_count=616882.0, ans=15.0
2024-09-24 23:54:56,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=617022.0, ans=0.0
2024-09-24 23:54:57,438 INFO [train.py:1198] (1/4) Epoch 34, batch 3650, loss[loss=0.1721, ctc_loss=0.1096, cr_loss=0.3128, over 17044.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1268, cr_loss=0.3434, over 3344599.52 frames. ], batch size: 39, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:55:03,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=617022.0, ans=0.025
2024-09-24 23:55:27,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=617068.6666666666, ans=0.1
2024-09-24 23:55:35,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=617115.3333333334, ans=0.125
2024-09-24 23:55:46,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=617115.3333333334, ans=0.1
2024-09-24 23:55:56,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=617162.0, ans=0.0
2024-09-24 23:56:21,544 INFO [train.py:1198] (1/4) Epoch 34, batch 3700, loss[loss=0.164, ctc_loss=0.1036, cr_loss=0.3023, over 16745.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1267, cr_loss=0.3433, over 3336250.14 frames. ], batch size: 37, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:56:38,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=617302.0, ans=0.125
2024-09-24 23:56:41,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=617302.0, ans=0.125
2024-09-24 23:56:43,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=617302.0, ans=0.125
2024-09-24 23:56:48,014 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.259e+02 1.354e+02 1.435e+02 3.016e+02, threshold=2.708e+02, percent-clipped=2.0
2024-09-24 23:56:56,837 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.80 vs. limit=22.5
2024-09-24 23:57:16,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=617395.3333333334, ans=0.0
2024-09-24 23:57:17,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=617395.3333333334, ans=0.125
2024-09-24 23:57:39,592 INFO [train.py:1198] (1/4) Epoch 34, batch 3750, loss[loss=0.1803, ctc_loss=0.1157, cr_loss=0.3227, over 17161.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3443, over 3326127.75 frames. ], batch size: 45, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:57:47,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=617488.6666666666, ans=0.125
2024-09-24 23:57:50,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=617488.6666666666, ans=0.5
2024-09-24 23:58:05,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=617535.3333333334, ans=0.125
2024-09-24 23:58:23,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=617582.0, ans=0.025
2024-09-24 23:58:26,285 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-24 23:58:26,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=617628.6666666666, ans=0.0
2024-09-24 23:58:46,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=617675.3333333334, ans=0.125
2024-09-24 23:58:47,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=617675.3333333334, ans=0.125
2024-09-24 23:58:50,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=617675.3333333334, ans=0.0
2024-09-24 23:58:56,794 INFO [train.py:1198] (1/4) Epoch 34, batch 3800, loss[loss=0.2194, ctc_loss=0.1512, cr_loss=0.3409, over 11426.00 frames. ], tot_loss[loss=0.1966, ctc_loss=0.1277, cr_loss=0.3443, over 3314469.80 frames. ], batch size: 123, lr: 3.48e-03, grad_scale: 32.0
2024-09-24 23:59:12,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=617768.6666666666, ans=0.125
2024-09-24 23:59:22,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=617768.6666666666, ans=0.09899494936611666
2024-09-24 23:59:23,294 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.281e+02 1.379e+02 1.537e+02 2.661e+02, threshold=2.757e+02, percent-clipped=0.0
2024-09-24 23:59:30,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=617815.3333333334, ans=0.125
2024-09-24 23:59:36,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=617815.3333333334, ans=0.0
2024-09-24 23:59:39,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=617815.3333333334, ans=0.1
2024-09-25 00:00:15,669 INFO [train.py:1198] (1/4) Epoch 34, batch 3850, loss[loss=0.2012, ctc_loss=0.1304, cr_loss=0.3539, over 16910.00 frames. ], tot_loss[loss=0.1984, ctc_loss=0.1291, cr_loss=0.3463, over 3283344.13 frames. ], batch size: 58, lr: 3.48e-03, grad_scale: 32.0
2024-09-25 00:00:39,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=618002.0, ans=0.125
2024-09-25 00:00:55,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=618048.6666666666, ans=0.125
2024-09-25 00:01:12,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=618095.3333333334, ans=0.1
2024-09-25 00:01:21,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=618142.0, ans=0.125
2024-09-25 00:02:16,955 INFO [train.py:1198] (1/4) Epoch 35, batch 0, loss[loss=0.2105, ctc_loss=0.1373, cr_loss=0.3661, over 17317.00 frames. ], tot_loss[loss=0.2105, ctc_loss=0.1373, cr_loss=0.3661, over 17317.00 frames. ], batch size: 51, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:02:16,955 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 00:02:32,192 INFO [train.py:1230] (1/4) Epoch 35, validation: loss=0.03449, ctc_loss=0.03449, cr_loss=9.757e-15, over 944034.00 frames. 
2024-09-25 00:02:32,193 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 00:03:08,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=618263.3333333334, ans=0.0
2024-09-25 00:03:10,088 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.401e+02 1.522e+02 1.667e+02 2.435e+02, threshold=3.044e+02, percent-clipped=0.0
2024-09-25 00:03:16,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=618263.3333333334, ans=0.0
2024-09-25 00:03:19,323 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.46 vs. limit=22.5
2024-09-25 00:03:21,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=618263.3333333334, ans=0.0
2024-09-25 00:03:56,377 INFO [train.py:1198] (1/4) Epoch 35, batch 50, loss[loss=0.1414, ctc_loss=0.08881, cr_loss=0.2628, over 17181.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1255, cr_loss=0.3441, over 765766.90 frames. ], batch size: 41, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:04:08,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=618403.3333333334, ans=0.125
2024-09-25 00:04:35,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=618496.6666666666, ans=0.125
2024-09-25 00:04:46,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=618543.3333333334, ans=0.0
2024-09-25 00:04:51,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.24 vs. limit=22.5
2024-09-25 00:04:52,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=618543.3333333334, ans=0.125
2024-09-25 00:05:04,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=618590.0, ans=0.0
2024-09-25 00:05:05,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=618590.0, ans=0.125
2024-09-25 00:05:16,501 INFO [train.py:1198] (1/4) Epoch 35, batch 100, loss[loss=0.1956, ctc_loss=0.1256, cr_loss=0.3503, over 17140.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1249, cr_loss=0.3433, over 1343658.41 frames. ], batch size: 48, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:05:29,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=618636.6666666666, ans=0.125
2024-09-25 00:05:49,884 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.242e+02 1.304e+02 1.413e+02 1.730e+02, threshold=2.607e+02, percent-clipped=0.0
2024-09-25 00:06:34,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=618823.3333333334, ans=0.125
2024-09-25 00:06:37,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=618870.0, ans=0.125
2024-09-25 00:06:38,902 INFO [train.py:1198] (1/4) Epoch 35, batch 150, loss[loss=0.2096, ctc_loss=0.1342, cr_loss=0.3773, over 17087.00 frames. ], tot_loss[loss=0.1945, ctc_loss=0.1257, cr_loss=0.344, over 1792614.64 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:06:47,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=618870.0, ans=0.1
2024-09-25 00:06:47,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=618870.0, ans=0.125
2024-09-25 00:07:19,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=618963.3333333334, ans=0.125
2024-09-25 00:07:33,766 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 00:07:38,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=619010.0, ans=0.125
2024-09-25 00:08:05,713 INFO [train.py:1198] (1/4) Epoch 35, batch 200, loss[loss=0.2138, ctc_loss=0.142, cr_loss=0.359, over 16806.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1244, cr_loss=0.3412, over 2149317.76 frames. ], batch size: 58, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:08:22,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=619150.0, ans=0.1
2024-09-25 00:08:23,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=619150.0, ans=0.1
2024-09-25 00:08:30,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=619150.0, ans=0.1
2024-09-25 00:08:39,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=619196.6666666666, ans=0.1
2024-09-25 00:08:41,112 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.252e+02 1.342e+02 1.492e+02 1.753e+02, threshold=2.683e+02, percent-clipped=0.0
2024-09-25 00:08:44,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=619196.6666666666, ans=0.125
2024-09-25 00:08:57,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=619243.3333333334, ans=0.125
2024-09-25 00:09:05,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=619243.3333333334, ans=0.125
2024-09-25 00:09:07,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=619243.3333333334, ans=0.125
2024-09-25 00:09:07,308 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=10.44 vs. limit=12.0
2024-09-25 00:09:07,629 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.71 vs. limit=12.0
2024-09-25 00:09:27,867 INFO [train.py:1198] (1/4) Epoch 35, batch 250, loss[loss=0.1739, ctc_loss=0.1089, cr_loss=0.3252, over 17155.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1251, cr_loss=0.3418, over 2410975.82 frames. ], batch size: 45, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:10:28,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=619476.6666666666, ans=0.1
2024-09-25 00:10:44,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=619523.3333333334, ans=0.0
2024-09-25 00:10:47,176 INFO [train.py:1198] (1/4) Epoch 35, batch 300, loss[loss=0.2129, ctc_loss=0.1426, cr_loss=0.3515, over 17063.00 frames. ], tot_loss[loss=0.1944, ctc_loss=0.1258, cr_loss=0.3431, over 2628307.99 frames. ], batch size: 52, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:10:49,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=619570.0, ans=0.125
2024-09-25 00:11:12,532 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.43 vs. limit=15.0
2024-09-25 00:11:20,985 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.276e+02 1.333e+02 1.416e+02 3.334e+02, threshold=2.666e+02, percent-clipped=1.0
2024-09-25 00:11:24,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=619663.3333333334, ans=0.125
2024-09-25 00:11:54,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=619756.6666666666, ans=0.125
2024-09-25 00:12:07,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=619756.6666666666, ans=0.125
2024-09-25 00:12:10,250 INFO [train.py:1198] (1/4) Epoch 35, batch 350, loss[loss=0.1838, ctc_loss=0.1167, cr_loss=0.3359, over 16949.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1262, cr_loss=0.3441, over 2783269.22 frames. ], batch size: 42, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:12:13,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=619803.3333333334, ans=0.0
2024-09-25 00:12:18,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=619803.3333333334, ans=0.5
2024-09-25 00:13:23,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=619990.0, ans=0.0
2024-09-25 00:13:39,190 INFO [train.py:1198] (1/4) Epoch 35, batch 400, loss[loss=0.1956, ctc_loss=0.126, cr_loss=0.3479, over 17314.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1267, cr_loss=0.345, over 2900538.02 frames. ], batch size: 51, lr: 3.42e-03, grad_scale: 32.0
2024-09-25 00:13:47,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=620036.6666666666, ans=0.5
2024-09-25 00:13:48,203 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.89 vs. limit=22.5
2024-09-25 00:14:12,447 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.297e+02 1.357e+02 1.460e+02 2.001e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-25 00:14:14,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=620130.0, ans=0.125
2024-09-25 00:14:23,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=620130.0, ans=0.0
2024-09-25 00:14:53,451 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.44 vs. limit=15.0
2024-09-25 00:14:59,149 INFO [train.py:1198] (1/4) Epoch 35, batch 450, loss[loss=0.1764, ctc_loss=0.1123, cr_loss=0.3204, over 17095.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1256, cr_loss=0.3429, over 3010296.97 frames. ], batch size: 40, lr: 3.42e-03, grad_scale: 16.0
2024-09-25 00:14:59,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=620270.0, ans=0.0
2024-09-25 00:14:59,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1.whitening_limit, batch_count=620270.0, ans=10.0
2024-09-25 00:15:12,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=620270.0, ans=0.0
2024-09-25 00:15:31,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=620363.3333333334, ans=0.2
2024-09-25 00:15:31,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=620363.3333333334, ans=0.125
2024-09-25 00:15:34,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=620363.3333333334, ans=0.025
2024-09-25 00:15:36,962 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.63 vs. limit=22.5
2024-09-25 00:15:43,061 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.54 vs. limit=22.5
2024-09-25 00:16:19,362 INFO [train.py:1198] (1/4) Epoch 35, batch 500, loss[loss=0.1982, ctc_loss=0.128, cr_loss=0.3513, over 16998.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1267, cr_loss=0.345, over 3081506.30 frames. ], batch size: 53, lr: 3.42e-03, grad_scale: 16.0
2024-09-25 00:16:21,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=620503.3333333334, ans=0.125
2024-09-25 00:16:22,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=620503.3333333334, ans=0.1
2024-09-25 00:16:55,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=620596.6666666666, ans=0.125
2024-09-25 00:16:57,097 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.242e+02 1.333e+02 1.438e+02 2.516e+02, threshold=2.666e+02, percent-clipped=0.0
2024-09-25 00:17:32,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=620690.0, ans=0.125
2024-09-25 00:17:35,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=620690.0, ans=0.05
2024-09-25 00:17:39,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=620690.0, ans=0.04949747468305833
2024-09-25 00:17:44,911 INFO [train.py:1198] (1/4) Epoch 35, batch 550, loss[loss=0.2116, ctc_loss=0.1362, cr_loss=0.3769, over 16994.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1265, cr_loss=0.3446, over 3143910.25 frames. ], batch size: 53, lr: 3.42e-03, grad_scale: 16.0
2024-09-25 00:17:51,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=620736.6666666666, ans=0.125
2024-09-25 00:18:00,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer_ff3.min_abs, batch_count=620736.6666666666, ans=0.2
2024-09-25 00:18:18,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=620783.3333333334, ans=0.125
2024-09-25 00:18:44,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=620876.6666666666, ans=0.125
2024-09-25 00:18:56,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=620923.3333333334, ans=0.5
2024-09-25 00:19:00,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=620923.3333333334, ans=0.0
2024-09-25 00:19:10,362 INFO [train.py:1198] (1/4) Epoch 35, batch 600, loss[loss=0.1921, ctc_loss=0.1252, cr_loss=0.3342, over 17222.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1264, cr_loss=0.3447, over 3187290.11 frames. ], batch size: 50, lr: 3.42e-03, grad_scale: 16.0
2024-09-25 00:19:28,849 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.20 vs. limit=15.0
2024-09-25 00:19:45,441 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.279e+02 1.403e+02 1.511e+02 1.952e+02, threshold=2.806e+02, percent-clipped=0.0
2024-09-25 00:19:50,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=621063.3333333334, ans=0.125
2024-09-25 00:19:52,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=621063.3333333334, ans=0.5
2024-09-25 00:20:03,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=621110.0, ans=0.125
2024-09-25 00:20:04,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=621110.0, ans=0.125
2024-09-25 00:20:08,931 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.65 vs. limit=15.0
2024-09-25 00:20:27,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=621156.6666666666, ans=0.1
2024-09-25 00:20:30,168 INFO [train.py:1198] (1/4) Epoch 35, batch 650, loss[loss=0.183, ctc_loss=0.1169, cr_loss=0.3306, over 17111.00 frames. ], tot_loss[loss=0.194, ctc_loss=0.1254, cr_loss=0.343, over 3231583.88 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 16.0
2024-09-25 00:20:33,916 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 00:20:35,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=621203.3333333334, ans=0.125
2024-09-25 00:20:37,214 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 00:20:48,349 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.58 vs. limit=15.0
2024-09-25 00:21:40,653 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 00:21:53,028 INFO [train.py:1198] (1/4) Epoch 35, batch 700, loss[loss=0.1862, ctc_loss=0.1178, cr_loss=0.3419, over 17266.00 frames. ], tot_loss[loss=0.1944, ctc_loss=0.1257, cr_loss=0.3438, over 3262569.99 frames. ], batch size: 44, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:22:30,889 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.276e+02 1.349e+02 1.443e+02 1.700e+02, threshold=2.699e+02, percent-clipped=0.0
2024-09-25 00:22:43,448 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.32 vs. limit=15.0
2024-09-25 00:22:50,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=621576.6666666666, ans=0.125
2024-09-25 00:23:21,253 INFO [train.py:1198] (1/4) Epoch 35, batch 750, loss[loss=0.2008, ctc_loss=0.1287, cr_loss=0.3602, over 17209.00 frames. ], tot_loss[loss=0.1952, ctc_loss=0.1263, cr_loss=0.3445, over 3272340.39 frames. ], batch size: 50, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:23:24,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=621670.0, ans=0.1
2024-09-25 00:23:29,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=621670.0, ans=0.125
2024-09-25 00:23:38,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=621716.6666666666, ans=0.1
2024-09-25 00:23:42,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=621716.6666666666, ans=0.0
2024-09-25 00:23:47,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=621716.6666666666, ans=0.2
2024-09-25 00:24:06,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=621763.3333333334, ans=0.025
2024-09-25 00:24:37,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=621856.6666666666, ans=0.2
2024-09-25 00:24:41,687 INFO [train.py:1198] (1/4) Epoch 35, batch 800, loss[loss=0.225, ctc_loss=0.1518, cr_loss=0.3663, over 14917.00 frames. ], tot_loss[loss=0.1945, ctc_loss=0.1259, cr_loss=0.3429, over 3283030.07 frames. ], batch size: 89, lr: 3.41e-03, grad_scale: 32.0
2024-09-25 00:25:06,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=621950.0, ans=0.125
2024-09-25 00:25:07,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=621950.0, ans=0.125
2024-09-25 00:25:09,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=621950.0, ans=0.0
2024-09-25 00:25:14,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=621996.6666666666, ans=0.2
2024-09-25 00:25:16,940 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.280e+02 1.366e+02 1.481e+02 2.443e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-25 00:25:30,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=622043.3333333334, ans=0.0
2024-09-25 00:25:35,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=622043.3333333334, ans=0.125
2024-09-25 00:26:00,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=622136.6666666666, ans=0.125
2024-09-25 00:26:00,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=622136.6666666666, ans=0.2
2024-09-25 00:26:01,918 INFO [train.py:1198] (1/4) Epoch 35, batch 850, loss[loss=0.2051, ctc_loss=0.1335, cr_loss=0.3579, over 15988.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1266, cr_loss=0.3446, over 3299715.38 frames. ], batch size: 74, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:27:26,638 INFO [train.py:1198] (1/4) Epoch 35, batch 900, loss[loss=0.1451, ctc_loss=0.09134, cr_loss=0.2689, over 17268.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1267, cr_loss=0.3446, over 3306293.72 frames. ], batch size: 42, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:27:47,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=622416.6666666666, ans=0.025
2024-09-25 00:28:08,746 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.305e+02 1.377e+02 1.455e+02 1.762e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-25 00:28:22,498 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.81 vs. limit=12.0
2024-09-25 00:28:36,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=622556.6666666666, ans=0.0
2024-09-25 00:28:42,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=622556.6666666666, ans=0.125
2024-09-25 00:28:52,251 INFO [train.py:1198] (1/4) Epoch 35, batch 950, loss[loss=0.2076, ctc_loss=0.1382, cr_loss=0.347, over 16106.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1271, cr_loss=0.3449, over 3320427.98 frames. ], batch size: 74, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:28:57,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=622603.3333333334, ans=0.2
2024-09-25 00:29:03,192 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.74 vs. limit=10.0
2024-09-25 00:29:27,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=622696.6666666666, ans=0.1
2024-09-25 00:29:52,155 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=622743.3333333334, ans=0.125
2024-09-25 00:29:55,321 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 00:29:55,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=622790.0, ans=0.125
2024-09-25 00:30:06,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=622790.0, ans=0.2
2024-09-25 00:30:06,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=622790.0, ans=0.025
2024-09-25 00:30:09,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=622790.0, ans=0.2
2024-09-25 00:30:12,544 INFO [train.py:1198] (1/4) Epoch 35, batch 1000, loss[loss=0.2164, ctc_loss=0.1413, cr_loss=0.3756, over 17027.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1266, cr_loss=0.344, over 3336201.37 frames. ], batch size: 51, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:30:14,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=622836.6666666666, ans=0.1
2024-09-25 00:30:49,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.266e+02 1.340e+02 1.444e+02 2.744e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-25 00:31:03,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=622976.6666666666, ans=0.04949747468305833
2024-09-25 00:31:08,949 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.81 vs. limit=22.5
2024-09-25 00:31:10,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=622976.6666666666, ans=0.125
2024-09-25 00:31:35,002 INFO [train.py:1198] (1/4) Epoch 35, batch 1050, loss[loss=0.1671, ctc_loss=0.1056, cr_loss=0.3075, over 17174.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1266, cr_loss=0.3445, over 3346009.91 frames. ], batch size: 41, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:31:36,312 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.80 vs. limit=15.0
2024-09-25 00:32:27,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=623210.0, ans=0.125
2024-09-25 00:32:33,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=623210.0, ans=0.2
2024-09-25 00:32:41,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=623256.6666666666, ans=0.07
2024-09-25 00:32:50,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=623256.6666666666, ans=0.125
2024-09-25 00:32:58,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten.whitening_limit, batch_count=623256.6666666666, ans=22.5
2024-09-25 00:32:59,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=623256.6666666666, ans=0.125
2024-09-25 00:33:01,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=623303.3333333334, ans=0.0
2024-09-25 00:33:02,769 INFO [train.py:1198] (1/4) Epoch 35, batch 1100, loss[loss=0.208, ctc_loss=0.1337, cr_loss=0.3714, over 17058.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1271, cr_loss=0.3452, over 3349533.11 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:33:25,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=623350.0, ans=0.2
2024-09-25 00:33:31,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=623350.0, ans=0.2
2024-09-25 00:33:39,206 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.286e+02 1.347e+02 1.466e+02 2.468e+02, threshold=2.694e+02, percent-clipped=0.0
2024-09-25 00:34:02,355 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 00:34:02,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.07 vs. limit=15.0
2024-09-25 00:34:12,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=623490.0, ans=0.125
2024-09-25 00:34:15,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=623490.0, ans=0.0
2024-09-25 00:34:22,932 INFO [train.py:1198] (1/4) Epoch 35, batch 1150, loss[loss=0.2319, ctc_loss=0.1544, cr_loss=0.3875, over 15060.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3448, over 3346016.00 frames. ], batch size: 89, lr: 3.41e-03, grad_scale: 16.0
2024-09-25 00:34:37,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=623583.3333333334, ans=0.125
2024-09-25 00:34:40,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=623583.3333333334, ans=0.05
2024-09-25 00:35:12,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=623676.6666666666, ans=0.125
2024-09-25 00:35:31,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=623723.3333333334, ans=0.04949747468305833
2024-09-25 00:35:42,222 INFO [train.py:1198] (1/4) Epoch 35, batch 1200, loss[loss=0.1975, ctc_loss=0.1293, cr_loss=0.3409, over 17073.00 frames. ], tot_loss[loss=0.1949, ctc_loss=0.1262, cr_loss=0.3432, over 3353951.29 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 32.0
2024-09-25 00:35:42,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=623770.0, ans=0.1
2024-09-25 00:36:18,955 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.278e+02 1.375e+02 1.481e+02 3.565e+02, threshold=2.751e+02, percent-clipped=1.0
2024-09-25 00:36:49,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=623956.6666666666, ans=0.2
2024-09-25 00:36:49,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=623956.6666666666, ans=0.1
2024-09-25 00:37:04,858 INFO [train.py:1198] (1/4) Epoch 35, batch 1250, loss[loss=0.202, ctc_loss=0.1326, cr_loss=0.3474, over 17135.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1265, cr_loss=0.3438, over 3354958.95 frames. ], batch size: 48, lr: 3.41e-03, grad_scale: 32.0
2024-09-25 00:37:13,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=624003.3333333334, ans=0.0
2024-09-25 00:37:17,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=624003.3333333334, ans=0.2
2024-09-25 00:37:44,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=624096.6666666666, ans=0.2
2024-09-25 00:37:53,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=624096.6666666666, ans=0.125
2024-09-25 00:38:00,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=624143.3333333334, ans=0.2
2024-09-25 00:38:32,164 INFO [train.py:1198] (1/4) Epoch 35, batch 1300, loss[loss=0.2172, ctc_loss=0.1401, cr_loss=0.3856, over 16905.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1269, cr_loss=0.3447, over 3364291.74 frames. ], batch size: 58, lr: 3.41e-03, grad_scale: 32.0
2024-09-25 00:38:33,035 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.65 vs. limit=15.0
2024-09-25 00:38:38,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=624236.6666666666, ans=0.0
2024-09-25 00:38:40,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=624236.6666666666, ans=0.0
2024-09-25 00:38:46,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=624283.3333333334, ans=0.125
2024-09-25 00:39:07,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=624330.0, ans=0.0
2024-09-25 00:39:08,851 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.267e+02 1.378e+02 1.452e+02 1.774e+02, threshold=2.755e+02, percent-clipped=0.0
2024-09-25 00:39:26,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.min_positive, batch_count=624376.6666666666, ans=0.05
2024-09-25 00:39:28,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=624376.6666666666, ans=0.2
2024-09-25 00:39:38,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=624423.3333333334, ans=0.2
2024-09-25 00:39:52,134 INFO [train.py:1198] (1/4) Epoch 35, batch 1350, loss[loss=0.2104, ctc_loss=0.1372, cr_loss=0.3662, over 17108.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3445, over 3363794.84 frames. ], batch size: 49, lr: 3.41e-03, grad_scale: 32.0
2024-09-25 00:39:52,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.69 vs. limit=6.0
2024-09-25 00:39:54,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=624470.0, ans=0.2
2024-09-25 00:39:58,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=624470.0, ans=0.2
2024-09-25 00:40:06,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=624516.6666666666, ans=0.025
2024-09-25 00:40:51,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=624610.0, ans=0.0
2024-09-25 00:41:03,037 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.54 vs. limit=10.0
2024-09-25 00:41:11,914 INFO [train.py:1198] (1/4) Epoch 35, batch 1400, loss[loss=0.2261, ctc_loss=0.1501, cr_loss=0.38, over 16907.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1269, cr_loss=0.3444, over 3357830.54 frames. ], batch size: 58, lr: 3.41e-03, grad_scale: 32.0
2024-09-25 00:41:51,231 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.292e+02 1.373e+02 1.479e+02 2.692e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-25 00:41:59,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=624796.6666666666, ans=0.0
2024-09-25 00:42:02,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=624843.3333333334, ans=0.0
2024-09-25 00:42:18,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=624843.3333333334, ans=0.1
2024-09-25 00:42:37,117 INFO [train.py:1198] (1/4) Epoch 35, batch 1450, loss[loss=0.1894, ctc_loss=0.1244, cr_loss=0.3252, over 17023.00 frames. ], tot_loss[loss=0.1952, ctc_loss=0.1265, cr_loss=0.3437, over 3362373.17 frames. ], batch size: 51, lr: 3.41e-03, grad_scale: 32.0
2024-09-25 00:42:48,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=624936.6666666666, ans=0.1
2024-09-25 00:43:44,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=625123.3333333334, ans=0.1
2024-09-25 00:43:57,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=625123.3333333334, ans=0.2
2024-09-25 00:44:02,342 INFO [train.py:1198] (1/4) Epoch 35, batch 1500, loss[loss=0.1978, ctc_loss=0.1274, cr_loss=0.3519, over 17207.00 frames. ], tot_loss[loss=0.1944, ctc_loss=0.1259, cr_loss=0.3425, over 3361532.69 frames. ], batch size: 47, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:44:06,646 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.57 vs. limit=15.0
2024-09-25 00:44:15,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=625170.0, ans=0.5
2024-09-25 00:44:35,231 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.49 vs. limit=15.0
2024-09-25 00:44:39,191 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.249e+02 1.364e+02 1.434e+02 2.230e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-25 00:45:23,011 INFO [train.py:1198] (1/4) Epoch 35, batch 1550, loss[loss=0.1958, ctc_loss=0.128, cr_loss=0.3389, over 17279.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.3438, over 3355702.66 frames. ], batch size: 49, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:45:39,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=625450.0, ans=0.2
2024-09-25 00:45:45,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=625450.0, ans=0.1
2024-09-25 00:45:53,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=625496.6666666666, ans=0.125
2024-09-25 00:46:01,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=625496.6666666666, ans=0.025
2024-09-25 00:46:11,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=625543.3333333334, ans=0.125
2024-09-25 00:46:27,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=625590.0, ans=0.125
2024-09-25 00:46:29,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=625590.0, ans=0.125
2024-09-25 00:46:30,976 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.85 vs. limit=15.0
2024-09-25 00:46:45,799 INFO [train.py:1198] (1/4) Epoch 35, batch 1600, loss[loss=0.2207, ctc_loss=0.1451, cr_loss=0.378, over 16542.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1272, cr_loss=0.3448, over 3358981.60 frames. ], batch size: 66, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:46:55,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=625636.6666666666, ans=0.2
2024-09-25 00:47:19,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=625730.0, ans=0.125
2024-09-25 00:47:21,340 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.14 vs. limit=6.0
2024-09-25 00:47:25,306 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.273e+02 1.348e+02 1.440e+02 1.761e+02, threshold=2.695e+02, percent-clipped=0.0
2024-09-25 00:47:29,258 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.73 vs. limit=15.0
2024-09-25 00:47:33,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer_ff2.min_abs, batch_count=625730.0, ans=0.1
2024-09-25 00:47:46,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=625776.6666666666, ans=0.0
2024-09-25 00:47:50,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=625776.6666666666, ans=0.1
2024-09-25 00:48:14,374 INFO [train.py:1198] (1/4) Epoch 35, batch 1650, loss[loss=0.1627, ctc_loss=0.1034, cr_loss=0.2967, over 16962.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1273, cr_loss=0.3451, over 3343838.58 frames. ], batch size: 42, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:48:18,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=625870.0, ans=0.0
2024-09-25 00:48:28,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=625916.6666666666, ans=0.125
2024-09-25 00:49:08,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=626010.0, ans=0.125
2024-09-25 00:49:09,100 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=626010.0, ans=0.125
2024-09-25 00:49:21,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=626056.6666666666, ans=0.1
2024-09-25 00:49:23,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=626056.6666666666, ans=0.125
2024-09-25 00:49:23,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=626056.6666666666, ans=0.125
2024-09-25 00:49:26,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=626056.6666666666, ans=0.0
2024-09-25 00:49:34,599 INFO [train.py:1198] (1/4) Epoch 35, batch 1700, loss[loss=0.2253, ctc_loss=0.1498, cr_loss=0.3771, over 15932.00 frames. ], tot_loss[loss=0.1968, ctc_loss=0.1277, cr_loss=0.3455, over 3341263.70 frames. ], batch size: 74, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:50:10,954 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.273e+02 1.359e+02 1.469e+02 2.264e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 00:50:14,679 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.14 vs. limit=15.0
2024-09-25 00:50:31,041 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.36 vs. limit=15.0
2024-09-25 00:50:34,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=626243.3333333334, ans=0.2
2024-09-25 00:50:40,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=626290.0, ans=0.125
2024-09-25 00:50:44,280 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.35 vs. limit=22.5
2024-09-25 00:50:51,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=626290.0, ans=0.1
2024-09-25 00:50:54,471 INFO [train.py:1198] (1/4) Epoch 35, batch 1750, loss[loss=0.1505, ctc_loss=0.09355, cr_loss=0.2847, over 17212.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1272, cr_loss=0.345, over 3347287.47 frames. ], batch size: 41, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:51:04,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=626336.6666666666, ans=0.04949747468305833
2024-09-25 00:51:05,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=626336.6666666666, ans=0.0
2024-09-25 00:51:17,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.40 vs. limit=15.0
2024-09-25 00:51:51,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=626476.6666666666, ans=0.0
2024-09-25 00:51:51,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=626476.6666666666, ans=0.125
2024-09-25 00:52:07,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=626523.3333333334, ans=0.125
2024-09-25 00:52:19,237 INFO [train.py:1198] (1/4) Epoch 35, batch 1800, loss[loss=0.2052, ctc_loss=0.1326, cr_loss=0.3629, over 17026.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1269, cr_loss=0.3441, over 3358907.84 frames. ], batch size: 56, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:52:24,682 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.82 vs. limit=15.0
2024-09-25 00:52:48,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=626616.6666666666, ans=0.1
2024-09-25 00:53:00,986 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.248e+02 1.342e+02 1.447e+02 1.901e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-25 00:53:44,285 INFO [train.py:1198] (1/4) Epoch 35, batch 1850, loss[loss=0.218, ctc_loss=0.1441, cr_loss=0.3695, over 15996.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1267, cr_loss=0.3435, over 3362625.25 frames. ], batch size: 74, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:54:03,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=626850.0, ans=0.0
2024-09-25 00:54:24,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=626896.6666666666, ans=0.07
2024-09-25 00:54:38,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=626943.3333333334, ans=0.1
2024-09-25 00:54:56,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=626990.0, ans=0.2
2024-09-25 00:55:04,162 INFO [train.py:1198] (1/4) Epoch 35, batch 1900, loss[loss=0.2056, ctc_loss=0.1333, cr_loss=0.3615, over 16503.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.127, cr_loss=0.3447, over 3363806.08 frames. ], batch size: 66, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:55:41,025 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.333e+02 1.409e+02 1.508e+02 2.528e+02, threshold=2.819e+02, percent-clipped=0.0
2024-09-25 00:55:46,575 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.08 vs. limit=22.5
2024-09-25 00:55:49,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=627130.0, ans=0.0
2024-09-25 00:56:03,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=627176.6666666666, ans=0.2
2024-09-25 00:56:05,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=627176.6666666666, ans=0.125
2024-09-25 00:56:19,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=627223.3333333334, ans=0.125
2024-09-25 00:56:24,333 INFO [train.py:1198] (1/4) Epoch 35, batch 1950, loss[loss=0.1881, ctc_loss=0.1207, cr_loss=0.3375, over 17236.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.3439, over 3372137.35 frames. ], batch size: 50, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:56:30,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=627270.0, ans=0.025
2024-09-25 00:56:38,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=627270.0, ans=0.05
2024-09-25 00:56:39,983 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 00:57:07,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=627363.3333333334, ans=0.125
2024-09-25 00:57:07,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=627363.3333333334, ans=22.5
2024-09-25 00:57:19,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=627410.0, ans=0.125
2024-09-25 00:57:32,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=627456.6666666666, ans=0.0
2024-09-25 00:57:37,147 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.66 vs. limit=22.5
2024-09-25 00:57:44,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=627456.6666666666, ans=0.1
2024-09-25 00:57:49,087 INFO [train.py:1198] (1/4) Epoch 35, batch 2000, loss[loss=0.1696, ctc_loss=0.105, cr_loss=0.3229, over 16937.00 frames. ], tot_loss[loss=0.1951, ctc_loss=0.1264, cr_loss=0.3436, over 3372039.84 frames. ], batch size: 42, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:58:00,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=627503.3333333334, ans=0.0
2024-09-25 00:58:00,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=627503.3333333334, ans=0.025
2024-09-25 00:58:16,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=627550.0, ans=0.0
2024-09-25 00:58:21,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=627550.0, ans=0.2
2024-09-25 00:58:27,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=627596.6666666666, ans=0.1
2024-09-25 00:58:30,774 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.295e+02 1.390e+02 1.531e+02 2.683e+02, threshold=2.780e+02, percent-clipped=0.0
2024-09-25 00:58:33,605 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.59 vs. limit=5.0
2024-09-25 00:58:40,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=627643.3333333334, ans=0.2
2024-09-25 00:58:40,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=627643.3333333334, ans=0.2
2024-09-25 00:58:46,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=627643.3333333334, ans=0.04949747468305833
2024-09-25 00:58:47,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.92 vs. limit=6.0
2024-09-25 00:58:50,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=627643.3333333334, ans=0.125
2024-09-25 00:58:53,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=627643.3333333334, ans=0.2
2024-09-25 00:58:59,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=627690.0, ans=0.0
2024-09-25 00:59:13,924 INFO [train.py:1198] (1/4) Epoch 35, batch 2050, loss[loss=0.2251, ctc_loss=0.1462, cr_loss=0.3942, over 17026.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.1261, cr_loss=0.3431, over 3368234.99 frames. ], batch size: 56, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 00:59:51,394 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.73 vs. limit=15.0
2024-09-25 01:00:07,539 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.84 vs. limit=10.0
2024-09-25 01:00:25,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=627923.3333333334, ans=0.125
2024-09-25 01:00:29,370 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.23 vs. limit=15.0
2024-09-25 01:00:33,291 INFO [train.py:1198] (1/4) Epoch 35, batch 2100, loss[loss=0.1766, ctc_loss=0.1119, cr_loss=0.3233, over 17195.00 frames. ], tot_loss[loss=0.1948, ctc_loss=0.1261, cr_loss=0.3434, over 3363634.04 frames. ], batch size: 41, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 01:00:38,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=627970.0, ans=0.125
2024-09-25 01:00:51,699 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:01:10,508 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.271e+02 1.342e+02 1.455e+02 1.760e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-25 01:01:40,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=628156.6666666666, ans=0.125
2024-09-25 01:01:42,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=628156.6666666666, ans=0.125
2024-09-25 01:01:56,292 INFO [train.py:1198] (1/4) Epoch 35, batch 2150, loss[loss=0.1919, ctc_loss=0.1242, cr_loss=0.3386, over 17311.00 frames. ], tot_loss[loss=0.1944, ctc_loss=0.1258, cr_loss=0.343, over 3373857.01 frames. ], batch size: 51, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 01:02:09,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2.whitening_limit, batch_count=628203.3333333334, ans=15.0
2024-09-25 01:02:25,886 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.94 vs. limit=5.0
2024-09-25 01:02:36,683 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.26 vs. limit=12.0
2024-09-25 01:03:13,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=628390.0, ans=0.125
2024-09-25 01:03:14,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=628390.0, ans=0.1
2024-09-25 01:03:18,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=628390.0, ans=0.025
2024-09-25 01:03:24,295 INFO [train.py:1198] (1/4) Epoch 35, batch 2200, loss[loss=0.214, ctc_loss=0.1408, cr_loss=0.3659, over 17027.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1256, cr_loss=0.3425, over 3381837.44 frames. ], batch size: 52, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 01:03:26,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.min_positive, batch_count=628436.6666666666, ans=0.05
2024-09-25 01:03:34,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=628436.6666666666, ans=0.0
2024-09-25 01:03:59,277 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.24 vs. limit=15.0
2024-09-25 01:04:01,427 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.291e+02 1.359e+02 1.454e+02 2.045e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 01:04:22,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=628576.6666666666, ans=0.0
2024-09-25 01:04:30,208 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.34 vs. limit=10.0
2024-09-25 01:04:33,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=628623.3333333334, ans=0.0
2024-09-25 01:04:45,059 INFO [train.py:1198] (1/4) Epoch 35, batch 2250, loss[loss=0.2098, ctc_loss=0.1344, cr_loss=0.3772, over 17078.00 frames. ], tot_loss[loss=0.1939, ctc_loss=0.1253, cr_loss=0.3426, over 3385616.96 frames. ], batch size: 43, lr: 3.40e-03, grad_scale: 32.0
2024-09-25 01:04:51,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer_na.min_abs, batch_count=628670.0, ans=0.02
2024-09-25 01:04:58,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=628670.0, ans=0.125
2024-09-25 01:05:03,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=628716.6666666666, ans=0.125
2024-09-25 01:05:31,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=628810.0, ans=0.2
2024-09-25 01:05:54,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=628856.6666666666, ans=0.0
2024-09-25 01:05:57,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=628856.6666666666, ans=0.125
2024-09-25 01:05:58,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.whiten.whitening_limit, batch_count=628856.6666666666, ans=15.0
2024-09-25 01:05:59,099 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=628856.6666666666, ans=0.125
2024-09-25 01:06:05,297 INFO [train.py:1198] (1/4) Epoch 35, batch 2300, loss[loss=0.2266, ctc_loss=0.1503, cr_loss=0.3816, over 15066.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1253, cr_loss=0.3422, over 3383747.41 frames. ], batch size: 89, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:06:38,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=628996.6666666666, ans=0.07
2024-09-25 01:06:44,417 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.311e+02 1.390e+02 1.542e+02 2.527e+02, threshold=2.780e+02, percent-clipped=0.0
2024-09-25 01:06:48,286 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.90 vs. limit=15.0
2024-09-25 01:07:19,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=629090.0, ans=0.125
2024-09-25 01:07:26,190 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.41 vs. limit=15.0
2024-09-25 01:07:30,082 INFO [train.py:1198] (1/4) Epoch 35, batch 2350, loss[loss=0.1708, ctc_loss=0.1062, cr_loss=0.3231, over 17263.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1252, cr_loss=0.3422, over 3382371.48 frames. ], batch size: 44, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:07:36,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=629136.6666666666, ans=0.125
2024-09-25 01:07:52,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=629183.3333333334, ans=0.2
2024-09-25 01:07:55,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=629183.3333333334, ans=0.1
2024-09-25 01:08:15,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=629230.0, ans=0.125
2024-09-25 01:08:16,004 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.87 vs. limit=6.0
2024-09-25 01:08:16,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.83 vs. limit=10.0
2024-09-25 01:08:34,903 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:08:55,602 INFO [train.py:1198] (1/4) Epoch 35, batch 2400, loss[loss=0.2076, ctc_loss=0.1331, cr_loss=0.3729, over 17187.00 frames. ], tot_loss[loss=0.194, ctc_loss=0.1255, cr_loss=0.3423, over 3378002.59 frames. ], batch size: 55, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:08:57,612 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:09:08,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=629370.0, ans=0.125
2024-09-25 01:09:09,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=629370.0, ans=0.125
2024-09-25 01:09:18,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=629416.6666666666, ans=0.2
2024-09-25 01:09:22,044 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.35 vs. limit=22.5
2024-09-25 01:09:32,456 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.256e+02 1.334e+02 1.415e+02 2.339e+02, threshold=2.669e+02, percent-clipped=0.0
2024-09-25 01:09:34,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=629463.3333333334, ans=0.125
2024-09-25 01:09:47,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=629510.0, ans=0.2
2024-09-25 01:09:51,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=629510.0, ans=0.0
2024-09-25 01:09:58,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_positive, batch_count=629556.6666666666, ans=0.05
2024-09-25 01:10:15,371 INFO [train.py:1198] (1/4) Epoch 35, batch 2450, loss[loss=0.1916, ctc_loss=0.1227, cr_loss=0.3449, over 17156.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1255, cr_loss=0.3418, over 3359746.12 frames. ], batch size: 45, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:10:33,033 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:10:46,291 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=629696.6666666666, ans=0.125
2024-09-25 01:11:22,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.42 vs. limit=10.0
2024-09-25 01:11:37,758 INFO [train.py:1198] (1/4) Epoch 35, batch 2500, loss[loss=0.1441, ctc_loss=0.08861, cr_loss=0.2776, over 17053.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1256, cr_loss=0.3423, over 3352001.34 frames. ], batch size: 39, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:11:47,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=629836.6666666666, ans=0.125
2024-09-25 01:12:16,800 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.245e+02 1.351e+02 1.450e+02 1.965e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-25 01:12:25,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=629930.0, ans=0.125
2024-09-25 01:12:31,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=629976.6666666666, ans=0.125
2024-09-25 01:12:39,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=629976.6666666666, ans=0.1
2024-09-25 01:12:39,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=629976.6666666666, ans=0.07
2024-09-25 01:13:05,480 INFO [train.py:1198] (1/4) Epoch 35, batch 2550, loss[loss=0.1767, ctc_loss=0.1133, cr_loss=0.3173, over 16319.00 frames. ], tot_loss[loss=0.1944, ctc_loss=0.1259, cr_loss=0.3427, over 3355657.42 frames. ], batch size: 36, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:13:11,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=630070.0, ans=0.125
2024-09-25 01:13:32,932 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.99 vs. limit=15.0
2024-09-25 01:14:25,411 INFO [train.py:1198] (1/4) Epoch 35, batch 2600, loss[loss=0.1809, ctc_loss=0.1173, cr_loss=0.3182, over 17299.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1265, cr_loss=0.3441, over 3358621.89 frames. ], batch size: 49, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:14:54,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=630350.0, ans=0.125
2024-09-25 01:15:02,358 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.261e+02 1.318e+02 1.420e+02 1.911e+02, threshold=2.636e+02, percent-clipped=0.0
2024-09-25 01:15:16,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=630443.3333333334, ans=0.125
2024-09-25 01:15:26,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=630443.3333333334, ans=0.125
2024-09-25 01:15:45,579 INFO [train.py:1198] (1/4) Epoch 35, batch 2650, loss[loss=0.2143, ctc_loss=0.1385, cr_loss=0.3789, over 15867.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1272, cr_loss=0.3452, over 3364225.82 frames. ], batch size: 74, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:15:45,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=630536.6666666666, ans=0.1
2024-09-25 01:15:52,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=630536.6666666666, ans=0.125
2024-09-25 01:15:52,877 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.20 vs. limit=6.0
2024-09-25 01:16:03,791 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.38 vs. limit=22.5
2024-09-25 01:16:13,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=630583.3333333334, ans=0.125
2024-09-25 01:16:18,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten.whitening_limit, batch_count=630630.0, ans=22.5
2024-09-25 01:16:19,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=630630.0, ans=0.0
2024-09-25 01:16:58,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=630723.3333333334, ans=0.0
2024-09-25 01:16:59,637 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.31 vs. limit=12.0
2024-09-25 01:17:05,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=630723.3333333334, ans=0.2
2024-09-25 01:17:08,143 INFO [train.py:1198] (1/4) Epoch 35, batch 2700, loss[loss=0.2024, ctc_loss=0.1338, cr_loss=0.3431, over 17322.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1273, cr_loss=0.3459, over 3359706.48 frames. ], batch size: 51, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:17:23,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=630770.0, ans=0.1
2024-09-25 01:17:30,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=630816.6666666666, ans=0.125
2024-09-25 01:17:44,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=630863.3333333334, ans=0.125
2024-09-25 01:17:52,694 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.276e+02 1.350e+02 1.459e+02 1.855e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-25 01:18:10,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=630910.0, ans=0.2
2024-09-25 01:18:25,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=630956.6666666666, ans=0.1
2024-09-25 01:18:33,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=630956.6666666666, ans=0.125
2024-09-25 01:18:36,126 INFO [train.py:1198] (1/4) Epoch 35, batch 2750, loss[loss=0.2285, ctc_loss=0.1502, cr_loss=0.3915, over 16675.00 frames. ], tot_loss[loss=0.1966, ctc_loss=0.1274, cr_loss=0.3459, over 3363153.15 frames. ], batch size: 61, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:18:46,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=631003.3333333334, ans=0.025
2024-09-25 01:18:46,748 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.54 vs. limit=22.5
2024-09-25 01:19:12,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=631096.6666666666, ans=0.125
2024-09-25 01:19:23,409 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:19:29,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=631143.3333333334, ans=0.1
2024-09-25 01:19:48,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=631190.0, ans=0.125
2024-09-25 01:19:49,193 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.87 vs. limit=15.0
2024-09-25 01:19:50,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=631190.0, ans=0.125
2024-09-25 01:19:54,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=631236.6666666666, ans=0.0
2024-09-25 01:19:56,334 INFO [train.py:1198] (1/4) Epoch 35, batch 2800, loss[loss=0.2165, ctc_loss=0.1418, cr_loss=0.3736, over 17211.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1271, cr_loss=0.3447, over 3359267.80 frames. ], batch size: 55, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:20:08,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=631236.6666666666, ans=0.125
2024-09-25 01:20:33,122 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.248e+02 1.324e+02 1.411e+02 2.015e+02, threshold=2.647e+02, percent-clipped=0.0
2024-09-25 01:20:49,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=631376.6666666666, ans=0.125
2024-09-25 01:20:56,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.85 vs. limit=15.0
2024-09-25 01:21:16,333 INFO [train.py:1198] (1/4) Epoch 35, batch 2850, loss[loss=0.1668, ctc_loss=0.1084, cr_loss=0.2922, over 17060.00 frames. ], tot_loss[loss=0.1956, ctc_loss=0.1268, cr_loss=0.3441, over 3358085.67 frames. ], batch size: 46, lr: 3.39e-03, grad_scale: 64.0
2024-09-25 01:21:29,537 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.88 vs. limit=10.0
2024-09-25 01:21:36,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=631516.6666666666, ans=0.125
2024-09-25 01:21:38,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=631516.6666666666, ans=0.125
2024-09-25 01:21:56,256 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.30 vs. limit=22.5
2024-09-25 01:22:00,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=631563.3333333334, ans=10.0
2024-09-25 01:22:02,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=631563.3333333334, ans=0.0
2024-09-25 01:22:14,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=631610.0, ans=0.0
2024-09-25 01:22:22,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=631610.0, ans=0.0
2024-09-25 01:22:46,461 INFO [train.py:1198] (1/4) Epoch 35, batch 2900, loss[loss=0.1752, ctc_loss=0.1133, cr_loss=0.3096, over 17066.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3448, over 3358625.14 frames. ], batch size: 43, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:22:46,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=631703.3333333334, ans=0.0
2024-09-25 01:22:51,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=631703.3333333334, ans=0.1
2024-09-25 01:22:51,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=631703.3333333334, ans=0.125
2024-09-25 01:23:04,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=631750.0, ans=0.125
2024-09-25 01:23:12,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1.whitening_limit, batch_count=631750.0, ans=10.0
2024-09-25 01:23:24,648 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.283e+02 1.348e+02 1.424e+02 1.926e+02, threshold=2.696e+02, percent-clipped=0.0
2024-09-25 01:23:40,186 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.90 vs. limit=15.0
2024-09-25 01:24:01,042 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.05 vs. limit=15.0
2024-09-25 01:24:06,673 INFO [train.py:1198] (1/4) Epoch 35, batch 2950, loss[loss=0.1915, ctc_loss=0.122, cr_loss=0.3475, over 17054.00 frames. ], tot_loss[loss=0.1963, ctc_loss=0.1272, cr_loss=0.3452, over 3352662.47 frames. ], batch size: 46, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:24:15,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=631936.6666666666, ans=0.1
2024-09-25 01:24:23,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.79 vs. limit=22.5
2024-09-25 01:24:34,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=631983.3333333334, ans=0.0
2024-09-25 01:24:55,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=632076.6666666666, ans=0.035
2024-09-25 01:25:20,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=632123.3333333334, ans=0.07
2024-09-25 01:25:26,966 INFO [train.py:1198] (1/4) Epoch 35, batch 3000, loss[loss=0.1962, ctc_loss=0.1269, cr_loss=0.3465, over 17363.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1274, cr_loss=0.3457, over 3358051.88 frames. ], batch size: 48, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:25:26,967 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 01:25:42,187 INFO [train.py:1230] (1/4) Epoch 35, validation: loss=0.03538, ctc_loss=0.03538, cr_loss=9.094e-15, over 944034.00 frames. 
2024-09-25 01:25:42,188 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 01:25:42,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=632170.0, ans=0.0
2024-09-25 01:25:42,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=632170.0, ans=0.125
2024-09-25 01:25:45,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=632170.0, ans=0.0
2024-09-25 01:26:19,541 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.295e+02 1.357e+02 1.448e+02 2.181e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-25 01:26:37,428 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.75 vs. limit=15.0
2024-09-25 01:26:54,252 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.56 vs. limit=15.0
2024-09-25 01:27:02,790 INFO [train.py:1198] (1/4) Epoch 35, batch 3050, loss[loss=0.2071, ctc_loss=0.1338, cr_loss=0.3666, over 17209.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1271, cr_loss=0.3455, over 3356536.95 frames. ], batch size: 55, lr: 3.39e-03, grad_scale: 32.0
2024-09-25 01:28:02,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=632543.3333333334, ans=0.2
2024-09-25 01:28:16,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=632590.0, ans=0.025
2024-09-25 01:28:23,385 INFO [train.py:1198] (1/4) Epoch 35, batch 3100, loss[loss=0.2074, ctc_loss=0.1326, cr_loss=0.3742, over 17235.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1272, cr_loss=0.346, over 3355530.01 frames. ], batch size: 50, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:28:55,678 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.62 vs. limit=15.0
2024-09-25 01:29:01,045 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.287e+02 1.342e+02 1.471e+02 1.803e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-25 01:29:06,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=632730.0, ans=0.0
2024-09-25 01:29:46,584 INFO [train.py:1198] (1/4) Epoch 35, batch 3150, loss[loss=0.2087, ctc_loss=0.1363, cr_loss=0.3622, over 17170.00 frames. ], tot_loss[loss=0.1963, ctc_loss=0.1271, cr_loss=0.3463, over 3356770.97 frames. ], batch size: 48, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:30:07,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=632916.6666666666, ans=0.0
2024-09-25 01:30:13,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=632916.6666666666, ans=0.1
2024-09-25 01:30:20,719 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.71 vs. limit=15.0
2024-09-25 01:30:27,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=632963.3333333334, ans=0.0
2024-09-25 01:30:46,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=633010.0, ans=0.1
2024-09-25 01:30:48,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=633056.6666666666, ans=0.025
2024-09-25 01:30:55,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=633056.6666666666, ans=0.5
2024-09-25 01:30:58,116 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.89 vs. limit=22.5
2024-09-25 01:31:04,923 INFO [train.py:1198] (1/4) Epoch 35, batch 3200, loss[loss=0.2074, ctc_loss=0.134, cr_loss=0.3665, over 17307.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1263, cr_loss=0.3448, over 3365204.15 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:31:28,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=633150.0, ans=0.125
2024-09-25 01:31:42,383 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.282e+02 1.354e+02 1.513e+02 2.603e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-25 01:31:52,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=633243.3333333334, ans=0.125
2024-09-25 01:32:09,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=633290.0, ans=0.1
2024-09-25 01:32:12,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=633290.0, ans=0.1
2024-09-25 01:32:14,591 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.22 vs. limit=10.0
2024-09-25 01:32:23,311 INFO [train.py:1198] (1/4) Epoch 35, batch 3250, loss[loss=0.2251, ctc_loss=0.1491, cr_loss=0.3801, over 15016.00 frames. ], tot_loss[loss=0.1948, ctc_loss=0.1261, cr_loss=0.3438, over 3364148.44 frames. ], batch size: 89, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:32:29,827 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:32:52,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=633430.0, ans=0.0
2024-09-25 01:32:59,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=633430.0, ans=0.125
2024-09-25 01:33:19,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=633476.6666666666, ans=0.0
2024-09-25 01:33:30,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=633523.3333333334, ans=0.0
2024-09-25 01:33:40,961 INFO [train.py:1198] (1/4) Epoch 35, batch 3300, loss[loss=0.1829, ctc_loss=0.1173, cr_loss=0.3281, over 17074.00 frames. ], tot_loss[loss=0.1949, ctc_loss=0.1261, cr_loss=0.3441, over 3362850.95 frames. ], batch size: 46, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:34:05,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=633616.6666666666, ans=0.125
2024-09-25 01:34:10,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=633616.6666666666, ans=0.1
2024-09-25 01:34:16,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.max_abs, batch_count=633663.3333333334, ans=10.0
2024-09-25 01:34:19,045 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.276e+02 1.335e+02 1.450e+02 3.347e+02, threshold=2.669e+02, percent-clipped=1.0
2024-09-25 01:34:39,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=633710.0, ans=0.125
2024-09-25 01:34:54,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=633756.6666666666, ans=0.0
2024-09-25 01:35:00,134 INFO [train.py:1198] (1/4) Epoch 35, batch 3350, loss[loss=0.1811, ctc_loss=0.1151, cr_loss=0.3301, over 17144.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.127, cr_loss=0.3459, over 3361496.83 frames. ], batch size: 48, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:35:17,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=633850.0, ans=0.1
2024-09-25 01:35:22,663 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.86 vs. limit=15.0
2024-09-25 01:35:24,401 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.74 vs. limit=10.0
2024-09-25 01:35:47,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=633943.3333333334, ans=0.125
2024-09-25 01:36:12,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=633990.0, ans=0.125
2024-09-25 01:36:17,056 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:36:18,433 INFO [train.py:1198] (1/4) Epoch 35, batch 3400, loss[loss=0.1718, ctc_loss=0.1095, cr_loss=0.3112, over 17187.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.126, cr_loss=0.3437, over 3371899.68 frames. ], batch size: 41, lr: 3.38e-03, grad_scale: 16.0
2024-09-25 01:36:22,294 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.99 vs. limit=12.0
2024-09-25 01:36:31,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=634036.6666666666, ans=0.0
2024-09-25 01:36:34,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=634083.3333333334, ans=0.2
2024-09-25 01:36:37,912 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.85 vs. limit=6.0
2024-09-25 01:36:57,297 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.280e+02 1.352e+02 1.458e+02 2.510e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-25 01:37:03,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=634176.6666666666, ans=0.0
2024-09-25 01:37:26,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=634223.3333333334, ans=0.125
2024-09-25 01:37:38,588 INFO [train.py:1198] (1/4) Epoch 35, batch 3450, loss[loss=0.1667, ctc_loss=0.1069, cr_loss=0.299, over 16983.00 frames. ], tot_loss[loss=0.1946, ctc_loss=0.126, cr_loss=0.3431, over 3369130.34 frames. ], batch size: 42, lr: 3.38e-03, grad_scale: 16.0
2024-09-25 01:38:38,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=634410.0, ans=0.125
2024-09-25 01:38:38,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=634410.0, ans=0.0
2024-09-25 01:38:58,784 INFO [train.py:1198] (1/4) Epoch 35, batch 3500, loss[loss=0.2556, ctc_loss=0.1724, cr_loss=0.4159, over 16429.00 frames. ], tot_loss[loss=0.1946, ctc_loss=0.1261, cr_loss=0.3428, over 3373378.64 frames. ], batch size: 66, lr: 3.38e-03, grad_scale: 16.0
2024-09-25 01:39:17,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=634550.0, ans=0.1
2024-09-25 01:39:24,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer_ff2.min_abs, batch_count=634550.0, ans=0.1
2024-09-25 01:39:29,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=634550.0, ans=0.1
2024-09-25 01:39:34,463 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.62 vs. limit=12.0
2024-09-25 01:39:37,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=634596.6666666666, ans=0.025
2024-09-25 01:39:40,109 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.254e+02 1.331e+02 1.459e+02 2.382e+02, threshold=2.663e+02, percent-clipped=0.0
2024-09-25 01:40:13,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=634690.0, ans=0.0
2024-09-25 01:40:19,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=634690.0, ans=0.125
2024-09-25 01:40:22,747 INFO [train.py:1198] (1/4) Epoch 35, batch 3550, loss[loss=0.2203, ctc_loss=0.1458, cr_loss=0.3726, over 16439.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1268, cr_loss=0.3432, over 3344941.14 frames. ], batch size: 66, lr: 3.38e-03, grad_scale: 16.0
2024-09-25 01:40:33,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=634736.6666666666, ans=0.125
2024-09-25 01:41:00,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=634830.0, ans=0.125
2024-09-25 01:41:18,998 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:41:40,836 INFO [train.py:1198] (1/4) Epoch 35, batch 3600, loss[loss=0.188, ctc_loss=0.1226, cr_loss=0.3268, over 16994.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1271, cr_loss=0.3444, over 3339684.07 frames. ], batch size: 56, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:41:59,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=635016.6666666666, ans=0.2
2024-09-25 01:42:11,513 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.37 vs. limit=15.0
2024-09-25 01:42:17,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=635063.3333333334, ans=0.125
2024-09-25 01:42:19,816 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.292e+02 1.369e+02 1.496e+02 2.107e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-25 01:42:43,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=635156.6666666666, ans=0.0
2024-09-25 01:42:49,567 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:42:58,846 INFO [train.py:1198] (1/4) Epoch 35, batch 3650, loss[loss=0.2025, ctc_loss=0.1318, cr_loss=0.3533, over 17300.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1266, cr_loss=0.3439, over 3346890.82 frames. ], batch size: 51, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:43:38,664 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.39 vs. limit=15.0
2024-09-25 01:43:41,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=635296.6666666666, ans=0.0
2024-09-25 01:43:52,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=635343.3333333334, ans=0.125
2024-09-25 01:43:55,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=635343.3333333334, ans=0.1
2024-09-25 01:44:08,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=635390.0, ans=0.125
2024-09-25 01:44:17,733 INFO [train.py:1198] (1/4) Epoch 35, batch 3700, loss[loss=0.2267, ctc_loss=0.1489, cr_loss=0.3893, over 17206.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1266, cr_loss=0.3443, over 3343806.21 frames. ], batch size: 50, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:44:27,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=635436.6666666666, ans=0.125
2024-09-25 01:44:46,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=635483.3333333334, ans=0.1
2024-09-25 01:44:56,874 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.290e+02 1.367e+02 1.473e+02 2.318e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-25 01:45:13,720 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.40 vs. limit=15.0
2024-09-25 01:45:33,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=635623.3333333334, ans=0.1
2024-09-25 01:45:36,670 INFO [train.py:1198] (1/4) Epoch 35, batch 3750, loss[loss=0.2153, ctc_loss=0.1414, cr_loss=0.3695, over 17012.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1264, cr_loss=0.3442, over 3346214.20 frames. ], batch size: 51, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:45:54,365 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:45:56,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.64 vs. limit=15.0
2024-09-25 01:46:03,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=635716.6666666666, ans=0.1
2024-09-25 01:46:27,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=635810.0, ans=0.125
2024-09-25 01:46:51,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=635856.6666666666, ans=0.1
2024-09-25 01:46:55,869 INFO [train.py:1198] (1/4) Epoch 35, batch 3800, loss[loss=0.2208, ctc_loss=0.1457, cr_loss=0.3755, over 15009.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1268, cr_loss=0.3446, over 3334824.82 frames. ], batch size: 89, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:47:07,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.max_abs, batch_count=635903.3333333334, ans=10.0
2024-09-25 01:47:16,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=635950.0, ans=0.125
2024-09-25 01:47:27,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=635996.6666666666, ans=0.125
2024-09-25 01:47:35,577 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.304e+02 1.388e+02 1.504e+02 2.196e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-25 01:48:16,221 INFO [train.py:1198] (1/4) Epoch 35, batch 3850, loss[loss=0.2535, ctc_loss=0.173, cr_loss=0.4025, over 11766.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1276, cr_loss=0.3443, over 3277207.08 frames. ], batch size: 123, lr: 3.38e-03, grad_scale: 32.0
2024-09-25 01:48:21,532 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.78 vs. limit=15.0
2024-09-25 01:48:46,765 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.70 vs. limit=22.5
2024-09-25 01:48:50,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=636230.0, ans=0.125
2024-09-25 01:49:16,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=636323.3333333334, ans=0.0
2024-09-25 01:49:22,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=636323.3333333334, ans=0.125
2024-09-25 01:50:19,159 INFO [train.py:1198] (1/4) Epoch 36, batch 0, loss[loss=0.2001, ctc_loss=0.1295, cr_loss=0.353, over 17017.00 frames. ], tot_loss[loss=0.2001, ctc_loss=0.1295, cr_loss=0.353, over 17017.00 frames. ], batch size: 51, lr: 3.33e-03, grad_scale: 32.0
2024-09-25 01:50:19,160 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 01:50:34,830 INFO [train.py:1230] (1/4) Epoch 36, validation: loss=0.0356, ctc_loss=0.0356, cr_loss=9.615e-15, over 944034.00 frames. 
2024-09-25 01:50:34,831 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 01:50:35,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=636351.3333333334, ans=0.0
2024-09-25 01:50:49,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=636398.0, ans=0.125
2024-09-25 01:50:54,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=636398.0, ans=0.2
2024-09-25 01:50:54,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=636398.0, ans=0.04949747468305833
2024-09-25 01:51:10,973 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.54 vs. limit=10.0
2024-09-25 01:51:21,245 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.459e+02 1.589e+02 1.794e+02 2.988e+02, threshold=3.179e+02, percent-clipped=1.0
2024-09-25 01:51:49,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=636538.0, ans=0.1
2024-09-25 01:51:55,583 INFO [train.py:1198] (1/4) Epoch 36, batch 50, loss[loss=0.2065, ctc_loss=0.1346, cr_loss=0.3594, over 17300.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1248, cr_loss=0.3408, over 756421.45 frames. ], batch size: 51, lr: 3.33e-03, grad_scale: 32.0
2024-09-25 01:51:57,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=636584.6666666666, ans=0.0
2024-09-25 01:52:08,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=636584.6666666666, ans=0.125
2024-09-25 01:52:08,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=636584.6666666666, ans=0.025
2024-09-25 01:52:11,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=636631.3333333334, ans=0.05
2024-09-25 01:52:26,401 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=7.63 vs. limit=15.0
2024-09-25 01:53:12,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=636771.3333333334, ans=0.0
2024-09-25 01:53:12,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=636771.3333333334, ans=0.125
2024-09-25 01:53:21,605 INFO [train.py:1198] (1/4) Epoch 36, batch 100, loss[loss=0.2086, ctc_loss=0.1351, cr_loss=0.3673, over 17304.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1273, cr_loss=0.3447, over 1333810.93 frames. ], batch size: 49, lr: 3.33e-03, grad_scale: 32.0
2024-09-25 01:53:44,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=636864.6666666666, ans=0.125
2024-09-25 01:53:46,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=636864.6666666666, ans=0.125
2024-09-25 01:53:49,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=636864.6666666666, ans=0.0
2024-09-25 01:54:11,061 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.276e+02 1.354e+02 1.441e+02 1.843e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-25 01:54:17,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=636958.0, ans=0.0
2024-09-25 01:54:34,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=637004.6666666666, ans=0.125
2024-09-25 01:54:47,635 INFO [train.py:1198] (1/4) Epoch 36, batch 150, loss[loss=0.2118, ctc_loss=0.1382, cr_loss=0.3682, over 17008.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.127, cr_loss=0.3451, over 1787905.14 frames. ], batch size: 53, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 01:54:49,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=637051.3333333334, ans=0.0
2024-09-25 01:55:10,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=637098.0, ans=0.125
2024-09-25 01:55:52,020 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:55:55,105 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 01:56:06,037 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=637284.6666666666, ans=0.125
2024-09-25 01:56:07,385 INFO [train.py:1198] (1/4) Epoch 36, batch 200, loss[loss=0.1993, ctc_loss=0.1306, cr_loss=0.3432, over 17063.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1265, cr_loss=0.3449, over 2131665.85 frames. ], batch size: 46, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 01:56:44,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=637378.0, ans=0.025
2024-09-25 01:56:44,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=637378.0, ans=0.125
2024-09-25 01:56:55,269 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.279e+02 1.373e+02 1.478e+02 2.081e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-25 01:57:08,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=637424.6666666666, ans=0.125
2024-09-25 01:57:13,668 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.96 vs. limit=12.0
2024-09-25 01:57:29,959 INFO [train.py:1198] (1/4) Epoch 36, batch 250, loss[loss=0.2349, ctc_loss=0.1596, cr_loss=0.3765, over 11748.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1262, cr_loss=0.3441, over 2403162.02 frames. ], batch size: 123, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 01:57:34,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=637518.0, ans=0.125
2024-09-25 01:57:49,957 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.98 vs. limit=15.0
2024-09-25 01:58:21,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=637658.0, ans=0.0
2024-09-25 01:58:29,442 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.28 vs. limit=15.0
2024-09-25 01:58:52,445 INFO [train.py:1198] (1/4) Epoch 36, batch 300, loss[loss=0.1626, ctc_loss=0.1038, cr_loss=0.2938, over 17087.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1253, cr_loss=0.3418, over 2617138.10 frames. ], batch size: 40, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 01:59:13,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=637798.0, ans=0.09899494936611666
2024-09-25 01:59:40,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=637844.6666666666, ans=0.1
2024-09-25 01:59:46,603 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.272e+02 1.363e+02 1.432e+02 1.912e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-25 01:59:50,437 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.20 vs. limit=15.0
2024-09-25 01:59:54,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=637891.3333333334, ans=0.125
2024-09-25 02:00:18,409 INFO [train.py:1198] (1/4) Epoch 36, batch 350, loss[loss=0.1625, ctc_loss=0.1038, cr_loss=0.2939, over 17132.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1256, cr_loss=0.3427, over 2775757.95 frames. ], batch size: 40, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 02:00:28,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=637984.6666666666, ans=0.025
2024-09-25 02:00:31,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=637984.6666666666, ans=0.0
2024-09-25 02:00:32,269 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.36 vs. limit=6.0
2024-09-25 02:00:43,489 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.93 vs. limit=22.5
2024-09-25 02:01:23,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=638171.3333333334, ans=0.125
2024-09-25 02:01:37,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=638218.0, ans=0.125
2024-09-25 02:01:38,834 INFO [train.py:1198] (1/4) Epoch 36, batch 400, loss[loss=0.2217, ctc_loss=0.1456, cr_loss=0.3805, over 16867.00 frames. ], tot_loss[loss=0.194, ctc_loss=0.1255, cr_loss=0.3429, over 2917424.84 frames. ], batch size: 58, lr: 3.32e-03, grad_scale: 32.0
2024-09-25 02:01:42,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=638218.0, ans=0.2
2024-09-25 02:02:01,495 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 02:02:28,328 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.05 vs. limit=15.0
2024-09-25 02:02:29,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.293e+02 1.367e+02 1.475e+02 2.656e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-25 02:02:34,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=638358.0, ans=0.5
2024-09-25 02:02:46,208 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.04 vs. limit=15.0
2024-09-25 02:02:52,052 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=638404.6666666666, ans=0.1
2024-09-25 02:03:01,277 INFO [train.py:1198] (1/4) Epoch 36, batch 450, loss[loss=0.215, ctc_loss=0.1406, cr_loss=0.3721, over 17258.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.1259, cr_loss=0.3438, over 3017175.70 frames. ], batch size: 44, lr: 3.32e-03, grad_scale: 32.0
2024-09-25 02:03:33,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=638498.0, ans=0.125
2024-09-25 02:03:33,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=638498.0, ans=0.0
2024-09-25 02:03:52,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=638591.3333333334, ans=0.125
2024-09-25 02:04:11,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=638638.0, ans=0.2
2024-09-25 02:04:27,072 INFO [train.py:1198] (1/4) Epoch 36, batch 500, loss[loss=0.1675, ctc_loss=0.1065, cr_loss=0.3051, over 17246.00 frames. ], tot_loss[loss=0.1948, ctc_loss=0.1261, cr_loss=0.3439, over 3090944.17 frames. ], batch size: 44, lr: 3.32e-03, grad_scale: 32.0
2024-09-25 02:04:41,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=638684.6666666666, ans=0.0
2024-09-25 02:05:19,275 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.244e+02 1.315e+02 1.459e+02 2.015e+02, threshold=2.630e+02, percent-clipped=0.0
2024-09-25 02:05:29,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=638824.6666666666, ans=0.025
2024-09-25 02:05:38,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=638871.3333333334, ans=0.0
2024-09-25 02:05:49,777 INFO [train.py:1198] (1/4) Epoch 36, batch 550, loss[loss=0.1714, ctc_loss=0.1099, cr_loss=0.3072, over 17249.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1248, cr_loss=0.342, over 3159411.28 frames. ], batch size: 44, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 02:05:59,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=638918.0, ans=0.1
2024-09-25 02:06:04,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=638964.6666666666, ans=0.0
2024-09-25 02:06:15,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=638964.6666666666, ans=0.1
2024-09-25 02:06:31,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=639011.3333333334, ans=0.2
2024-09-25 02:06:33,720 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.78 vs. limit=6.0
2024-09-25 02:07:01,850 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=639104.6666666666, ans=0.125
2024-09-25 02:07:09,442 INFO [train.py:1198] (1/4) Epoch 36, batch 600, loss[loss=0.1959, ctc_loss=0.1244, cr_loss=0.3577, over 17009.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1252, cr_loss=0.3427, over 3205546.85 frames. ], batch size: 56, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 02:07:19,397 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.42 vs. limit=6.0
2024-09-25 02:07:25,417 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.10 vs. limit=6.0
2024-09-25 02:08:01,748 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.285e+02 1.361e+02 1.464e+02 2.442e+02, threshold=2.722e+02, percent-clipped=0.0
2024-09-25 02:08:06,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=639291.3333333334, ans=0.2
2024-09-25 02:08:19,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=639338.0, ans=0.125
2024-09-25 02:08:24,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=639338.0, ans=0.0
2024-09-25 02:08:34,769 INFO [train.py:1198] (1/4) Epoch 36, batch 650, loss[loss=0.1953, ctc_loss=0.1239, cr_loss=0.3569, over 17015.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1254, cr_loss=0.3433, over 3240749.85 frames. ], batch size: 56, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 02:08:38,624 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.83 vs. limit=6.0
2024-09-25 02:08:58,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=639431.3333333334, ans=0.0
2024-09-25 02:09:04,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=639431.3333333334, ans=0.0
2024-09-25 02:09:07,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=639478.0, ans=0.125
2024-09-25 02:09:09,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=639478.0, ans=0.125
2024-09-25 02:09:59,836 INFO [train.py:1198] (1/4) Epoch 36, batch 700, loss[loss=0.208, ctc_loss=0.1396, cr_loss=0.3424, over 16882.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.125, cr_loss=0.3423, over 3270723.86 frames. ], batch size: 58, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 02:10:04,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=639618.0, ans=0.0
2024-09-25 02:10:21,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=639664.6666666666, ans=0.0
2024-09-25 02:10:29,553 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.34 vs. limit=15.0
2024-09-25 02:10:40,641 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.77 vs. limit=15.0
2024-09-25 02:10:49,613 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.257e+02 1.341e+02 1.454e+02 1.758e+02, threshold=2.682e+02, percent-clipped=0.0
2024-09-25 02:11:13,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=639804.6666666666, ans=0.04949747468305833
2024-09-25 02:11:19,965 INFO [train.py:1198] (1/4) Epoch 36, batch 750, loss[loss=0.1767, ctc_loss=0.1151, cr_loss=0.308, over 17299.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1249, cr_loss=0.3416, over 3291067.76 frames. ], batch size: 51, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 02:11:46,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=639898.0, ans=0.0
2024-09-25 02:12:17,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=639991.3333333334, ans=0.1
2024-09-25 02:12:42,670 INFO [train.py:1198] (1/4) Epoch 36, batch 800, loss[loss=0.1835, ctc_loss=0.1188, cr_loss=0.3234, over 17261.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1249, cr_loss=0.3418, over 3305642.11 frames. ], batch size: 42, lr: 3.32e-03, grad_scale: 32.0
2024-09-25 02:12:52,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=640084.6666666666, ans=0.2
2024-09-25 02:13:22,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=640178.0, ans=0.125
2024-09-25 02:13:22,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=640178.0, ans=0.125
2024-09-25 02:13:24,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=640178.0, ans=0.0
2024-09-25 02:13:35,135 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.278e+02 1.351e+02 1.457e+02 1.942e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-25 02:13:37,983 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.62 vs. limit=15.0
2024-09-25 02:14:00,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=640271.3333333334, ans=0.0
2024-09-25 02:14:08,465 INFO [train.py:1198] (1/4) Epoch 36, batch 850, loss[loss=0.192, ctc_loss=0.1241, cr_loss=0.3398, over 17283.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1251, cr_loss=0.3414, over 3325764.21 frames. ], batch size: 49, lr: 3.32e-03, grad_scale: 32.0
2024-09-25 02:14:09,544 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.89 vs. limit=15.0
2024-09-25 02:14:23,504 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.55 vs. limit=12.0
2024-09-25 02:14:36,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=640364.6666666666, ans=0.125
2024-09-25 02:14:39,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=640364.6666666666, ans=0.125
2024-09-25 02:14:43,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=640411.3333333334, ans=0.0
2024-09-25 02:14:43,229 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 02:15:12,432 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.76 vs. limit=22.5
2024-09-25 02:15:26,531 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.43 vs. limit=15.0
2024-09-25 02:15:27,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=640504.6666666666, ans=0.125
2024-09-25 02:15:30,656 INFO [train.py:1198] (1/4) Epoch 36, batch 900, loss[loss=0.1728, ctc_loss=0.1095, cr_loss=0.3165, over 17286.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1256, cr_loss=0.3423, over 3332424.40 frames. ], batch size: 46, lr: 3.32e-03, grad_scale: 32.0
2024-09-25 02:16:06,291 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=640644.6666666666, ans=0.0
2024-09-25 02:16:22,098 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.267e+02 1.323e+02 1.416e+02 1.789e+02, threshold=2.647e+02, percent-clipped=0.0
2024-09-25 02:16:23,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=640691.3333333334, ans=0.0
2024-09-25 02:16:36,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=640738.0, ans=10.0
2024-09-25 02:16:43,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=640738.0, ans=0.0
2024-09-25 02:16:51,066 INFO [train.py:1198] (1/4) Epoch 36, batch 950, loss[loss=0.2142, ctc_loss=0.1414, cr_loss=0.3643, over 16897.00 frames. ], tot_loss[loss=0.1945, ctc_loss=0.126, cr_loss=0.3429, over 3340364.32 frames. ], batch size: 58, lr: 3.32e-03, grad_scale: 16.0
2024-09-25 02:16:56,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=640784.6666666666, ans=0.0
2024-09-25 02:17:06,672 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.21 vs. limit=15.0
2024-09-25 02:17:22,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.max_abs, batch_count=640831.3333333334, ans=10.0
2024-09-25 02:17:32,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten.whitening_limit, batch_count=640878.0, ans=22.5
2024-09-25 02:17:38,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=640878.0, ans=0.125
2024-09-25 02:17:41,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=640924.6666666666, ans=0.2
2024-09-25 02:18:16,109 INFO [train.py:1198] (1/4) Epoch 36, batch 1000, loss[loss=0.219, ctc_loss=0.1424, cr_loss=0.3833, over 17049.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1264, cr_loss=0.3444, over 3348442.83 frames. ], batch size: 52, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:18:24,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=641018.0, ans=0.1
2024-09-25 02:18:42,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=641064.6666666666, ans=0.1
2024-09-25 02:18:49,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=641111.3333333334, ans=0.0
2024-09-25 02:19:10,045 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.268e+02 1.360e+02 1.441e+02 1.989e+02, threshold=2.720e+02, percent-clipped=0.0
2024-09-25 02:19:14,175 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.37 vs. limit=22.5
2024-09-25 02:19:27,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=641204.6666666666, ans=0.0
2024-09-25 02:19:29,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=641204.6666666666, ans=0.125
2024-09-25 02:19:41,706 INFO [train.py:1198] (1/4) Epoch 36, batch 1050, loss[loss=0.191, ctc_loss=0.1245, cr_loss=0.3328, over 17093.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1265, cr_loss=0.3448, over 3341831.35 frames. ], batch size: 49, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:20:01,400 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.04 vs. limit=10.0
2024-09-25 02:20:04,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=641298.0, ans=0.0
2024-09-25 02:20:12,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=641344.6666666666, ans=0.2
2024-09-25 02:20:17,941 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.33 vs. limit=15.0
2024-09-25 02:20:18,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=641344.6666666666, ans=0.0
2024-09-25 02:20:36,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=641391.3333333334, ans=0.2
2024-09-25 02:21:01,838 INFO [train.py:1198] (1/4) Epoch 36, batch 1100, loss[loss=0.201, ctc_loss=0.1286, cr_loss=0.3617, over 17326.00 frames. ], tot_loss[loss=0.1952, ctc_loss=0.1264, cr_loss=0.3438, over 3327206.87 frames. ], batch size: 49, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:21:52,920 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.290e+02 1.394e+02 1.498e+02 2.022e+02, threshold=2.788e+02, percent-clipped=0.0
2024-09-25 02:21:53,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=641624.6666666666, ans=0.0
2024-09-25 02:22:15,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=641671.3333333334, ans=0.125
2024-09-25 02:22:15,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=641671.3333333334, ans=0.125
2024-09-25 02:22:24,255 INFO [train.py:1198] (1/4) Epoch 36, batch 1150, loss[loss=0.1863, ctc_loss=0.1195, cr_loss=0.3341, over 17167.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1257, cr_loss=0.3425, over 3338815.75 frames. ], batch size: 45, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:22:43,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=641764.6666666666, ans=0.1
2024-09-25 02:22:53,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=641764.6666666666, ans=0.125
2024-09-25 02:22:53,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=641764.6666666666, ans=0.025
2024-09-25 02:23:23,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=641858.0, ans=0.125
2024-09-25 02:23:30,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=641904.6666666666, ans=0.025
2024-09-25 02:23:33,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=641904.6666666666, ans=0.0
2024-09-25 02:23:33,895 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=8.41 vs. limit=15.0
2024-09-25 02:23:48,969 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.80 vs. limit=10.0
2024-09-25 02:23:49,551 INFO [train.py:1198] (1/4) Epoch 36, batch 1200, loss[loss=0.1822, ctc_loss=0.1174, cr_loss=0.3237, over 17104.00 frames. ], tot_loss[loss=0.1939, ctc_loss=0.1255, cr_loss=0.342, over 3335887.92 frames. ], batch size: 40, lr: 3.31e-03, grad_scale: 32.0
2024-09-25 02:24:22,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=642044.6666666666, ans=0.125
2024-09-25 02:24:25,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=642044.6666666666, ans=0.125
2024-09-25 02:24:26,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=642044.6666666666, ans=0.0
2024-09-25 02:24:43,285 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.264e+02 1.352e+02 1.439e+02 3.839e+02, threshold=2.704e+02, percent-clipped=1.0
2024-09-25 02:25:12,608 INFO [train.py:1198] (1/4) Epoch 36, batch 1250, loss[loss=0.2084, ctc_loss=0.1344, cr_loss=0.3698, over 17350.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1246, cr_loss=0.3409, over 3343874.69 frames. ], batch size: 48, lr: 3.31e-03, grad_scale: 32.0
2024-09-25 02:25:19,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=642184.6666666666, ans=0.125
2024-09-25 02:25:19,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=642184.6666666666, ans=0.125
2024-09-25 02:25:35,423 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 02:25:46,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=642278.0, ans=0.125
2024-09-25 02:25:55,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=642278.0, ans=0.125
2024-09-25 02:26:02,354 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.33 vs. limit=15.0
2024-09-25 02:26:06,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=642324.6666666666, ans=0.0
2024-09-25 02:26:07,265 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.48 vs. limit=15.0
2024-09-25 02:26:32,184 INFO [train.py:1198] (1/4) Epoch 36, batch 1300, loss[loss=0.1971, ctc_loss=0.1266, cr_loss=0.3526, over 17160.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1253, cr_loss=0.3422, over 3345118.59 frames. ], batch size: 45, lr: 3.31e-03, grad_scale: 32.0
2024-09-25 02:26:48,652 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 02:26:59,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=642464.6666666666, ans=0.125
2024-09-25 02:27:25,952 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.281e+02 1.359e+02 1.510e+02 2.229e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 02:27:30,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=642558.0, ans=0.125
2024-09-25 02:27:37,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=642604.6666666666, ans=0.0
2024-09-25 02:27:49,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=642604.6666666666, ans=0.1
2024-09-25 02:27:57,336 INFO [train.py:1198] (1/4) Epoch 36, batch 1350, loss[loss=0.1907, ctc_loss=0.121, cr_loss=0.3487, over 17210.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.1251, cr_loss=0.3417, over 3358434.18 frames. ], batch size: 47, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:28:07,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=642651.3333333334, ans=0.125
2024-09-25 02:28:07,884 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.38 vs. limit=15.0
2024-09-25 02:28:16,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=642698.0, ans=0.125
2024-09-25 02:28:50,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=642791.3333333334, ans=0.1
2024-09-25 02:29:05,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=642838.0, ans=0.2
2024-09-25 02:29:21,997 INFO [train.py:1198] (1/4) Epoch 36, batch 1400, loss[loss=0.1798, ctc_loss=0.1144, cr_loss=0.327, over 17303.00 frames. ], tot_loss[loss=0.1946, ctc_loss=0.126, cr_loss=0.343, over 3361163.80 frames. ], batch size: 49, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:29:31,892 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 02:29:39,144 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.02 vs. limit=12.0
2024-09-25 02:30:00,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=642978.0, ans=0.125
2024-09-25 02:30:04,610 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.05 vs. limit=12.0
2024-09-25 02:30:10,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=643024.6666666666, ans=0.125
2024-09-25 02:30:14,813 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.323e+02 1.435e+02 1.580e+02 6.083e+02, threshold=2.870e+02, percent-clipped=1.0
2024-09-25 02:30:25,338 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.07 vs. limit=22.5
2024-09-25 02:30:41,857 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.14 vs. limit=22.5
2024-09-25 02:30:42,402 INFO [train.py:1198] (1/4) Epoch 36, batch 1450, loss[loss=0.2144, ctc_loss=0.1405, cr_loss=0.3698, over 17214.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1266, cr_loss=0.3441, over 3352509.20 frames. ], batch size: 55, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:31:11,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=643164.6666666666, ans=0.0
2024-09-25 02:31:13,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=643211.3333333334, ans=0.125
2024-09-25 02:31:19,775 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.25 vs. limit=10.0
2024-09-25 02:31:24,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=643211.3333333334, ans=0.125
2024-09-25 02:31:33,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=643258.0, ans=0.125
2024-09-25 02:31:55,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=643304.6666666666, ans=0.025
2024-09-25 02:32:04,901 INFO [train.py:1198] (1/4) Epoch 36, batch 1500, loss[loss=0.1608, ctc_loss=0.103, cr_loss=0.2888, over 17010.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.126, cr_loss=0.3434, over 3354029.80 frames. ], batch size: 44, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:32:10,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=643351.3333333334, ans=0.2
2024-09-25 02:32:30,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=643398.0, ans=0.1
2024-09-25 02:32:53,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=643444.6666666666, ans=0.1
2024-09-25 02:33:00,745 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.291e+02 1.396e+02 1.518e+02 2.099e+02, threshold=2.792e+02, percent-clipped=0.0
2024-09-25 02:33:05,025 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=10.78 vs. limit=15.0
2024-09-25 02:33:05,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=643491.3333333334, ans=0.125
2024-09-25 02:33:25,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=643538.0, ans=0.125
2024-09-25 02:33:30,539 INFO [train.py:1198] (1/4) Epoch 36, batch 1550, loss[loss=0.2062, ctc_loss=0.1346, cr_loss=0.3581, over 17014.00 frames. ], tot_loss[loss=0.1944, ctc_loss=0.1258, cr_loss=0.3432, over 3357249.33 frames. ], batch size: 52, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:33:37,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=643584.6666666666, ans=0.125
2024-09-25 02:33:38,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=643584.6666666666, ans=0.0
2024-09-25 02:33:52,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=8.06 vs. limit=15.0
2024-09-25 02:34:10,954 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=15.71 vs. limit=22.5
2024-09-25 02:34:18,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=643678.0, ans=0.1
2024-09-25 02:34:20,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=643724.6666666666, ans=0.09899494936611666
2024-09-25 02:34:35,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=643771.3333333334, ans=0.0
2024-09-25 02:34:53,198 INFO [train.py:1198] (1/4) Epoch 36, batch 1600, loss[loss=0.1999, ctc_loss=0.1294, cr_loss=0.3527, over 17214.00 frames. ], tot_loss[loss=0.1945, ctc_loss=0.1259, cr_loss=0.3432, over 3350738.71 frames. ], batch size: 47, lr: 3.31e-03, grad_scale: 32.0
2024-09-25 02:34:53,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=643818.0, ans=0.125
2024-09-25 02:35:04,222 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=10.03 vs. limit=22.5
2024-09-25 02:35:17,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.91 vs. limit=10.0
2024-09-25 02:35:19,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=643864.6666666666, ans=0.0
2024-09-25 02:35:28,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=643911.3333333334, ans=0.125
2024-09-25 02:35:46,524 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.268e+02 1.352e+02 1.455e+02 2.374e+02, threshold=2.704e+02, percent-clipped=0.0
2024-09-25 02:35:53,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=643958.0, ans=0.125
2024-09-25 02:36:01,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=644004.6666666666, ans=0.0
2024-09-25 02:36:14,435 INFO [train.py:1198] (1/4) Epoch 36, batch 1650, loss[loss=0.1793, ctc_loss=0.1146, cr_loss=0.324, over 17235.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1268, cr_loss=0.345, over 3349059.66 frames. ], batch size: 47, lr: 3.31e-03, grad_scale: 32.0
2024-09-25 02:36:35,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=644098.0, ans=0.0
2024-09-25 02:36:48,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=644144.6666666666, ans=0.015
2024-09-25 02:36:51,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=644144.6666666666, ans=0.04949747468305833
2024-09-25 02:37:22,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=644238.0, ans=0.125
2024-09-25 02:37:37,127 INFO [train.py:1198] (1/4) Epoch 36, batch 1700, loss[loss=0.1843, ctc_loss=0.1176, cr_loss=0.3338, over 17213.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1278, cr_loss=0.3467, over 3347327.67 frames. ], batch size: 47, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:37:55,168 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=7.51 vs. limit=15.0
2024-09-25 02:37:56,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=644331.3333333334, ans=0.04949747468305833
2024-09-25 02:38:07,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=644331.3333333334, ans=0.0
2024-09-25 02:38:25,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=644378.0, ans=0.2
2024-09-25 02:38:37,024 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.256e+02 1.337e+02 1.426e+02 1.735e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-25 02:38:37,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=644424.6666666666, ans=0.09899494936611666
2024-09-25 02:38:43,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=644424.6666666666, ans=0.2
2024-09-25 02:39:02,768 INFO [train.py:1198] (1/4) Epoch 36, batch 1750, loss[loss=0.2301, ctc_loss=0.1593, cr_loss=0.3541, over 12154.00 frames. ], tot_loss[loss=0.1972, ctc_loss=0.1279, cr_loss=0.3463, over 3348501.75 frames. ], batch size: 124, lr: 3.31e-03, grad_scale: 16.0
2024-09-25 02:39:26,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=644564.6666666666, ans=0.1
2024-09-25 02:40:00,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=644658.0, ans=0.025
2024-09-25 02:40:14,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=644704.6666666666, ans=0.07
2024-09-25 02:40:17,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=644704.6666666666, ans=0.0
2024-09-25 02:40:17,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=644704.6666666666, ans=0.1
2024-09-25 02:40:21,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys.whitening_limit, batch_count=644704.6666666666, ans=6.0
2024-09-25 02:40:22,850 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.54 vs. limit=15.0
2024-09-25 02:40:25,299 INFO [train.py:1198] (1/4) Epoch 36, batch 1800, loss[loss=0.2204, ctc_loss=0.1509, cr_loss=0.3473, over 11769.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1274, cr_loss=0.3454, over 3349426.32 frames. ], batch size: 123, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 02:40:33,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=644751.3333333334, ans=10.0
2024-09-25 02:40:47,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=644798.0, ans=0.0
2024-09-25 02:40:49,350 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=644798.0, ans=0.125
2024-09-25 02:40:59,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=644844.6666666666, ans=0.125
2024-09-25 02:41:00,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=644844.6666666666, ans=0.125
2024-09-25 02:41:19,904 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.269e+02 1.345e+02 1.435e+02 2.144e+02, threshold=2.691e+02, percent-clipped=0.0
2024-09-25 02:41:21,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=644891.3333333334, ans=0.125
2024-09-25 02:41:45,750 INFO [train.py:1198] (1/4) Epoch 36, batch 1850, loss[loss=0.1994, ctc_loss=0.1274, cr_loss=0.3601, over 17270.00 frames. ], tot_loss[loss=0.1956, ctc_loss=0.1267, cr_loss=0.3445, over 3359723.20 frames. ], batch size: 44, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 02:41:51,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=644984.6666666666, ans=0.125
2024-09-25 02:42:09,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=645031.3333333334, ans=0.2
2024-09-25 02:42:16,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=645031.3333333334, ans=0.2
2024-09-25 02:42:21,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=645078.0, ans=0.09899494936611666
2024-09-25 02:42:21,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=645078.0, ans=0.0
2024-09-25 02:42:54,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=645171.3333333334, ans=0.5
2024-09-25 02:42:59,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=645171.3333333334, ans=0.125
2024-09-25 02:43:08,285 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.55 vs. limit=22.5
2024-09-25 02:43:10,545 INFO [train.py:1198] (1/4) Epoch 36, batch 1900, loss[loss=0.1888, ctc_loss=0.1215, cr_loss=0.3366, over 16901.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1267, cr_loss=0.3446, over 3359413.49 frames. ], batch size: 58, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 02:43:15,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=645218.0, ans=0.125
2024-09-25 02:43:39,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=645264.6666666666, ans=0.0
2024-09-25 02:44:10,239 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.279e+02 1.358e+02 1.455e+02 1.986e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-25 02:44:36,299 INFO [train.py:1198] (1/4) Epoch 36, batch 1950, loss[loss=0.2292, ctc_loss=0.1517, cr_loss=0.3876, over 17007.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1257, cr_loss=0.343, over 3357171.30 frames. ], batch size: 56, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 02:44:47,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=645451.3333333334, ans=0.125
2024-09-25 02:44:58,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=645498.0, ans=0.07
2024-09-25 02:45:06,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=645544.6666666666, ans=0.04949747468305833
2024-09-25 02:45:20,094 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.76 vs. limit=6.0
2024-09-25 02:45:24,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten.whitening_limit, batch_count=645591.3333333334, ans=15.0
2024-09-25 02:45:31,107 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.70 vs. limit=12.0
2024-09-25 02:45:47,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.06 vs. limit=12.0
2024-09-25 02:45:53,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=645638.0, ans=0.125
2024-09-25 02:45:56,173 INFO [train.py:1198] (1/4) Epoch 36, batch 2000, loss[loss=0.1866, ctc_loss=0.121, cr_loss=0.328, over 17288.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1252, cr_loss=0.3422, over 3356979.30 frames. ], batch size: 46, lr: 3.30e-03, grad_scale: 32.0
2024-09-25 02:46:02,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=645684.6666666666, ans=0.0
2024-09-25 02:46:18,874 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 02:46:24,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=645731.3333333334, ans=0.1
2024-09-25 02:46:34,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=645778.0, ans=0.125
2024-09-25 02:46:46,191 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.10 vs. limit=15.0
2024-09-25 02:46:50,156 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.284e+02 1.382e+02 1.479e+02 3.122e+02, threshold=2.765e+02, percent-clipped=1.0
2024-09-25 02:46:59,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=645824.6666666666, ans=0.0
2024-09-25 02:47:06,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=645871.3333333334, ans=0.0
2024-09-25 02:47:18,702 INFO [train.py:1198] (1/4) Epoch 36, batch 2050, loss[loss=0.2265, ctc_loss=0.1561, cr_loss=0.3518, over 11829.00 frames. ], tot_loss[loss=0.1939, ctc_loss=0.1255, cr_loss=0.3423, over 3358807.79 frames. ], batch size: 123, lr: 3.30e-03, grad_scale: 32.0
2024-09-25 02:47:26,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=645918.0, ans=0.125
2024-09-25 02:48:26,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=646104.6666666666, ans=0.0
2024-09-25 02:48:43,712 INFO [train.py:1198] (1/4) Epoch 36, batch 2100, loss[loss=0.1866, ctc_loss=0.1222, cr_loss=0.322, over 17073.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1254, cr_loss=0.3419, over 3362699.57 frames. ], batch size: 43, lr: 3.30e-03, grad_scale: 32.0
2024-09-25 02:49:12,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=646198.0, ans=0.1
2024-09-25 02:49:28,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=646244.6666666666, ans=0.07
2024-09-25 02:49:40,606 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.268e+02 1.352e+02 1.474e+02 3.330e+02, threshold=2.705e+02, percent-clipped=1.0
2024-09-25 02:49:54,635 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.11 vs. limit=15.0
2024-09-25 02:50:06,069 INFO [train.py:1198] (1/4) Epoch 36, batch 2150, loss[loss=0.2054, ctc_loss=0.1336, cr_loss=0.3592, over 15952.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1253, cr_loss=0.3414, over 3357985.83 frames. ], batch size: 74, lr: 3.30e-03, grad_scale: 32.0
2024-09-25 02:50:14,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=646384.6666666666, ans=0.125
2024-09-25 02:50:23,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=646431.3333333334, ans=0.2
2024-09-25 02:50:50,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=646478.0, ans=0.04949747468305833
2024-09-25 02:50:53,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=646524.6666666666, ans=0.0
2024-09-25 02:50:54,849 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.79 vs. limit=22.5
2024-09-25 02:50:55,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=646524.6666666666, ans=0.0
2024-09-25 02:51:10,952 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.29 vs. limit=15.0
2024-09-25 02:51:13,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=646571.3333333334, ans=0.1
2024-09-25 02:51:25,842 INFO [train.py:1198] (1/4) Epoch 36, batch 2200, loss[loss=0.2177, ctc_loss=0.1432, cr_loss=0.3725, over 17030.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1254, cr_loss=0.3414, over 3361955.25 frames. ], batch size: 52, lr: 3.30e-03, grad_scale: 32.0
2024-09-25 02:51:51,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=646664.6666666666, ans=0.125
2024-09-25 02:52:08,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=646711.3333333334, ans=0.125
2024-09-25 02:52:21,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=646758.0, ans=0.2
2024-09-25 02:52:22,864 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.282e+02 1.343e+02 1.455e+02 1.826e+02, threshold=2.686e+02, percent-clipped=0.0
2024-09-25 02:52:31,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=646758.0, ans=0.0
2024-09-25 02:52:49,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=646851.3333333334, ans=0.125
2024-09-25 02:52:51,064 INFO [train.py:1198] (1/4) Epoch 36, batch 2250, loss[loss=0.2223, ctc_loss=0.1453, cr_loss=0.3846, over 16569.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1257, cr_loss=0.3426, over 3364915.65 frames. ], batch size: 66, lr: 3.30e-03, grad_scale: 32.0
2024-09-25 02:53:07,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=646898.0, ans=0.1
2024-09-25 02:53:18,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=646898.0, ans=0.025
2024-09-25 02:53:24,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=646944.6666666666, ans=0.0
2024-09-25 02:53:24,886 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.74 vs. limit=22.5
2024-09-25 02:53:27,884 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.85 vs. limit=10.0
2024-09-25 02:53:37,414 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.16 vs. limit=15.0
2024-09-25 02:53:38,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=646944.6666666666, ans=0.125
2024-09-25 02:53:41,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=646991.3333333334, ans=0.125
2024-09-25 02:53:45,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=646991.3333333334, ans=0.0
2024-09-25 02:53:58,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=647038.0, ans=0.125
2024-09-25 02:54:00,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=647038.0, ans=0.0
2024-09-25 02:54:02,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=647038.0, ans=0.2
2024-09-25 02:54:16,047 INFO [train.py:1198] (1/4) Epoch 36, batch 2300, loss[loss=0.1662, ctc_loss=0.1074, cr_loss=0.2943, over 15840.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.3441, over 3363751.11 frames. ], batch size: 35, lr: 3.30e-03, grad_scale: 8.0
2024-09-25 02:54:47,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=647178.0, ans=0.125
2024-09-25 02:54:48,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=647178.0, ans=0.125
2024-09-25 02:55:03,494 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.09 vs. limit=15.0
2024-09-25 02:55:14,011 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.282e+02 1.356e+02 1.478e+02 2.427e+02, threshold=2.712e+02, percent-clipped=0.0
2024-09-25 02:55:36,334 INFO [train.py:1198] (1/4) Epoch 36, batch 2350, loss[loss=0.1701, ctc_loss=0.1091, cr_loss=0.3048, over 17310.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1267, cr_loss=0.3444, over 3366008.53 frames. ], batch size: 42, lr: 3.30e-03, grad_scale: 8.0
2024-09-25 02:55:45,223 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=10.85 vs. limit=15.0
2024-09-25 02:56:08,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=647411.3333333334, ans=0.0
2024-09-25 02:56:58,845 INFO [train.py:1198] (1/4) Epoch 36, batch 2400, loss[loss=0.2409, ctc_loss=0.1597, cr_loss=0.4058, over 15070.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.127, cr_loss=0.3444, over 3357006.98 frames. ], batch size: 89, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 02:57:13,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=647598.0, ans=0.0
2024-09-25 02:57:22,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=647598.0, ans=0.0
2024-09-25 02:57:25,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=647598.0, ans=0.125
2024-09-25 02:57:27,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=647598.0, ans=0.125
2024-09-25 02:57:36,850 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=647644.6666666666, ans=0.125
2024-09-25 02:57:38,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=647644.6666666666, ans=0.0
2024-09-25 02:57:52,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=647691.3333333334, ans=0.0
2024-09-25 02:57:55,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=647691.3333333334, ans=0.015
2024-09-25 02:57:58,955 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.280e+02 1.384e+02 1.461e+02 3.295e+02, threshold=2.768e+02, percent-clipped=1.0
2024-09-25 02:58:02,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=647691.3333333334, ans=0.0
2024-09-25 02:58:24,117 INFO [train.py:1198] (1/4) Epoch 36, batch 2450, loss[loss=0.2042, ctc_loss=0.1326, cr_loss=0.358, over 17298.00 frames. ], tot_loss[loss=0.1951, ctc_loss=0.1264, cr_loss=0.3436, over 3358293.82 frames. ], batch size: 46, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 02:58:30,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=647784.6666666666, ans=0.125
2024-09-25 02:58:55,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=647831.3333333334, ans=0.125
2024-09-25 02:59:12,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=647878.0, ans=0.0
2024-09-25 02:59:43,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=647971.3333333334, ans=0.2
2024-09-25 02:59:46,719 INFO [train.py:1198] (1/4) Epoch 36, batch 2500, loss[loss=0.2212, ctc_loss=0.1466, cr_loss=0.3734, over 15093.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1258, cr_loss=0.3421, over 3364402.13 frames. ], batch size: 89, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 02:59:55,376 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.57 vs. limit=12.0
2024-09-25 03:00:19,654 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.56 vs. limit=12.0
2024-09-25 03:00:44,516 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.286e+02 1.353e+02 1.494e+02 3.015e+02, threshold=2.705e+02, percent-clipped=1.0
2024-09-25 03:00:51,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=648204.6666666666, ans=0.2
2024-09-25 03:01:06,741 INFO [train.py:1198] (1/4) Epoch 36, batch 2550, loss[loss=0.1552, ctc_loss=0.09939, cr_loss=0.279, over 17033.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1257, cr_loss=0.3419, over 3370961.00 frames. ], batch size: 39, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 03:01:13,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=648251.3333333334, ans=0.2
2024-09-25 03:01:23,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=648298.0, ans=0.1
2024-09-25 03:01:26,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=648298.0, ans=0.125
2024-09-25 03:02:06,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=648391.3333333334, ans=0.05
2024-09-25 03:02:31,701 INFO [train.py:1198] (1/4) Epoch 36, batch 2600, loss[loss=0.1743, ctc_loss=0.1107, cr_loss=0.3178, over 17078.00 frames. ], tot_loss[loss=0.1929, ctc_loss=0.1248, cr_loss=0.3405, over 3377299.93 frames. ], batch size: 43, lr: 3.30e-03, grad_scale: 16.0
2024-09-25 03:02:35,680 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.19 vs. limit=15.0
2024-09-25 03:02:47,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=648531.3333333334, ans=0.125
2024-09-25 03:02:58,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=648531.3333333334, ans=0.09899494936611666
2024-09-25 03:03:31,214 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.288e+02 1.366e+02 1.471e+02 2.094e+02, threshold=2.731e+02, percent-clipped=0.0
2024-09-25 03:03:56,675 INFO [train.py:1198] (1/4) Epoch 36, batch 2650, loss[loss=0.2173, ctc_loss=0.1404, cr_loss=0.3843, over 16510.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1247, cr_loss=0.3399, over 3368526.69 frames. ], batch size: 66, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:03:58,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=648718.0, ans=0.0
2024-09-25 03:04:23,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=648764.6666666666, ans=0.125
2024-09-25 03:05:16,416 INFO [train.py:1198] (1/4) Epoch 36, batch 2700, loss[loss=0.1793, ctc_loss=0.1139, cr_loss=0.327, over 17308.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.125, cr_loss=0.3414, over 3370980.93 frames. ], batch size: 46, lr: 3.29e-03, grad_scale: 8.0
2024-09-25 03:05:29,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=648951.3333333334, ans=0.0
2024-09-25 03:05:46,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=649044.6666666666, ans=0.0
2024-09-25 03:05:51,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=649044.6666666666, ans=0.2
2024-09-25 03:05:54,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=649044.6666666666, ans=0.025
2024-09-25 03:06:04,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=649091.3333333334, ans=0.0
2024-09-25 03:06:12,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=649091.3333333334, ans=0.0
2024-09-25 03:06:15,209 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.266e+02 1.332e+02 1.436e+02 2.982e+02, threshold=2.664e+02, percent-clipped=1.0
2024-09-25 03:06:35,962 INFO [train.py:1198] (1/4) Epoch 36, batch 2750, loss[loss=0.1686, ctc_loss=0.1067, cr_loss=0.3096, over 17249.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1246, cr_loss=0.3408, over 3366059.07 frames. ], batch size: 44, lr: 3.29e-03, grad_scale: 8.0
2024-09-25 03:06:36,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.89 vs. limit=10.0
2024-09-25 03:07:04,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=649231.3333333334, ans=0.125
2024-09-25 03:07:42,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=649324.6666666666, ans=0.125
2024-09-25 03:07:53,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=649371.3333333334, ans=0.125
2024-09-25 03:08:01,227 INFO [train.py:1198] (1/4) Epoch 36, batch 2800, loss[loss=0.1931, ctc_loss=0.1233, cr_loss=0.349, over 17310.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1241, cr_loss=0.3394, over 3364566.67 frames. ], batch size: 51, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:08:20,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=649464.6666666666, ans=0.0
2024-09-25 03:08:39,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=649511.3333333334, ans=0.1
2024-09-25 03:08:41,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=649511.3333333334, ans=0.1
2024-09-25 03:08:48,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=649511.3333333334, ans=0.125
2024-09-25 03:09:05,753 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.257e+02 1.332e+02 1.427e+02 2.151e+02, threshold=2.663e+02, percent-clipped=0.0
2024-09-25 03:09:07,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=649558.0, ans=0.04949747468305833
2024-09-25 03:09:21,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=649604.6666666666, ans=0.0
2024-09-25 03:09:27,092 INFO [train.py:1198] (1/4) Epoch 36, batch 2850, loss[loss=0.2112, ctc_loss=0.1383, cr_loss=0.3647, over 17287.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.125, cr_loss=0.3406, over 3353199.57 frames. ], batch size: 46, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:09:38,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=649651.3333333334, ans=0.125
2024-09-25 03:10:04,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=649744.6666666666, ans=0.0
2024-09-25 03:10:21,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=649791.3333333334, ans=0.05
2024-09-25 03:10:31,386 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.10 vs. limit=15.0
2024-09-25 03:10:32,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=649838.0, ans=0.2
2024-09-25 03:10:38,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=649838.0, ans=0.0
2024-09-25 03:10:46,266 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.47 vs. limit=10.0
2024-09-25 03:10:46,601 INFO [train.py:1198] (1/4) Epoch 36, batch 2900, loss[loss=0.1858, ctc_loss=0.1193, cr_loss=0.3323, over 17087.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1245, cr_loss=0.3396, over 3351009.28 frames. ], batch size: 43, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:10:59,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=649884.6666666666, ans=0.0
2024-09-25 03:11:45,703 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.278e+02 1.363e+02 1.442e+02 1.816e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-25 03:11:46,413 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.46 vs. limit=22.5
2024-09-25 03:12:03,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.19 vs. limit=15.0
2024-09-25 03:12:08,898 INFO [train.py:1198] (1/4) Epoch 36, batch 2950, loss[loss=0.2223, ctc_loss=0.1463, cr_loss=0.3802, over 16984.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.125, cr_loss=0.3406, over 3346814.46 frames. ], batch size: 53, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:12:54,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=650211.3333333334, ans=0.125
2024-09-25 03:13:02,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=650258.0, ans=0.125
2024-09-25 03:13:06,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.59 vs. limit=15.0
2024-09-25 03:13:07,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=650258.0, ans=0.125
2024-09-25 03:13:10,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=650258.0, ans=0.125
2024-09-25 03:13:23,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=650304.6666666666, ans=0.95
2024-09-25 03:13:25,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=650304.6666666666, ans=0.07
2024-09-25 03:13:29,061 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.23 vs. limit=15.0
2024-09-25 03:13:32,846 INFO [train.py:1198] (1/4) Epoch 36, batch 3000, loss[loss=0.2077, ctc_loss=0.1358, cr_loss=0.3595, over 17152.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1247, cr_loss=0.3402, over 3350192.94 frames. ], batch size: 48, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:13:32,847 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 03:13:48,652 INFO [train.py:1230] (1/4) Epoch 36, validation: loss=0.03616, ctc_loss=0.03616, cr_loss=9.264e-15, over 944034.00 frames. 
2024-09-25 03:13:48,653 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 03:13:49,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=650351.3333333334, ans=10.0
2024-09-25 03:13:58,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=650351.3333333334, ans=0.2
2024-09-25 03:14:26,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=650444.6666666666, ans=0.2
2024-09-25 03:14:32,900 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:14:47,522 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=5.085e-02
2024-09-25 03:14:48,918 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.273e+02 1.377e+02 1.498e+02 1.977e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-25 03:14:52,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=650538.0, ans=0.0
2024-09-25 03:14:59,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=650538.0, ans=0.0
2024-09-25 03:15:06,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=650538.0, ans=0.125
2024-09-25 03:15:09,705 INFO [train.py:1198] (1/4) Epoch 36, batch 3050, loss[loss=0.2213, ctc_loss=0.1451, cr_loss=0.381, over 16484.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1249, cr_loss=0.3409, over 3356781.48 frames. ], batch size: 66, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:15:22,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=650584.6666666666, ans=0.125
2024-09-25 03:15:34,480 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.84 vs. limit=6.0
2024-09-25 03:15:47,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=650678.0, ans=0.0
2024-09-25 03:15:49,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=650678.0, ans=0.2
2024-09-25 03:15:50,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=650678.0, ans=0.025
2024-09-25 03:15:59,549 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=4.05 vs. limit=15.0
2024-09-25 03:16:03,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=650724.6666666666, ans=0.125
2024-09-25 03:16:08,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=650724.6666666666, ans=0.025
2024-09-25 03:16:11,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=650771.3333333334, ans=0.125
2024-09-25 03:16:19,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=650771.3333333334, ans=0.2
2024-09-25 03:16:22,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=650771.3333333334, ans=0.125
2024-09-25 03:16:24,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=650771.3333333334, ans=10.0
2024-09-25 03:16:28,398 INFO [train.py:1198] (1/4) Epoch 36, batch 3100, loss[loss=0.207, ctc_loss=0.135, cr_loss=0.3598, over 17259.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.125, cr_loss=0.3419, over 3367355.32 frames. ], batch size: 44, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:16:28,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=650818.0, ans=0.1
2024-09-25 03:16:30,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=650818.0, ans=0.0
2024-09-25 03:16:34,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=650818.0, ans=0.0
2024-09-25 03:16:40,191 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=6.02 vs. limit=15.0
2024-09-25 03:16:56,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=650864.6666666666, ans=0.1
2024-09-25 03:17:01,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=650911.3333333334, ans=0.1
2024-09-25 03:17:07,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=650911.3333333334, ans=0.125
2024-09-25 03:17:10,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=650911.3333333334, ans=0.0
2024-09-25 03:17:26,168 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.257e+02 1.357e+02 1.451e+02 2.020e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-25 03:17:41,564 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.34 vs. limit=22.5
2024-09-25 03:17:46,709 INFO [train.py:1198] (1/4) Epoch 36, batch 3150, loss[loss=0.1756, ctc_loss=0.1118, cr_loss=0.3189, over 17092.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1252, cr_loss=0.342, over 3370003.98 frames. ], batch size: 40, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:17:53,680 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.55 vs. limit=6.0
2024-09-25 03:17:59,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.63 vs. limit=15.0
2024-09-25 03:18:01,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=651098.0, ans=0.0
2024-09-25 03:18:04,518 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.45 vs. limit=10.0
2024-09-25 03:18:23,312 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.21 vs. limit=15.0
2024-09-25 03:18:24,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=651144.6666666666, ans=0.125
2024-09-25 03:18:55,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=651238.0, ans=0.0
2024-09-25 03:19:05,093 INFO [train.py:1198] (1/4) Epoch 36, batch 3200, loss[loss=0.2302, ctc_loss=0.1512, cr_loss=0.3949, over 14828.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1246, cr_loss=0.3408, over 3369709.87 frames. ], batch size: 89, lr: 3.29e-03, grad_scale: 32.0
2024-09-25 03:19:19,885 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.30 vs. limit=22.5
2024-09-25 03:19:21,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=651331.3333333334, ans=0.025
2024-09-25 03:19:28,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=651331.3333333334, ans=0.125
2024-09-25 03:19:56,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=651424.6666666666, ans=0.125
2024-09-25 03:20:02,961 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.272e+02 1.367e+02 1.451e+02 1.708e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-25 03:20:09,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=651471.3333333334, ans=0.125
2024-09-25 03:20:23,375 INFO [train.py:1198] (1/4) Epoch 36, batch 3250, loss[loss=0.1769, ctc_loss=0.1139, cr_loss=0.3151, over 17137.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1245, cr_loss=0.3399, over 3367887.93 frames. ], batch size: 48, lr: 3.29e-03, grad_scale: 32.0
2024-09-25 03:20:26,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=651518.0, ans=0.125
2024-09-25 03:20:31,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=651518.0, ans=0.1
2024-09-25 03:20:36,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=651518.0, ans=0.125
2024-09-25 03:20:54,927 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.04 vs. limit=15.0
2024-09-25 03:21:05,535 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.40 vs. limit=15.0
2024-09-25 03:21:11,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=651658.0, ans=0.1
2024-09-25 03:21:19,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=651658.0, ans=0.125
2024-09-25 03:21:19,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=651658.0, ans=0.0
2024-09-25 03:21:20,972 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.78 vs. limit=15.0
2024-09-25 03:21:43,851 INFO [train.py:1198] (1/4) Epoch 36, batch 3300, loss[loss=0.2261, ctc_loss=0.1545, cr_loss=0.3579, over 11590.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1247, cr_loss=0.3401, over 3362080.99 frames. ], batch size: 123, lr: 3.29e-03, grad_scale: 32.0
2024-09-25 03:21:53,039 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.99 vs. limit=15.0
2024-09-25 03:22:25,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.98 vs. limit=15.0
2024-09-25 03:22:29,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=651844.6666666666, ans=0.0
2024-09-25 03:22:43,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=651891.3333333334, ans=0.1
2024-09-25 03:22:44,635 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.278e+02 1.344e+02 1.481e+02 2.113e+02, threshold=2.688e+02, percent-clipped=0.0
2024-09-25 03:22:46,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=651938.0, ans=0.2
2024-09-25 03:23:03,251 INFO [train.py:1198] (1/4) Epoch 36, batch 3350, loss[loss=0.1832, ctc_loss=0.1205, cr_loss=0.3136, over 17014.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1242, cr_loss=0.3393, over 3359569.98 frames. ], batch size: 51, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:23:19,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=652031.3333333334, ans=0.1
2024-09-25 03:23:23,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=652031.3333333334, ans=0.2
2024-09-25 03:23:33,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=652031.3333333334, ans=0.2
2024-09-25 03:23:40,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=652078.0, ans=0.125
2024-09-25 03:24:00,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=652124.6666666666, ans=0.125
2024-09-25 03:24:00,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=652124.6666666666, ans=0.125
2024-09-25 03:24:03,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=652124.6666666666, ans=0.125
2024-09-25 03:24:04,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.92 vs. limit=12.0
2024-09-25 03:24:23,598 INFO [train.py:1198] (1/4) Epoch 36, batch 3400, loss[loss=0.2174, ctc_loss=0.1417, cr_loss=0.3787, over 17344.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1251, cr_loss=0.3414, over 3358906.97 frames. ], batch size: 52, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:24:23,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=652218.0, ans=0.125
2024-09-25 03:24:36,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=652218.0, ans=0.2
2024-09-25 03:24:45,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=652264.6666666666, ans=0.0
2024-09-25 03:24:50,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=652264.6666666666, ans=0.04949747468305833
2024-09-25 03:24:51,318 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.31 vs. limit=15.0
2024-09-25 03:24:53,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=652311.3333333334, ans=0.2
2024-09-25 03:24:55,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=652311.3333333334, ans=0.1
2024-09-25 03:25:22,900 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.287e+02 1.358e+02 1.469e+02 2.050e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-25 03:25:33,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=652404.6666666666, ans=0.125
2024-09-25 03:25:44,230 INFO [train.py:1198] (1/4) Epoch 36, batch 3450, loss[loss=0.2175, ctc_loss=0.1427, cr_loss=0.3739, over 17300.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1251, cr_loss=0.3408, over 3358168.66 frames. ], batch size: 46, lr: 3.29e-03, grad_scale: 16.0
2024-09-25 03:25:48,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1.whitening_limit, batch_count=652451.3333333334, ans=10.0
2024-09-25 03:25:53,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=652451.3333333334, ans=0.2
2024-09-25 03:26:04,936 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:26:17,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=652544.6666666666, ans=0.2
2024-09-25 03:26:18,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=652544.6666666666, ans=0.07
2024-09-25 03:26:24,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=652544.6666666666, ans=15.0
2024-09-25 03:26:41,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=652591.3333333334, ans=0.0
2024-09-25 03:27:02,027 INFO [train.py:1198] (1/4) Epoch 36, batch 3500, loss[loss=0.1855, ctc_loss=0.1187, cr_loss=0.3338, over 17252.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1258, cr_loss=0.3426, over 3356331.97 frames. ], batch size: 44, lr: 3.28e-03, grad_scale: 16.0
2024-09-25 03:27:02,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=652684.6666666666, ans=0.125
2024-09-25 03:27:13,385 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.07 vs. limit=15.0
2024-09-25 03:27:16,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=652731.3333333334, ans=0.025
2024-09-25 03:27:26,236 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.84 vs. limit=15.0
2024-09-25 03:27:30,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=652731.3333333334, ans=0.0
2024-09-25 03:27:58,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=652824.6666666666, ans=0.125
2024-09-25 03:28:02,936 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.297e+02 1.416e+02 1.575e+02 3.387e+02, threshold=2.833e+02, percent-clipped=1.0
2024-09-25 03:28:20,554 INFO [train.py:1198] (1/4) Epoch 36, batch 3550, loss[loss=0.1909, ctc_loss=0.125, cr_loss=0.3298, over 17015.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.126, cr_loss=0.3438, over 3359237.08 frames. ], batch size: 51, lr: 3.28e-03, grad_scale: 8.0
2024-09-25 03:28:38,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.43 vs. limit=6.0
2024-09-25 03:28:39,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=652964.6666666666, ans=0.0
2024-09-25 03:29:03,509 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.46 vs. limit=15.0
2024-09-25 03:29:04,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=653011.3333333334, ans=0.125
2024-09-25 03:29:20,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=653058.0, ans=0.2
2024-09-25 03:29:38,745 INFO [train.py:1198] (1/4) Epoch 36, batch 3600, loss[loss=0.1757, ctc_loss=0.1117, cr_loss=0.3202, over 17310.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1262, cr_loss=0.344, over 3361433.16 frames. ], batch size: 46, lr: 3.28e-03, grad_scale: 16.0
2024-09-25 03:30:38,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=653291.3333333334, ans=0.125
2024-09-25 03:30:43,995 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.277e+02 1.347e+02 1.445e+02 1.925e+02, threshold=2.694e+02, percent-clipped=0.0
2024-09-25 03:30:48,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=653338.0, ans=0.2
2024-09-25 03:30:50,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=653338.0, ans=0.1
2024-09-25 03:31:00,963 INFO [train.py:1198] (1/4) Epoch 36, batch 3650, loss[loss=0.2108, ctc_loss=0.1402, cr_loss=0.3528, over 17172.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1269, cr_loss=0.3451, over 3359612.46 frames. ], batch size: 48, lr: 3.28e-03, grad_scale: 16.0
2024-09-25 03:31:05,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=653384.6666666666, ans=0.125
2024-09-25 03:31:17,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=653431.3333333334, ans=0.125
2024-09-25 03:31:42,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=653478.0, ans=0.0
2024-09-25 03:31:54,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=653524.6666666666, ans=0.2
2024-09-25 03:32:22,170 INFO [train.py:1198] (1/4) Epoch 36, batch 3700, loss[loss=0.2178, ctc_loss=0.1423, cr_loss=0.3773, over 17208.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.1261, cr_loss=0.3434, over 3364425.38 frames. ], batch size: 55, lr: 3.28e-03, grad_scale: 16.0
2024-09-25 03:32:33,930 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.33 vs. limit=6.0
2024-09-25 03:32:41,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=653664.6666666666, ans=0.0
2024-09-25 03:33:22,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=653758.0, ans=0.0
2024-09-25 03:33:24,003 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.267e+02 1.341e+02 1.465e+02 1.805e+02, threshold=2.682e+02, percent-clipped=0.0
2024-09-25 03:33:30,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=653804.6666666666, ans=0.1
2024-09-25 03:33:40,749 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.86 vs. limit=10.0
2024-09-25 03:33:41,505 INFO [train.py:1198] (1/4) Epoch 36, batch 3750, loss[loss=0.1983, ctc_loss=0.1288, cr_loss=0.3472, over 16795.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1262, cr_loss=0.3437, over 3353957.48 frames. ], batch size: 61, lr: 3.28e-03, grad_scale: 16.0
2024-09-25 03:34:11,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=653944.6666666666, ans=0.0
2024-09-25 03:34:22,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=653944.6666666666, ans=0.025
2024-09-25 03:34:40,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=653991.3333333334, ans=0.0
2024-09-25 03:35:01,334 INFO [train.py:1198] (1/4) Epoch 36, batch 3800, loss[loss=0.1712, ctc_loss=0.1099, cr_loss=0.3068, over 17021.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1254, cr_loss=0.3422, over 3351622.34 frames. ], batch size: 44, lr: 3.28e-03, grad_scale: 16.0
2024-09-25 03:35:21,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=654131.3333333334, ans=0.05
2024-09-25 03:35:42,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=654178.0, ans=0.125
2024-09-25 03:35:48,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=654224.6666666666, ans=0.0
2024-09-25 03:35:50,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=654224.6666666666, ans=0.125
2024-09-25 03:36:02,509 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.295e+02 1.360e+02 1.464e+02 2.754e+02, threshold=2.720e+02, percent-clipped=1.0
2024-09-25 03:36:02,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=654271.3333333334, ans=0.95
2024-09-25 03:36:18,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=654318.0, ans=0.0
2024-09-25 03:36:19,640 INFO [train.py:1198] (1/4) Epoch 36, batch 3850, loss[loss=0.1922, ctc_loss=0.1223, cr_loss=0.3492, over 17148.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1254, cr_loss=0.3413, over 3311253.31 frames. ], batch size: 45, lr: 3.28e-03, grad_scale: 16.0
2024-09-25 03:36:36,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=654364.6666666666, ans=0.0
2024-09-25 03:36:41,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=654364.6666666666, ans=0.2
2024-09-25 03:36:52,236 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=7.03 vs. limit=10.0
2024-09-25 03:37:15,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=654458.0, ans=0.0
2024-09-25 03:37:23,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=654504.6666666666, ans=0.125
2024-09-25 03:38:18,970 INFO [train.py:1198] (1/4) Epoch 37, batch 0, loss[loss=0.2543, ctc_loss=0.1686, cr_loss=0.4282, over 15100.00 frames. ], tot_loss[loss=0.2543, ctc_loss=0.1686, cr_loss=0.4282, over 15100.00 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:38:18,971 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 03:38:34,307 INFO [train.py:1230] (1/4) Epoch 37, validation: loss=0.03489, ctc_loss=0.03489, cr_loss=9.463e-15, over 944034.00 frames. 
2024-09-25 03:38:34,308 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 03:38:36,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=654532.6666666666, ans=0.0
2024-09-25 03:38:42,780 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:38:55,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=654579.3333333334, ans=0.125
2024-09-25 03:38:58,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=654579.3333333334, ans=0.1
2024-09-25 03:39:05,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=654626.0, ans=0.125
2024-09-25 03:39:16,638 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.12 vs. limit=15.0
2024-09-25 03:39:20,774 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=14.43 vs. limit=22.5
2024-09-25 03:39:26,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=654672.6666666666, ans=0.0
2024-09-25 03:39:38,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=654672.6666666666, ans=0.125
2024-09-25 03:39:39,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=654719.3333333334, ans=0.09899494936611666
2024-09-25 03:39:47,467 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.358e+02 1.528e+02 1.726e+02 8.114e+02, threshold=3.057e+02, percent-clipped=1.0
2024-09-25 03:39:52,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=654719.3333333334, ans=0.0
2024-09-25 03:39:57,023 INFO [train.py:1198] (1/4) Epoch 37, batch 50, loss[loss=0.1814, ctc_loss=0.1181, cr_loss=0.3169, over 17173.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.1273, cr_loss=0.3452, over 760292.74 frames. ], batch size: 45, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:41:09,883 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:41:14,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=654952.6666666666, ans=0.2
2024-09-25 03:41:16,430 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.48 vs. limit=15.0
2024-09-25 03:41:19,002 INFO [train.py:1198] (1/4) Epoch 37, batch 100, loss[loss=0.1853, ctc_loss=0.1192, cr_loss=0.3307, over 17012.00 frames. ], tot_loss[loss=0.1948, ctc_loss=0.1261, cr_loss=0.3433, over 1334591.19 frames. ], batch size: 44, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:41:21,010 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:41:25,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=654999.3333333334, ans=0.125
2024-09-25 03:42:03,250 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.48 vs. limit=6.0
2024-09-25 03:42:29,428 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.259e+02 1.323e+02 1.391e+02 2.896e+02, threshold=2.646e+02, percent-clipped=0.0
2024-09-25 03:42:32,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=655186.0, ans=0.1
2024-09-25 03:42:39,024 INFO [train.py:1198] (1/4) Epoch 37, batch 150, loss[loss=0.2032, ctc_loss=0.1338, cr_loss=0.347, over 17214.00 frames. ], tot_loss[loss=0.1965, ctc_loss=0.1274, cr_loss=0.3452, over 1768179.54 frames. ], batch size: 55, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:43:52,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=655419.3333333334, ans=0.09899494936611666
2024-09-25 03:44:01,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=655419.3333333334, ans=0.07
2024-09-25 03:44:07,479 INFO [train.py:1198] (1/4) Epoch 37, batch 200, loss[loss=0.2117, ctc_loss=0.136, cr_loss=0.3789, over 17015.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.127, cr_loss=0.3456, over 2133382.07 frames. ], batch size: 53, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:44:11,078 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.94 vs. limit=6.0
2024-09-25 03:44:12,510 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:44:31,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=655512.6666666666, ans=0.125
2024-09-25 03:44:59,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=655606.0, ans=0.07
2024-09-25 03:45:09,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=655652.6666666666, ans=0.0
2024-09-25 03:45:19,825 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.270e+02 1.344e+02 1.450e+02 1.913e+02, threshold=2.687e+02, percent-clipped=0.0
2024-09-25 03:45:29,187 INFO [train.py:1198] (1/4) Epoch 37, batch 250, loss[loss=0.216, ctc_loss=0.142, cr_loss=0.3701, over 17030.00 frames. ], tot_loss[loss=0.1952, ctc_loss=0.1266, cr_loss=0.3431, over 2401951.55 frames. ], batch size: 52, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:45:35,183 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.34 vs. limit=15.0
2024-09-25 03:45:43,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.21 vs. limit=12.0
2024-09-25 03:45:52,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=655746.0, ans=0.05
2024-09-25 03:46:03,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=655792.6666666666, ans=0.125
2024-09-25 03:46:30,625 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=10.64 vs. limit=22.5
2024-09-25 03:46:32,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=655886.0, ans=0.0
2024-09-25 03:46:36,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=655886.0, ans=0.0
2024-09-25 03:46:50,164 INFO [train.py:1198] (1/4) Epoch 37, batch 300, loss[loss=0.197, ctc_loss=0.1285, cr_loss=0.3427, over 17312.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1257, cr_loss=0.3423, over 2625241.16 frames. ], batch size: 49, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:46:55,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=655932.6666666666, ans=0.0
2024-09-25 03:47:06,595 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:47:17,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=655979.3333333334, ans=0.125
2024-09-25 03:47:27,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=656026.0, ans=0.125
2024-09-25 03:47:41,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=656072.6666666666, ans=0.125
2024-09-25 03:47:57,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=656119.3333333334, ans=0.0
2024-09-25 03:48:00,707 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.285e+02 1.370e+02 1.458e+02 2.873e+02, threshold=2.740e+02, percent-clipped=1.0
2024-09-25 03:48:02,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=656119.3333333334, ans=0.0
2024-09-25 03:48:10,573 INFO [train.py:1198] (1/4) Epoch 37, batch 350, loss[loss=0.1874, ctc_loss=0.121, cr_loss=0.3319, over 17210.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1252, cr_loss=0.3422, over 2789364.56 frames. ], batch size: 47, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:48:19,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=656166.0, ans=0.0
2024-09-25 03:48:21,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=656166.0, ans=0.0
2024-09-25 03:48:24,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=656166.0, ans=0.125
2024-09-25 03:48:36,404 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.68 vs. limit=6.0
2024-09-25 03:48:54,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=656259.3333333334, ans=0.2
2024-09-25 03:48:58,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=656259.3333333334, ans=0.125
2024-09-25 03:49:10,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=656306.0, ans=0.125
2024-09-25 03:49:11,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=656306.0, ans=0.125
2024-09-25 03:49:24,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=656352.6666666666, ans=0.0
2024-09-25 03:49:38,694 INFO [train.py:1198] (1/4) Epoch 37, batch 400, loss[loss=0.1879, ctc_loss=0.1222, cr_loss=0.3283, over 16880.00 frames. ], tot_loss[loss=0.1952, ctc_loss=0.1265, cr_loss=0.3439, over 2891969.80 frames. ], batch size: 58, lr: 3.23e-03, grad_scale: 32.0
2024-09-25 03:49:39,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=656399.3333333334, ans=0.1
2024-09-25 03:50:01,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=656446.0, ans=0.125
2024-09-25 03:50:01,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=656446.0, ans=0.07
2024-09-25 03:50:19,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=656492.6666666666, ans=0.0
2024-09-25 03:50:25,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=656492.6666666666, ans=0.125
2024-09-25 03:50:25,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=656492.6666666666, ans=0.2
2024-09-25 03:50:34,325 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.45 vs. limit=15.0
2024-09-25 03:50:37,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=656539.3333333334, ans=0.125
2024-09-25 03:50:41,648 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.73 vs. limit=6.0
2024-09-25 03:50:51,380 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.258e+02 1.358e+02 1.447e+02 2.750e+02, threshold=2.715e+02, percent-clipped=1.0
2024-09-25 03:51:01,056 INFO [train.py:1198] (1/4) Epoch 37, batch 450, loss[loss=0.2272, ctc_loss=0.1511, cr_loss=0.3806, over 16883.00 frames. ], tot_loss[loss=0.196, ctc_loss=0.1271, cr_loss=0.3447, over 2990006.97 frames. ], batch size: 58, lr: 3.23e-03, grad_scale: 32.0
2024-09-25 03:51:06,531 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.01 vs. limit=15.0
2024-09-25 03:51:07,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=656632.6666666666, ans=0.1
2024-09-25 03:51:37,276 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.99 vs. limit=15.0
2024-09-25 03:51:41,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=656726.0, ans=0.0
2024-09-25 03:51:54,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=656772.6666666666, ans=0.0
2024-09-25 03:51:59,797 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=5.07 vs. limit=15.0
2024-09-25 03:52:10,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=656819.3333333334, ans=0.125
2024-09-25 03:52:19,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=656866.0, ans=0.0
2024-09-25 03:52:21,376 INFO [train.py:1198] (1/4) Epoch 37, batch 500, loss[loss=0.1985, ctc_loss=0.1283, cr_loss=0.3509, over 17090.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1265, cr_loss=0.3441, over 3072014.53 frames. ], batch size: 49, lr: 3.23e-03, grad_scale: 32.0
2024-09-25 03:52:27,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=656866.0, ans=0.125
2024-09-25 03:52:58,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.88 vs. limit=15.0
2024-09-25 03:52:58,983 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.66 vs. limit=15.0
2024-09-25 03:53:10,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=657006.0, ans=0.125
2024-09-25 03:53:20,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=657006.0, ans=0.125
2024-09-25 03:53:22,244 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.21 vs. limit=15.0
2024-09-25 03:53:37,041 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.284e+02 1.360e+02 1.514e+02 2.432e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-25 03:53:46,708 INFO [train.py:1198] (1/4) Epoch 37, batch 550, loss[loss=0.147, ctc_loss=0.09314, cr_loss=0.2692, over 16357.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1257, cr_loss=0.3429, over 3143619.36 frames. ], batch size: 36, lr: 3.23e-03, grad_scale: 32.0
2024-09-25 03:54:45,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=657239.3333333334, ans=0.125
2024-09-25 03:55:00,873 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.20 vs. limit=15.0
2024-09-25 03:55:07,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=657286.0, ans=0.5
2024-09-25 03:55:09,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=657286.0, ans=0.125
2024-09-25 03:55:12,193 INFO [train.py:1198] (1/4) Epoch 37, batch 600, loss[loss=0.209, ctc_loss=0.1363, cr_loss=0.3632, over 16641.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.126, cr_loss=0.3438, over 3190587.43 frames. ], batch size: 66, lr: 3.23e-03, grad_scale: 32.0
2024-09-25 03:55:12,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=657332.6666666666, ans=0.0
2024-09-25 03:55:15,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=657332.6666666666, ans=0.025
2024-09-25 03:55:25,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=657332.6666666666, ans=0.0
2024-09-25 03:55:54,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=657426.0, ans=0.0
2024-09-25 03:56:10,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=657472.6666666666, ans=0.125
2024-09-25 03:56:12,082 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.87 vs. limit=15.0
2024-09-25 03:56:22,507 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.287e+02 1.356e+02 1.479e+02 3.442e+02, threshold=2.712e+02, percent-clipped=1.0
2024-09-25 03:56:32,226 INFO [train.py:1198] (1/4) Epoch 37, batch 650, loss[loss=0.1814, ctc_loss=0.1174, cr_loss=0.32, over 17227.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1257, cr_loss=0.3432, over 3230248.77 frames. ], batch size: 47, lr: 3.23e-03, grad_scale: 32.0
2024-09-25 03:56:38,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=657566.0, ans=0.2
2024-09-25 03:57:17,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.89 vs. limit=15.0
2024-09-25 03:57:31,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=657706.0, ans=0.1
2024-09-25 03:57:41,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=657752.6666666666, ans=0.04949747468305833
2024-09-25 03:57:52,560 INFO [train.py:1198] (1/4) Epoch 37, batch 700, loss[loss=0.2107, ctc_loss=0.1388, cr_loss=0.3592, over 15404.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.1251, cr_loss=0.342, over 3257299.72 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:58:03,707 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 03:58:08,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=657799.3333333334, ans=0.1
2024-09-25 03:58:18,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=657846.0, ans=0.125
2024-09-25 03:59:13,008 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.266e+02 1.364e+02 1.473e+02 1.883e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-25 03:59:21,270 INFO [train.py:1198] (1/4) Epoch 37, batch 750, loss[loss=0.1551, ctc_loss=0.09499, cr_loss=0.3008, over 17175.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1252, cr_loss=0.3425, over 3281157.07 frames. ], batch size: 41, lr: 3.23e-03, grad_scale: 16.0
2024-09-25 03:59:25,130 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.95 vs. limit=15.0
2024-09-25 03:59:39,634 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.91 vs. limit=6.0
2024-09-25 03:59:45,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=658079.3333333334, ans=0.0
2024-09-25 04:00:05,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=658126.0, ans=0.0
2024-09-25 04:00:31,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.38 vs. limit=15.0
2024-09-25 04:00:40,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=658219.3333333334, ans=0.025
2024-09-25 04:00:43,389 INFO [train.py:1198] (1/4) Epoch 37, batch 800, loss[loss=0.2405, ctc_loss=0.1557, cr_loss=0.424, over 16766.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1247, cr_loss=0.3417, over 3296568.96 frames. ], batch size: 61, lr: 3.23e-03, grad_scale: 32.0
2024-09-25 04:00:55,011 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.94 vs. limit=15.0
2024-09-25 04:01:38,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=9.43 vs. limit=15.0
2024-09-25 04:01:54,703 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.298e+02 1.352e+02 1.475e+02 2.154e+02, threshold=2.705e+02, percent-clipped=0.0
2024-09-25 04:01:56,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=658452.6666666666, ans=0.035
2024-09-25 04:02:02,798 INFO [train.py:1198] (1/4) Epoch 37, batch 850, loss[loss=0.2103, ctc_loss=0.1311, cr_loss=0.3962, over 17091.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1255, cr_loss=0.3432, over 3296728.48 frames. ], batch size: 49, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:02:35,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=658592.6666666666, ans=0.09899494936611666
2024-09-25 04:03:25,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=658686.0, ans=0.125
2024-09-25 04:03:28,268 INFO [train.py:1198] (1/4) Epoch 37, batch 900, loss[loss=0.2045, ctc_loss=0.1344, cr_loss=0.3507, over 16988.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1251, cr_loss=0.3423, over 3311124.67 frames. ], batch size: 51, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:03:35,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=658732.6666666666, ans=0.05
2024-09-25 04:04:14,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=658826.0, ans=0.125
2024-09-25 04:04:33,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=658919.3333333334, ans=0.025
2024-09-25 04:04:41,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=658919.3333333334, ans=0.2
2024-09-25 04:04:43,135 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.276e+02 1.332e+02 1.387e+02 1.934e+02, threshold=2.664e+02, percent-clipped=0.0
2024-09-25 04:04:51,042 INFO [train.py:1198] (1/4) Epoch 37, batch 950, loss[loss=0.1757, ctc_loss=0.1133, cr_loss=0.3119, over 17086.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1253, cr_loss=0.3427, over 3312456.42 frames. ], batch size: 43, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:04:59,431 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.77 vs. limit=15.0
2024-09-25 04:05:00,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=658966.0, ans=0.0
2024-09-25 04:05:01,005 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=7.31 vs. limit=15.0
2024-09-25 04:05:35,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=659059.3333333334, ans=0.125
2024-09-25 04:05:40,021 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.43 vs. limit=10.0
2024-09-25 04:05:58,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=659152.6666666666, ans=0.125
2024-09-25 04:06:10,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=659152.6666666666, ans=0.09899494936611666
2024-09-25 04:06:13,085 INFO [train.py:1198] (1/4) Epoch 37, batch 1000, loss[loss=0.2144, ctc_loss=0.1375, cr_loss=0.3844, over 17085.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.125, cr_loss=0.3427, over 3324463.11 frames. ], batch size: 49, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:06:14,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=659199.3333333334, ans=0.125
2024-09-25 04:07:10,134 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.10 vs. limit=22.5
2024-09-25 04:07:25,165 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.283e+02 1.388e+02 1.495e+02 1.963e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-25 04:07:33,125 INFO [train.py:1198] (1/4) Epoch 37, batch 1050, loss[loss=0.1806, ctc_loss=0.1148, cr_loss=0.3288, over 16977.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3414, over 3341093.06 frames. ], batch size: 42, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:07:46,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=659432.6666666666, ans=0.0
2024-09-25 04:08:23,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=659572.6666666666, ans=0.125
2024-09-25 04:08:34,694 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.65 vs. limit=15.0
2024-09-25 04:08:36,853 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.27 vs. limit=15.0
2024-09-25 04:08:54,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=659619.3333333334, ans=0.2
2024-09-25 04:09:00,281 INFO [train.py:1198] (1/4) Epoch 37, batch 1100, loss[loss=0.2032, ctc_loss=0.1307, cr_loss=0.3622, over 17023.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1245, cr_loss=0.3417, over 3350502.94 frames. ], batch size: 44, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:09:32,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=659759.3333333334, ans=0.2
2024-09-25 04:09:53,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=659806.0, ans=0.025
2024-09-25 04:09:53,768 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.93 vs. limit=22.5
2024-09-25 04:10:05,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=659852.6666666666, ans=0.2
2024-09-25 04:10:11,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=659852.6666666666, ans=0.2
2024-09-25 04:10:14,845 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.287e+02 1.378e+02 1.545e+02 2.459e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 04:10:22,843 INFO [train.py:1198] (1/4) Epoch 37, batch 1150, loss[loss=0.1989, ctc_loss=0.1334, cr_loss=0.3278, over 17228.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1247, cr_loss=0.342, over 3354703.23 frames. ], batch size: 55, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:10:25,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten.whitening_limit, batch_count=659899.3333333334, ans=15.0
2024-09-25 04:11:14,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=660039.3333333334, ans=0.125
2024-09-25 04:11:35,996 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.91 vs. limit=22.5
2024-09-25 04:11:43,113 INFO [train.py:1198] (1/4) Epoch 37, batch 1200, loss[loss=0.1858, ctc_loss=0.1199, cr_loss=0.3298, over 17016.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1252, cr_loss=0.3434, over 3358236.37 frames. ], batch size: 56, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:11:49,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=660132.6666666666, ans=0.2
2024-09-25 04:12:21,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=660226.0, ans=0.125
2024-09-25 04:12:26,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=660226.0, ans=0.1
2024-09-25 04:12:48,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=660319.3333333334, ans=0.0
2024-09-25 04:12:57,285 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.254e+02 1.340e+02 1.432e+02 2.120e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-25 04:13:03,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=660366.0, ans=0.2
2024-09-25 04:13:05,187 INFO [train.py:1198] (1/4) Epoch 37, batch 1250, loss[loss=0.1834, ctc_loss=0.1203, cr_loss=0.3153, over 17210.00 frames. ], tot_loss[loss=0.1933, ctc_loss=0.1249, cr_loss=0.3421, over 3367903.01 frames. ], batch size: 47, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:13:14,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=660366.0, ans=0.1
2024-09-25 04:13:15,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.max_abs, batch_count=660366.0, ans=10.0
2024-09-25 04:13:53,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=660459.3333333334, ans=0.125
2024-09-25 04:14:30,159 INFO [train.py:1198] (1/4) Epoch 37, batch 1300, loss[loss=0.1927, ctc_loss=0.1239, cr_loss=0.344, over 17211.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1243, cr_loss=0.341, over 3374109.93 frames. ], batch size: 47, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:14:54,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=660646.0, ans=0.2
2024-09-25 04:15:08,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=660692.6666666666, ans=0.1
2024-09-25 04:15:12,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=660692.6666666666, ans=0.0
2024-09-25 04:15:31,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=660739.3333333334, ans=0.0
2024-09-25 04:15:35,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.88 vs. limit=6.0
2024-09-25 04:15:36,806 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 04:15:44,612 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.301e+02 1.386e+02 1.515e+02 1.831e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 04:15:52,776 INFO [train.py:1198] (1/4) Epoch 37, batch 1350, loss[loss=0.1807, ctc_loss=0.1126, cr_loss=0.3404, over 17253.00 frames. ], tot_loss[loss=0.1933, ctc_loss=0.1249, cr_loss=0.342, over 3377885.79 frames. ], batch size: 42, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:16:12,280 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 04:16:20,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=660879.3333333334, ans=0.125
2024-09-25 04:16:20,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=660879.3333333334, ans=0.2
2024-09-25 04:16:44,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=660972.6666666666, ans=0.09899494936611666
2024-09-25 04:16:57,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=661019.3333333334, ans=0.2
2024-09-25 04:17:08,954 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.05 vs. limit=15.0
2024-09-25 04:17:11,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=661066.0, ans=0.125
2024-09-25 04:17:13,042 INFO [train.py:1198] (1/4) Epoch 37, batch 1400, loss[loss=0.2069, ctc_loss=0.1314, cr_loss=0.3772, over 16750.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1256, cr_loss=0.3435, over 3370750.54 frames. ], batch size: 61, lr: 3.22e-03, grad_scale: 16.0
2024-09-25 04:17:13,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=661066.0, ans=0.0
2024-09-25 04:17:47,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=661159.3333333334, ans=0.1
2024-09-25 04:18:28,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=661252.6666666666, ans=0.1
2024-09-25 04:18:31,385 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.311e+02 1.398e+02 1.519e+02 2.404e+02, threshold=2.797e+02, percent-clipped=0.0
2024-09-25 04:18:40,188 INFO [train.py:1198] (1/4) Epoch 37, batch 1450, loss[loss=0.2101, ctc_loss=0.1382, cr_loss=0.3597, over 16718.00 frames. ], tot_loss[loss=0.1957, ctc_loss=0.1266, cr_loss=0.3457, over 3371899.25 frames. ], batch size: 61, lr: 3.22e-03, grad_scale: 16.0
2024-09-25 04:18:42,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=661299.3333333334, ans=0.125
2024-09-25 04:18:51,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=661299.3333333334, ans=0.125
2024-09-25 04:19:19,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=661392.6666666666, ans=0.125
2024-09-25 04:19:27,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=661439.3333333334, ans=0.0
2024-09-25 04:19:41,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=661439.3333333334, ans=0.0
2024-09-25 04:19:51,951 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.00 vs. limit=12.0
2024-09-25 04:20:02,950 INFO [train.py:1198] (1/4) Epoch 37, batch 1500, loss[loss=0.15, ctc_loss=0.09679, cr_loss=0.2663, over 17261.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.1259, cr_loss=0.3441, over 3377665.13 frames. ], batch size: 42, lr: 3.22e-03, grad_scale: 16.0
2024-09-25 04:20:06,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=661532.6666666666, ans=0.1
2024-09-25 04:20:09,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=661532.6666666666, ans=0.125
2024-09-25 04:20:16,272 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.78 vs. limit=22.5
2024-09-25 04:20:53,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=661672.6666666666, ans=0.125
2024-09-25 04:21:00,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=661672.6666666666, ans=0.0
2024-09-25 04:21:03,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.max_abs, batch_count=661672.6666666666, ans=10.0
2024-09-25 04:21:16,230 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.237e+02 1.312e+02 1.425e+02 1.916e+02, threshold=2.625e+02, percent-clipped=0.0
2024-09-25 04:21:16,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=661719.3333333334, ans=0.125
2024-09-25 04:21:22,594 INFO [train.py:1198] (1/4) Epoch 37, batch 1550, loss[loss=0.1713, ctc_loss=0.11, cr_loss=0.3066, over 16967.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1252, cr_loss=0.3422, over 3365733.90 frames. ], batch size: 42, lr: 3.22e-03, grad_scale: 16.0
2024-09-25 04:22:19,621 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 04:22:29,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=661952.6666666666, ans=0.125
2024-09-25 04:22:45,505 INFO [train.py:1198] (1/4) Epoch 37, batch 1600, loss[loss=0.1954, ctc_loss=0.1254, cr_loss=0.35, over 17080.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1248, cr_loss=0.3412, over 3354278.60 frames. ], batch size: 46, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:23:25,921 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.93 vs. limit=15.0
2024-09-25 04:23:44,623 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.63 vs. limit=15.0
2024-09-25 04:24:04,105 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.273e+02 1.358e+02 1.458e+02 2.530e+02, threshold=2.716e+02, percent-clipped=0.0
2024-09-25 04:24:07,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=662186.0, ans=0.2
2024-09-25 04:24:10,512 INFO [train.py:1198] (1/4) Epoch 37, batch 1650, loss[loss=0.1897, ctc_loss=0.1276, cr_loss=0.3108, over 17231.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.1242, cr_loss=0.3398, over 3357318.87 frames. ], batch size: 50, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:24:20,688 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.78 vs. limit=15.0
2024-09-25 04:24:21,119 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.87 vs. limit=6.0
2024-09-25 04:24:33,099 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=662279.3333333334, ans=0.125
2024-09-25 04:24:44,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=662326.0, ans=0.125
2024-09-25 04:24:51,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=662326.0, ans=0.125
2024-09-25 04:24:58,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=662326.0, ans=0.0
2024-09-25 04:25:07,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=662372.6666666666, ans=0.125
2024-09-25 04:25:13,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=662372.6666666666, ans=0.1
2024-09-25 04:25:32,851 INFO [train.py:1198] (1/4) Epoch 37, batch 1700, loss[loss=0.1752, ctc_loss=0.1106, cr_loss=0.3231, over 16968.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1246, cr_loss=0.3409, over 3355778.50 frames. ], batch size: 42, lr: 3.22e-03, grad_scale: 32.0
2024-09-25 04:26:07,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=662559.3333333334, ans=0.035
2024-09-25 04:26:37,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=662652.6666666666, ans=0.125
2024-09-25 04:26:41,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=662652.6666666666, ans=0.0
2024-09-25 04:26:45,874 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.286e+02 1.377e+02 1.479e+02 2.270e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-25 04:26:52,404 INFO [train.py:1198] (1/4) Epoch 37, batch 1750, loss[loss=0.1942, ctc_loss=0.1258, cr_loss=0.3421, over 17006.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1247, cr_loss=0.3411, over 3356056.46 frames. ], batch size: 56, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:27:08,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=662746.0, ans=0.025
2024-09-25 04:27:49,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=662839.3333333334, ans=0.0
2024-09-25 04:27:51,155 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 04:28:13,567 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.63 vs. limit=6.0
2024-09-25 04:28:14,221 INFO [train.py:1198] (1/4) Epoch 37, batch 1800, loss[loss=0.1749, ctc_loss=0.1116, cr_loss=0.3162, over 17068.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1249, cr_loss=0.341, over 3349195.56 frames. ], batch size: 46, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:28:18,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=662932.6666666666, ans=0.1
2024-09-25 04:28:42,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=662979.3333333334, ans=0.125
2024-09-25 04:28:45,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=662979.3333333334, ans=0.125
2024-09-25 04:29:12,698 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 04:29:27,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=663119.3333333334, ans=0.1
2024-09-25 04:29:30,047 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.280e+02 1.352e+02 1.451e+02 1.795e+02, threshold=2.704e+02, percent-clipped=0.0
2024-09-25 04:29:37,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=663166.0, ans=0.0
2024-09-25 04:29:39,048 INFO [train.py:1198] (1/4) Epoch 37, batch 1850, loss[loss=0.1913, ctc_loss=0.1234, cr_loss=0.3398, over 16663.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1253, cr_loss=0.3424, over 3349823.40 frames. ], batch size: 61, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:29:58,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=663212.6666666666, ans=0.1
2024-09-25 04:30:07,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=663212.6666666666, ans=0.125
2024-09-25 04:30:15,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=663259.3333333334, ans=0.125
2024-09-25 04:30:20,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=663259.3333333334, ans=0.025
2024-09-25 04:30:33,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=663306.0, ans=0.125
2024-09-25 04:30:39,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=663306.0, ans=0.2
2024-09-25 04:30:49,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=663352.6666666666, ans=0.0
2024-09-25 04:30:58,767 INFO [train.py:1198] (1/4) Epoch 37, batch 1900, loss[loss=0.2206, ctc_loss=0.1446, cr_loss=0.3804, over 17082.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1252, cr_loss=0.3424, over 3357757.70 frames. ], batch size: 49, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:31:08,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=663399.3333333334, ans=0.05
2024-09-25 04:32:08,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=663586.0, ans=0.1
2024-09-25 04:32:12,966 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.273e+02 1.351e+02 1.492e+02 1.881e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-25 04:32:19,249 INFO [train.py:1198] (1/4) Epoch 37, batch 1950, loss[loss=0.1616, ctc_loss=0.1009, cr_loss=0.3036, over 16958.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1261, cr_loss=0.3446, over 3364614.25 frames. ], batch size: 42, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:33:03,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=663726.0, ans=0.125
2024-09-25 04:33:06,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=663726.0, ans=0.2
2024-09-25 04:33:09,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=663772.6666666666, ans=0.0
2024-09-25 04:33:36,019 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.52 vs. limit=15.0
2024-09-25 04:33:45,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=663866.0, ans=0.05
2024-09-25 04:33:46,763 INFO [train.py:1198] (1/4) Epoch 37, batch 2000, loss[loss=0.1921, ctc_loss=0.124, cr_loss=0.3408, over 16988.00 frames. ], tot_loss[loss=0.1948, ctc_loss=0.126, cr_loss=0.3443, over 3370488.93 frames. ], batch size: 56, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:33:47,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=663866.0, ans=0.125
2024-09-25 04:34:09,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=663912.6666666666, ans=0.0
2024-09-25 04:34:28,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=663959.3333333334, ans=0.1
2024-09-25 04:35:02,492 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.268e+02 1.366e+02 1.468e+02 1.745e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-25 04:35:02,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=664052.6666666666, ans=0.2
2024-09-25 04:35:04,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=664052.6666666666, ans=0.1
2024-09-25 04:35:06,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=664052.6666666666, ans=0.125
2024-09-25 04:35:08,884 INFO [train.py:1198] (1/4) Epoch 37, batch 2050, loss[loss=0.1914, ctc_loss=0.1228, cr_loss=0.3428, over 17149.00 frames. ], tot_loss[loss=0.1954, ctc_loss=0.1264, cr_loss=0.345, over 3370932.36 frames. ], batch size: 48, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:35:33,222 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.85 vs. limit=22.5
2024-09-25 04:35:58,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=664239.3333333334, ans=0.0
2024-09-25 04:36:01,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=664239.3333333334, ans=0.2
2024-09-25 04:36:17,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=664286.0, ans=0.2
2024-09-25 04:36:28,588 INFO [train.py:1198] (1/4) Epoch 37, batch 2100, loss[loss=0.1645, ctc_loss=0.1029, cr_loss=0.3081, over 16959.00 frames. ], tot_loss[loss=0.1947, ctc_loss=0.1259, cr_loss=0.3444, over 3360631.80 frames. ], batch size: 42, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:36:39,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=664332.6666666666, ans=0.125
2024-09-25 04:36:46,529 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.17 vs. limit=10.0
2024-09-25 04:37:27,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=664472.6666666666, ans=0.125
2024-09-25 04:37:40,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=664519.3333333334, ans=0.125
2024-09-25 04:37:46,466 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.254e+02 1.343e+02 1.449e+02 1.904e+02, threshold=2.685e+02, percent-clipped=0.0
2024-09-25 04:37:46,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=664519.3333333334, ans=0.0
2024-09-25 04:37:51,357 INFO [train.py:1198] (1/4) Epoch 37, batch 2150, loss[loss=0.1852, ctc_loss=0.1192, cr_loss=0.33, over 17215.00 frames. ], tot_loss[loss=0.1948, ctc_loss=0.126, cr_loss=0.3442, over 3356057.40 frames. ], batch size: 47, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:37:54,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=664566.0, ans=0.1
2024-09-25 04:38:04,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=664566.0, ans=0.125
2024-09-25 04:38:26,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=664659.3333333334, ans=0.0
2024-09-25 04:38:36,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=664659.3333333334, ans=0.125
2024-09-25 04:38:55,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=664706.0, ans=0.0
2024-09-25 04:39:15,888 INFO [train.py:1198] (1/4) Epoch 37, batch 2200, loss[loss=0.1688, ctc_loss=0.1068, cr_loss=0.3096, over 17037.00 frames. ], tot_loss[loss=0.1941, ctc_loss=0.1255, cr_loss=0.3428, over 3348468.01 frames. ], batch size: 39, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:40:05,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=664939.3333333334, ans=0.0
2024-09-25 04:40:14,418 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.71 vs. limit=10.0
2024-09-25 04:40:34,121 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.296e+02 1.362e+02 1.438e+02 1.964e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-25 04:40:37,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=665032.6666666666, ans=0.2
2024-09-25 04:40:38,877 INFO [train.py:1198] (1/4) Epoch 37, batch 2250, loss[loss=0.2184, ctc_loss=0.1434, cr_loss=0.3751, over 17353.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1252, cr_loss=0.3423, over 3359040.34 frames. ], batch size: 48, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:40:40,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=665032.6666666666, ans=0.125
2024-09-25 04:40:46,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.69 vs. limit=22.5
2024-09-25 04:40:47,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=665032.6666666666, ans=0.1
2024-09-25 04:40:48,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=665032.6666666666, ans=0.2
2024-09-25 04:41:14,999 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.35 vs. limit=12.0
2024-09-25 04:41:53,314 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.24 vs. limit=15.0
2024-09-25 04:41:58,859 INFO [train.py:1198] (1/4) Epoch 37, batch 2300, loss[loss=0.2054, ctc_loss=0.1329, cr_loss=0.3627, over 17168.00 frames. ], tot_loss[loss=0.1949, ctc_loss=0.1261, cr_loss=0.3438, over 3357585.28 frames. ], batch size: 45, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:42:02,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=665266.0, ans=0.0
2024-09-25 04:42:25,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=665312.6666666666, ans=0.125
2024-09-25 04:42:51,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=665406.0, ans=0.125
2024-09-25 04:43:07,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=665406.0, ans=0.125
2024-09-25 04:43:21,780 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.297e+02 1.364e+02 1.435e+02 2.216e+02, threshold=2.727e+02, percent-clipped=0.0
2024-09-25 04:43:26,538 INFO [train.py:1198] (1/4) Epoch 37, batch 2350, loss[loss=0.2116, ctc_loss=0.1379, cr_loss=0.3686, over 17068.00 frames. ], tot_loss[loss=0.1946, ctc_loss=0.1259, cr_loss=0.3436, over 3357559.78 frames. ], batch size: 56, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:43:37,124 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.26 vs. limit=15.0
2024-09-25 04:43:52,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=665546.0, ans=0.1
2024-09-25 04:43:54,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=665546.0, ans=0.125
2024-09-25 04:44:02,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=665592.6666666666, ans=0.0
2024-09-25 04:44:42,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=665686.0, ans=0.1
2024-09-25 04:44:46,596 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.26 vs. limit=15.0
2024-09-25 04:44:48,923 INFO [train.py:1198] (1/4) Epoch 37, batch 2400, loss[loss=0.2061, ctc_loss=0.1336, cr_loss=0.3625, over 16914.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1262, cr_loss=0.3439, over 3360282.71 frames. ], batch size: 58, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:45:25,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=665826.0, ans=0.1
2024-09-25 04:45:59,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=665919.3333333334, ans=0.1
2024-09-25 04:46:03,759 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.277e+02 1.353e+02 1.451e+02 1.749e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-25 04:46:08,611 INFO [train.py:1198] (1/4) Epoch 37, batch 2450, loss[loss=0.1666, ctc_loss=0.1057, cr_loss=0.3045, over 17196.00 frames. ], tot_loss[loss=0.1946, ctc_loss=0.1259, cr_loss=0.3434, over 3357996.17 frames. ], batch size: 41, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:46:19,116 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.30 vs. limit=22.5
2024-09-25 04:46:34,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=666012.6666666666, ans=0.125
2024-09-25 04:46:34,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=666012.6666666666, ans=0.0
2024-09-25 04:46:45,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.09 vs. limit=6.0
2024-09-25 04:47:03,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=666106.0, ans=0.2
2024-09-25 04:47:10,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=666106.0, ans=0.0
2024-09-25 04:47:26,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=666152.6666666666, ans=0.025
2024-09-25 04:47:28,740 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.97 vs. limit=22.5
2024-09-25 04:47:30,969 INFO [train.py:1198] (1/4) Epoch 37, batch 2500, loss[loss=0.1826, ctc_loss=0.1187, cr_loss=0.3196, over 17032.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1251, cr_loss=0.3426, over 3363009.05 frames. ], batch size: 44, lr: 3.21e-03, grad_scale: 32.0
2024-09-25 04:47:42,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=666199.3333333334, ans=0.025
2024-09-25 04:47:49,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=666246.0, ans=0.125
2024-09-25 04:47:52,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=666246.0, ans=0.125
2024-09-25 04:48:08,757 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=666292.6666666666, ans=0.0
2024-09-25 04:48:27,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=666339.3333333334, ans=0.2
2024-09-25 04:48:33,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.87 vs. limit=12.0
2024-09-25 04:48:39,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=666386.0, ans=0.125
2024-09-25 04:48:41,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.85 vs. limit=15.0
2024-09-25 04:48:53,436 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.305e+02 1.391e+02 1.498e+02 2.183e+02, threshold=2.782e+02, percent-clipped=0.0
2024-09-25 04:48:53,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=666386.0, ans=0.1
2024-09-25 04:48:56,604 INFO [train.py:1198] (1/4) Epoch 37, batch 2550, loss[loss=0.2092, ctc_loss=0.1367, cr_loss=0.3625, over 17268.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1255, cr_loss=0.3435, over 3367828.20 frames. ], batch size: 44, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:49:00,457 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.02 vs. limit=6.0
2024-09-25 04:49:18,066 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.27 vs. limit=12.0
2024-09-25 04:49:33,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=666526.0, ans=0.1
2024-09-25 04:49:49,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=666572.6666666666, ans=0.2
2024-09-25 04:49:56,127 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.01 vs. limit=12.0
2024-09-25 04:49:57,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=666572.6666666666, ans=0.0
2024-09-25 04:50:19,600 INFO [train.py:1198] (1/4) Epoch 37, batch 2600, loss[loss=0.1872, ctc_loss=0.1228, cr_loss=0.3223, over 17308.00 frames. ], tot_loss[loss=0.1943, ctc_loss=0.1255, cr_loss=0.3441, over 3373291.03 frames. ], batch size: 51, lr: 3.21e-03, grad_scale: 16.0
2024-09-25 04:50:26,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=666666.0, ans=0.125
2024-09-25 04:50:26,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.42 vs. limit=22.5
2024-09-25 04:50:48,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=666712.6666666666, ans=0.125
2024-09-25 04:51:06,601 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.75 vs. limit=15.0
2024-09-25 04:51:08,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.07 vs. limit=15.0
2024-09-25 04:51:09,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=666806.0, ans=0.125
2024-09-25 04:51:36,207 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.272e+02 1.334e+02 1.506e+02 4.167e+02, threshold=2.667e+02, percent-clipped=1.0
2024-09-25 04:51:39,401 INFO [train.py:1198] (1/4) Epoch 37, batch 2650, loss[loss=0.2273, ctc_loss=0.1502, cr_loss=0.3853, over 17138.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1247, cr_loss=0.3418, over 3366699.55 frames. ], batch size: 48, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 04:51:47,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=666899.3333333334, ans=0.125
2024-09-25 04:51:47,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=666899.3333333334, ans=0.0
2024-09-25 04:51:51,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=666899.3333333334, ans=0.2
2024-09-25 04:51:51,432 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.89 vs. limit=6.0
2024-09-25 04:51:55,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=666946.0, ans=0.035
2024-09-25 04:51:58,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=666946.0, ans=0.0
2024-09-25 04:52:57,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=667086.0, ans=0.1
2024-09-25 04:53:06,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=667132.6666666666, ans=0.125
2024-09-25 04:53:07,639 INFO [train.py:1198] (1/4) Epoch 37, batch 2700, loss[loss=0.1882, ctc_loss=0.1218, cr_loss=0.332, over 17160.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1239, cr_loss=0.3404, over 3369611.22 frames. ], batch size: 45, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 04:53:20,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=667132.6666666666, ans=0.2
2024-09-25 04:53:33,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=667179.3333333334, ans=0.0
2024-09-25 04:53:58,290 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.79 vs. limit=10.0
2024-09-25 04:54:02,521 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.36 vs. limit=10.0
2024-09-25 04:54:27,064 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.290e+02 1.345e+02 1.477e+02 2.693e+02, threshold=2.691e+02, percent-clipped=1.0
2024-09-25 04:54:30,303 INFO [train.py:1198] (1/4) Epoch 37, batch 2750, loss[loss=0.2101, ctc_loss=0.1351, cr_loss=0.3746, over 17235.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1243, cr_loss=0.3418, over 3373716.99 frames. ], batch size: 50, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 04:54:54,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=667412.6666666666, ans=0.0
2024-09-25 04:55:28,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=667506.0, ans=0.125
2024-09-25 04:55:34,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=667552.6666666666, ans=0.0
2024-09-25 04:55:42,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=667552.6666666666, ans=0.125
2024-09-25 04:55:44,888 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.36 vs. limit=10.0
2024-09-25 04:55:50,322 INFO [train.py:1198] (1/4) Epoch 37, batch 2800, loss[loss=0.1716, ctc_loss=0.1096, cr_loss=0.31, over 16945.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1243, cr_loss=0.3419, over 3377632.77 frames. ], batch size: 42, lr: 3.20e-03, grad_scale: 32.0
2024-09-25 04:55:57,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=667599.3333333334, ans=0.0
2024-09-25 04:56:05,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=667646.0, ans=0.0
2024-09-25 04:56:09,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=667646.0, ans=0.0
2024-09-25 04:56:26,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=667692.6666666666, ans=0.2
2024-09-25 04:56:27,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.88 vs. limit=22.5
2024-09-25 04:56:39,573 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.62 vs. limit=15.0
2024-09-25 04:56:47,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=667739.3333333334, ans=0.125
2024-09-25 04:56:51,041 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.84 vs. limit=15.0
2024-09-25 04:57:10,009 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.310e+02 1.386e+02 1.476e+02 1.796e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 04:57:13,305 INFO [train.py:1198] (1/4) Epoch 37, batch 2850, loss[loss=0.1787, ctc_loss=0.1154, cr_loss=0.3163, over 17294.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1237, cr_loss=0.3408, over 3382473.49 frames. ], batch size: 46, lr: 3.20e-03, grad_scale: 32.0
2024-09-25 04:57:13,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=667832.6666666666, ans=0.0
2024-09-25 04:57:20,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=667832.6666666666, ans=0.1
2024-09-25 04:57:32,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=667879.3333333334, ans=0.125
2024-09-25 04:57:57,614 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 04:58:01,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=667926.0, ans=0.125
2024-09-25 04:58:38,163 INFO [train.py:1198] (1/4) Epoch 37, batch 2900, loss[loss=0.2259, ctc_loss=0.1483, cr_loss=0.3881, over 16535.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.1231, cr_loss=0.3392, over 3373158.93 frames. ], batch size: 66, lr: 3.20e-03, grad_scale: 32.0
2024-09-25 04:58:42,130 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.90 vs. limit=15.0
2024-09-25 04:58:57,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1.whitening_limit, batch_count=668112.6666666666, ans=10.0
2024-09-25 04:59:00,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=668112.6666666666, ans=0.0
2024-09-25 04:59:41,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=668206.0, ans=0.0
2024-09-25 04:59:51,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=668252.6666666666, ans=0.1
2024-09-25 04:59:57,885 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.277e+02 1.411e+02 1.557e+02 2.099e+02, threshold=2.822e+02, percent-clipped=0.0
2024-09-25 04:59:58,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=668252.6666666666, ans=0.1
2024-09-25 05:00:01,063 INFO [train.py:1198] (1/4) Epoch 37, batch 2950, loss[loss=0.2102, ctc_loss=0.1377, cr_loss=0.3624, over 17347.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1238, cr_loss=0.3402, over 3359544.10 frames. ], batch size: 52, lr: 3.20e-03, grad_scale: 32.0
2024-09-25 05:00:03,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=668299.3333333334, ans=0.125
2024-09-25 05:00:10,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=668299.3333333334, ans=0.125
2024-09-25 05:00:22,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=668346.0, ans=0.125
2024-09-25 05:00:27,577 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.69 vs. limit=12.0
2024-09-25 05:00:44,942 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.43 vs. limit=22.5
2024-09-25 05:00:45,093 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.41 vs. limit=15.0
2024-09-25 05:00:57,542 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.44 vs. limit=15.0
2024-09-25 05:01:09,966 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.57 vs. limit=15.0
2024-09-25 05:01:20,184 INFO [train.py:1198] (1/4) Epoch 37, batch 3000, loss[loss=0.171, ctc_loss=0.1082, cr_loss=0.3139, over 17092.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1247, cr_loss=0.3414, over 3340402.75 frames. ], batch size: 43, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:01:20,184 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 05:01:35,774 INFO [train.py:1230] (1/4) Epoch 37, validation: loss=0.03526, ctc_loss=0.03526, cr_loss=1.039e-14, over 944034.00 frames. 
2024-09-25 05:01:35,775 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 05:01:39,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=668532.6666666666, ans=0.125
2024-09-25 05:01:47,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=668532.6666666666, ans=0.0
2024-09-25 05:01:53,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=668579.3333333334, ans=0.125
2024-09-25 05:02:24,805 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.43 vs. limit=15.0
2024-09-25 05:02:54,649 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.290e+02 1.381e+02 1.487e+02 2.036e+02, threshold=2.762e+02, percent-clipped=0.0
2024-09-25 05:02:54,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=668766.0, ans=0.2
2024-09-25 05:02:56,257 INFO [train.py:1198] (1/4) Epoch 37, batch 3050, loss[loss=0.1831, ctc_loss=0.1171, cr_loss=0.3298, over 16901.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1244, cr_loss=0.3412, over 3346128.50 frames. ], batch size: 58, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:03:14,472 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.14 vs. limit=15.0
2024-09-25 05:04:02,575 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.92 vs. limit=15.0
2024-09-25 05:04:09,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=668952.6666666666, ans=0.0
2024-09-25 05:04:12,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=668999.3333333334, ans=0.1
2024-09-25 05:04:14,260 INFO [train.py:1198] (1/4) Epoch 37, batch 3100, loss[loss=0.1756, ctc_loss=0.1121, cr_loss=0.3178, over 17221.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1237, cr_loss=0.3398, over 3356574.09 frames. ], batch size: 41, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:04:14,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=668999.3333333334, ans=0.0
2024-09-25 05:04:28,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=669046.0, ans=0.0
2024-09-25 05:04:38,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=669046.0, ans=0.2
2024-09-25 05:05:00,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=669092.6666666666, ans=0.0
2024-09-25 05:05:08,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=669139.3333333334, ans=0.125
2024-09-25 05:05:11,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=669139.3333333334, ans=0.04949747468305833
2024-09-25 05:05:21,072 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:05:29,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=669186.0, ans=0.125
2024-09-25 05:05:34,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=669186.0, ans=0.025
2024-09-25 05:05:34,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=669186.0, ans=0.0
2024-09-25 05:05:35,823 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.257e+02 1.346e+02 1.447e+02 1.997e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-25 05:05:37,418 INFO [train.py:1198] (1/4) Epoch 37, batch 3150, loss[loss=0.1839, ctc_loss=0.1191, cr_loss=0.3238, over 17163.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1238, cr_loss=0.34, over 3362164.17 frames. ], batch size: 45, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:05:37,659 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:05:40,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=669232.6666666666, ans=0.125
2024-09-25 05:06:12,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=669326.0, ans=0.125
2024-09-25 05:06:33,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=669372.6666666666, ans=0.125
2024-09-25 05:06:54,645 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.65 vs. limit=15.0
2024-09-25 05:06:55,583 INFO [train.py:1198] (1/4) Epoch 37, batch 3200, loss[loss=0.1996, ctc_loss=0.1279, cr_loss=0.3585, over 17154.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1246, cr_loss=0.3411, over 3370300.22 frames. ], batch size: 48, lr: 3.20e-03, grad_scale: 32.0
2024-09-25 05:06:57,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=669466.0, ans=0.1
2024-09-25 05:07:06,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=669466.0, ans=0.125
2024-09-25 05:07:15,223 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.42 vs. limit=15.0
2024-09-25 05:07:40,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=669606.0, ans=0.125
2024-09-25 05:07:47,498 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.94 vs. limit=6.0
2024-09-25 05:08:04,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=669652.6666666666, ans=0.5
2024-09-25 05:08:15,376 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.285e+02 1.355e+02 1.464e+02 2.821e+02, threshold=2.709e+02, percent-clipped=1.0
2024-09-25 05:08:15,400 INFO [train.py:1198] (1/4) Epoch 37, batch 3250, loss[loss=0.2045, ctc_loss=0.1339, cr_loss=0.3533, over 17260.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1245, cr_loss=0.3414, over 3369516.21 frames. ], batch size: 44, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:08:24,142 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.54 vs. limit=10.0
2024-09-25 05:08:46,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=669792.6666666666, ans=0.0
2024-09-25 05:09:01,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=669839.3333333334, ans=0.125
2024-09-25 05:09:05,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=669839.3333333334, ans=0.0
2024-09-25 05:09:22,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=669886.0, ans=0.2
2024-09-25 05:09:29,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=669886.0, ans=0.0
2024-09-25 05:09:33,486 INFO [train.py:1198] (1/4) Epoch 37, batch 3300, loss[loss=0.2039, ctc_loss=0.1303, cr_loss=0.3677, over 17024.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.125, cr_loss=0.3426, over 3361843.90 frames. ], batch size: 52, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:10:13,143 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:10:26,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=670072.6666666666, ans=0.1
2024-09-25 05:10:48,081 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.03 vs. limit=15.0
2024-09-25 05:10:49,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=670119.3333333334, ans=0.125
2024-09-25 05:10:51,976 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.300e+02 1.355e+02 1.430e+02 2.152e+02, threshold=2.710e+02, percent-clipped=0.0
2024-09-25 05:10:52,001 INFO [train.py:1198] (1/4) Epoch 37, batch 3350, loss[loss=0.23, ctc_loss=0.1537, cr_loss=0.3812, over 16992.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1248, cr_loss=0.3414, over 3352658.45 frames. ], batch size: 53, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:10:56,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=670166.0, ans=0.125
2024-09-25 05:11:33,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=670259.3333333334, ans=0.0
2024-09-25 05:11:34,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=670259.3333333334, ans=0.1
2024-09-25 05:11:40,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=670306.0, ans=0.125
2024-09-25 05:12:01,421 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.23 vs. limit=15.0
2024-09-25 05:12:09,881 INFO [train.py:1198] (1/4) Epoch 37, batch 3400, loss[loss=0.1874, ctc_loss=0.1231, cr_loss=0.3214, over 17083.00 frames. ], tot_loss[loss=0.194, ctc_loss=0.1254, cr_loss=0.3428, over 3362032.37 frames. ], batch size: 46, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:12:13,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=670399.3333333334, ans=0.0
2024-09-25 05:12:13,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=670399.3333333334, ans=0.125
2024-09-25 05:12:32,698 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.89 vs. limit=22.5
2024-09-25 05:12:41,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=670492.6666666666, ans=0.0
2024-09-25 05:12:49,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=670492.6666666666, ans=0.0
2024-09-25 05:13:28,360 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.294e+02 1.379e+02 1.549e+02 2.225e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-25 05:13:28,385 INFO [train.py:1198] (1/4) Epoch 37, batch 3450, loss[loss=0.2071, ctc_loss=0.1345, cr_loss=0.3629, over 17054.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1252, cr_loss=0.3427, over 3363618.13 frames. ], batch size: 56, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:14:01,273 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.74 vs. limit=15.0
2024-09-25 05:14:05,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=670726.0, ans=0.0
2024-09-25 05:14:33,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=670819.3333333334, ans=0.0
2024-09-25 05:14:46,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=670819.3333333334, ans=0.0
2024-09-25 05:14:47,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer_ff2.min_abs, batch_count=670866.0, ans=0.1
2024-09-25 05:14:49,369 INFO [train.py:1198] (1/4) Epoch 37, batch 3500, loss[loss=0.1968, ctc_loss=0.1276, cr_loss=0.346, over 17314.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1252, cr_loss=0.3429, over 3369146.29 frames. ], batch size: 49, lr: 3.20e-03, grad_scale: 16.0
2024-09-25 05:15:01,509 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.82 vs. limit=12.0
2024-09-25 05:15:12,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=670912.6666666666, ans=0.1
2024-09-25 05:15:31,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=670959.3333333334, ans=0.125
2024-09-25 05:15:37,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=670959.3333333334, ans=0.125
2024-09-25 05:15:37,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=670959.3333333334, ans=0.125
2024-09-25 05:15:53,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=671006.0, ans=0.125
2024-09-25 05:16:03,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=671052.6666666666, ans=0.04949747468305833
2024-09-25 05:16:11,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=671099.3333333334, ans=0.1
2024-09-25 05:16:12,280 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.290e+02 1.375e+02 1.489e+02 2.034e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-25 05:16:12,305 INFO [train.py:1198] (1/4) Epoch 37, batch 3550, loss[loss=0.1971, ctc_loss=0.1292, cr_loss=0.3396, over 17224.00 frames. ], tot_loss[loss=0.1938, ctc_loss=0.1252, cr_loss=0.343, over 3373812.59 frames. ], batch size: 55, lr: 3.19e-03, grad_scale: 16.0
2024-09-25 05:16:16,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.28 vs. limit=12.0
2024-09-25 05:16:25,466 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.76 vs. limit=15.0
2024-09-25 05:16:42,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=671192.6666666666, ans=0.2
2024-09-25 05:16:49,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=671192.6666666666, ans=0.125
2024-09-25 05:16:57,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=671239.3333333334, ans=0.0
2024-09-25 05:17:28,014 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.49 vs. limit=15.0
2024-09-25 05:17:30,028 INFO [train.py:1198] (1/4) Epoch 37, batch 3600, loss[loss=0.2044, ctc_loss=0.1303, cr_loss=0.3706, over 17186.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1249, cr_loss=0.3427, over 3371842.34 frames. ], batch size: 55, lr: 3.19e-03, grad_scale: 32.0
2024-09-25 05:17:33,491 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:17:39,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=671332.6666666666, ans=0.125
2024-09-25 05:17:51,374 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.72 vs. limit=6.0
2024-09-25 05:18:03,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=671426.0, ans=0.0
2024-09-25 05:18:15,180 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.50 vs. limit=6.0
2024-09-25 05:18:19,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.84 vs. limit=15.0
2024-09-25 05:18:50,878 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.287e+02 1.354e+02 1.480e+02 1.820e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-25 05:18:50,902 INFO [train.py:1198] (1/4) Epoch 37, batch 3650, loss[loss=0.2093, ctc_loss=0.137, cr_loss=0.3612, over 17024.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1251, cr_loss=0.3428, over 3375715.60 frames. ], batch size: 56, lr: 3.19e-03, grad_scale: 32.0
2024-09-25 05:18:51,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.22 vs. limit=15.0
2024-09-25 05:19:18,159 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.63 vs. limit=15.0
2024-09-25 05:19:35,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=671659.3333333334, ans=0.1
2024-09-25 05:19:50,003 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.57 vs. limit=6.0
2024-09-25 05:19:54,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=671752.6666666666, ans=0.0
2024-09-25 05:20:08,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=671799.3333333334, ans=0.1
2024-09-25 05:20:09,452 INFO [train.py:1198] (1/4) Epoch 37, batch 3700, loss[loss=0.1925, ctc_loss=0.1228, cr_loss=0.3484, over 17220.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.125, cr_loss=0.3423, over 3372761.09 frames. ], batch size: 47, lr: 3.19e-03, grad_scale: 32.0
2024-09-25 05:20:14,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=671799.3333333334, ans=0.035
2024-09-25 05:20:50,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=671892.6666666666, ans=0.125
2024-09-25 05:20:59,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=671939.3333333334, ans=0.2
2024-09-25 05:21:07,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=671939.3333333334, ans=0.1
2024-09-25 05:21:10,246 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:21:11,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=671986.0, ans=0.125
2024-09-25 05:21:29,091 INFO [train.py:1198] (1/4) Epoch 37, batch 3750, loss[loss=0.206, ctc_loss=0.1307, cr_loss=0.3765, over 17308.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1243, cr_loss=0.341, over 3358703.83 frames. ], batch size: 51, lr: 3.19e-03, grad_scale: 16.0
2024-09-25 05:21:30,592 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.295e+02 1.401e+02 1.510e+02 2.261e+02, threshold=2.801e+02, percent-clipped=0.0
2024-09-25 05:21:38,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=672032.6666666666, ans=0.125
2024-09-25 05:21:48,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=672079.3333333334, ans=0.2
2024-09-25 05:22:28,395 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.10 vs. limit=12.0
2024-09-25 05:22:38,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=672219.3333333334, ans=0.125
2024-09-25 05:22:41,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=672219.3333333334, ans=0.125
2024-09-25 05:22:46,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=672266.0, ans=0.2
2024-09-25 05:22:47,832 INFO [train.py:1198] (1/4) Epoch 37, batch 3800, loss[loss=0.1763, ctc_loss=0.1141, cr_loss=0.3111, over 17173.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1236, cr_loss=0.3396, over 3351228.12 frames. ], batch size: 45, lr: 3.19e-03, grad_scale: 16.0
2024-09-25 05:22:58,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=672266.0, ans=0.125
2024-09-25 05:24:06,808 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.76 vs. limit=15.0
2024-09-25 05:24:07,328 INFO [train.py:1198] (1/4) Epoch 37, batch 3850, loss[loss=0.2318, ctc_loss=0.1541, cr_loss=0.3888, over 14852.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1234, cr_loss=0.3386, over 3317308.11 frames. ], batch size: 89, lr: 3.19e-03, grad_scale: 16.0
2024-09-25 05:24:08,856 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.255e+02 1.356e+02 1.479e+02 3.474e+02, threshold=2.713e+02, percent-clipped=1.0
2024-09-25 05:24:12,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=672499.3333333334, ans=0.125
2024-09-25 05:24:30,989 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:24:31,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=672546.0, ans=0.0
2024-09-25 05:24:34,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=672546.0, ans=0.1
2024-09-25 05:24:41,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=672592.6666666666, ans=0.125
2024-09-25 05:24:52,499 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.21 vs. limit=15.0
2024-09-25 05:24:55,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=672639.3333333334, ans=0.125
2024-09-25 05:25:04,100 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=672639.3333333334, ans=10.0
2024-09-25 05:25:10,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=672686.0, ans=0.0
2024-09-25 05:25:10,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer_na.min_abs, batch_count=672686.0, ans=0.02
2024-09-25 05:25:10,536 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.71 vs. limit=22.5
2024-09-25 05:26:04,404 INFO [train.py:1198] (1/4) Epoch 38, batch 0, loss[loss=0.189, ctc_loss=0.1206, cr_loss=0.3423, over 17296.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1206, cr_loss=0.3423, over 17296.00 frames. ], batch size: 46, lr: 3.15e-03, grad_scale: 32.0
2024-09-25 05:26:04,405 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 05:26:20,208 INFO [train.py:1230] (1/4) Epoch 38, validation: loss=0.03515, ctc_loss=0.03515, cr_loss=9.44e-15, over 944034.00 frames. 
2024-09-25 05:26:20,208 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 05:26:24,118 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=11.07 vs. limit=12.0
2024-09-25 05:26:27,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=672714.0, ans=0.125
2024-09-25 05:26:35,670 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=5.26 vs. limit=15.0
2024-09-25 05:26:41,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=672760.6666666666, ans=0.0
2024-09-25 05:26:46,401 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=672760.6666666666, ans=0.0
2024-09-25 05:26:54,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=672807.3333333334, ans=0.125
2024-09-25 05:27:16,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=672854.0, ans=0.1
2024-09-25 05:27:40,674 INFO [train.py:1198] (1/4) Epoch 38, batch 50, loss[loss=0.2159, ctc_loss=0.1405, cr_loss=0.3769, over 17230.00 frames. ], tot_loss[loss=0.1978, ctc_loss=0.1281, cr_loss=0.3488, over 748924.00 frames. ], batch size: 55, lr: 3.15e-03, grad_scale: 16.0
2024-09-25 05:27:44,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=672947.3333333334, ans=0.125
2024-09-25 05:27:50,497 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.374e+02 1.537e+02 1.726e+02 2.147e+02, threshold=3.075e+02, percent-clipped=0.0
2024-09-25 05:27:52,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=672947.3333333334, ans=0.2
2024-09-25 05:28:20,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=673040.6666666666, ans=0.1
2024-09-25 05:28:23,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=673040.6666666666, ans=0.0
2024-09-25 05:28:28,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=673040.6666666666, ans=0.125
2024-09-25 05:28:36,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=673087.3333333334, ans=0.2
2024-09-25 05:28:46,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=673134.0, ans=0.0
2024-09-25 05:29:03,636 INFO [train.py:1198] (1/4) Epoch 38, batch 100, loss[loss=0.1633, ctc_loss=0.1035, cr_loss=0.2988, over 17156.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1249, cr_loss=0.3423, over 1333764.04 frames. ], batch size: 45, lr: 3.15e-03, grad_scale: 8.0
2024-09-25 05:29:19,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=673227.3333333334, ans=0.125
2024-09-25 05:29:23,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=673227.3333333334, ans=0.0
2024-09-25 05:29:53,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=673274.0, ans=0.025
2024-09-25 05:30:16,123 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.39 vs. limit=15.0
2024-09-25 05:30:31,018 INFO [train.py:1198] (1/4) Epoch 38, batch 150, loss[loss=0.1909, ctc_loss=0.1234, cr_loss=0.3373, over 17348.00 frames. ], tot_loss[loss=0.1939, ctc_loss=0.1254, cr_loss=0.3428, over 1792629.59 frames. ], batch size: 48, lr: 3.15e-03, grad_scale: 8.0
2024-09-25 05:30:42,271 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.262e+02 1.329e+02 1.427e+02 1.998e+02, threshold=2.657e+02, percent-clipped=0.0
2024-09-25 05:31:38,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=673600.6666666666, ans=0.5
2024-09-25 05:31:40,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=673600.6666666666, ans=0.2
2024-09-25 05:31:51,203 INFO [train.py:1198] (1/4) Epoch 38, batch 200, loss[loss=0.2108, ctc_loss=0.1363, cr_loss=0.3727, over 16987.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1266, cr_loss=0.346, over 2143093.84 frames. ], batch size: 53, lr: 3.15e-03, grad_scale: 8.0
2024-09-25 05:31:53,089 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:32:34,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=673740.6666666666, ans=0.0
2024-09-25 05:32:39,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=673787.3333333334, ans=0.2
2024-09-25 05:32:43,263 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.75 vs. limit=15.0
2024-09-25 05:33:05,657 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=8.65 vs. limit=22.5
2024-09-25 05:33:13,725 INFO [train.py:1198] (1/4) Epoch 38, batch 250, loss[loss=0.2048, ctc_loss=0.1317, cr_loss=0.3656, over 17064.00 frames. ], tot_loss[loss=0.1964, ctc_loss=0.127, cr_loss=0.3467, over 2405226.39 frames. ], batch size: 46, lr: 3.15e-03, grad_scale: 8.0
2024-09-25 05:33:17,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=673880.6666666666, ans=0.2
2024-09-25 05:33:18,809 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:33:24,710 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.262e+02 1.325e+02 1.385e+02 3.377e+02, threshold=2.651e+02, percent-clipped=1.0
2024-09-25 05:34:04,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=674020.6666666666, ans=0.125
2024-09-25 05:34:06,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=674020.6666666666, ans=0.2
2024-09-25 05:34:21,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=674067.3333333334, ans=0.0
2024-09-25 05:34:35,576 INFO [train.py:1198] (1/4) Epoch 38, batch 300, loss[loss=0.1588, ctc_loss=0.09863, cr_loss=0.3011, over 17272.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1269, cr_loss=0.3467, over 2615575.66 frames. ], batch size: 42, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:34:38,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=674114.0, ans=0.125
2024-09-25 05:34:51,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=674114.0, ans=0.1
2024-09-25 05:34:56,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=674160.6666666666, ans=0.2
2024-09-25 05:35:07,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=674160.6666666666, ans=0.0
2024-09-25 05:35:19,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=674207.3333333334, ans=0.0
2024-09-25 05:35:28,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.68 vs. limit=15.0
2024-09-25 05:35:37,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=674254.0, ans=0.0
2024-09-25 05:35:48,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=674300.6666666666, ans=0.2
2024-09-25 05:36:01,135 INFO [train.py:1198] (1/4) Epoch 38, batch 350, loss[loss=0.1914, ctc_loss=0.1231, cr_loss=0.3416, over 17247.00 frames. ], tot_loss[loss=0.1958, ctc_loss=0.1266, cr_loss=0.3462, over 2785843.39 frames. ], batch size: 44, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:36:01,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=674347.3333333334, ans=0.125
2024-09-25 05:36:01,847 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.35 vs. limit=12.0
2024-09-25 05:36:12,222 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.293e+02 1.371e+02 1.501e+02 2.181e+02, threshold=2.742e+02, percent-clipped=0.0
2024-09-25 05:36:21,369 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.78 vs. limit=12.0
2024-09-25 05:36:30,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=674394.0, ans=0.1
2024-09-25 05:37:17,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=674534.0, ans=0.0
2024-09-25 05:37:20,498 INFO [train.py:1198] (1/4) Epoch 38, batch 400, loss[loss=0.2181, ctc_loss=0.1411, cr_loss=0.385, over 17233.00 frames. ], tot_loss[loss=0.1959, ctc_loss=0.1266, cr_loss=0.3466, over 2919068.48 frames. ], batch size: 55, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:37:23,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=674580.6666666666, ans=0.1
2024-09-25 05:37:28,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=674580.6666666666, ans=0.1
2024-09-25 05:37:30,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=674580.6666666666, ans=0.125
2024-09-25 05:37:45,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.max_abs, batch_count=674627.3333333334, ans=10.0
2024-09-25 05:38:10,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=674720.6666666666, ans=0.125
2024-09-25 05:38:16,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=674720.6666666666, ans=6.0
2024-09-25 05:38:17,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.13 vs. limit=12.0
2024-09-25 05:38:20,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=674720.6666666666, ans=0.0
2024-09-25 05:38:31,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=674767.3333333334, ans=0.025
2024-09-25 05:38:33,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=674767.3333333334, ans=0.2
2024-09-25 05:38:35,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=11.14 vs. limit=15.0
2024-09-25 05:38:42,900 INFO [train.py:1198] (1/4) Epoch 38, batch 450, loss[loss=0.2121, ctc_loss=0.1387, cr_loss=0.3669, over 17310.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1263, cr_loss=0.3456, over 3001052.11 frames. ], batch size: 49, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:38:52,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=674814.0, ans=0.0
2024-09-25 05:38:55,495 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.292e+02 1.363e+02 1.447e+02 2.119e+02, threshold=2.727e+02, percent-clipped=0.0
2024-09-25 05:39:10,641 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.74 vs. limit=6.0
2024-09-25 05:39:13,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=674907.3333333334, ans=0.0
2024-09-25 05:39:30,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=674907.3333333334, ans=0.0
2024-09-25 05:39:38,924 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:39:40,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=674954.0, ans=0.125
2024-09-25 05:39:40,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.04 vs. limit=12.0
2024-09-25 05:39:52,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=675000.6666666666, ans=0.2
2024-09-25 05:40:11,290 INFO [train.py:1198] (1/4) Epoch 38, batch 500, loss[loss=0.1562, ctc_loss=0.09802, cr_loss=0.2911, over 17104.00 frames. ], tot_loss[loss=0.1955, ctc_loss=0.1263, cr_loss=0.3457, over 3084935.95 frames. ], batch size: 40, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:40:22,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=675047.3333333334, ans=0.1
2024-09-25 05:40:30,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=675094.0, ans=0.125
2024-09-25 05:40:45,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=675140.6666666666, ans=0.0
2024-09-25 05:40:46,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=675140.6666666666, ans=0.125
2024-09-25 05:41:05,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=675187.3333333334, ans=0.1
2024-09-25 05:41:19,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=675234.0, ans=0.2
2024-09-25 05:41:30,505 INFO [train.py:1198] (1/4) Epoch 38, batch 550, loss[loss=0.1789, ctc_loss=0.1136, cr_loss=0.3263, over 17000.00 frames. ], tot_loss[loss=0.1961, ctc_loss=0.1267, cr_loss=0.3469, over 3143546.24 frames. ], batch size: 51, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:41:30,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=675280.6666666666, ans=0.1
2024-09-25 05:41:40,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=675280.6666666666, ans=0.0
2024-09-25 05:41:43,253 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.284e+02 1.364e+02 1.434e+02 1.794e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-25 05:41:51,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=675327.3333333334, ans=0.125
2024-09-25 05:42:26,593 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 05:42:33,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=675467.3333333334, ans=0.0
2024-09-25 05:42:45,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=675467.3333333334, ans=0.125
2024-09-25 05:42:50,161 INFO [train.py:1198] (1/4) Epoch 38, batch 600, loss[loss=0.1462, ctc_loss=0.09049, cr_loss=0.2787, over 17270.00 frames. ], tot_loss[loss=0.1953, ctc_loss=0.1261, cr_loss=0.3457, over 3192129.36 frames. ], batch size: 42, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:43:15,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=675560.6666666666, ans=0.2
2024-09-25 05:43:24,051 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.03 vs. limit=15.0
2024-09-25 05:43:38,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=675607.3333333334, ans=0.0
2024-09-25 05:43:42,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=675654.0, ans=0.125
2024-09-25 05:43:47,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=675654.0, ans=0.0
2024-09-25 05:44:13,109 INFO [train.py:1198] (1/4) Epoch 38, batch 650, loss[loss=0.1494, ctc_loss=0.09235, cr_loss=0.2851, over 17052.00 frames. ], tot_loss[loss=0.1944, ctc_loss=0.1255, cr_loss=0.3442, over 3229454.40 frames. ], batch size: 39, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:44:28,504 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.283e+02 1.368e+02 1.490e+02 2.037e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 05:44:30,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=675794.0, ans=0.125
2024-09-25 05:44:51,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.27 vs. limit=12.0
2024-09-25 05:45:34,219 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=675934.0, ans=0.0
2024-09-25 05:45:40,415 INFO [train.py:1198] (1/4) Epoch 38, batch 700, loss[loss=0.1855, ctc_loss=0.1175, cr_loss=0.3399, over 17215.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1255, cr_loss=0.3439, over 3261789.44 frames. ], batch size: 47, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:45:50,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=675980.6666666666, ans=0.125
2024-09-25 05:45:58,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=676027.3333333334, ans=0.125
2024-09-25 05:45:59,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=676027.3333333334, ans=10.0
2024-09-25 05:46:35,699 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.62 vs. limit=15.0
2024-09-25 05:46:54,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=676167.3333333334, ans=0.0
2024-09-25 05:46:54,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.98 vs. limit=15.0
2024-09-25 05:47:00,009 INFO [train.py:1198] (1/4) Epoch 38, batch 750, loss[loss=0.2039, ctc_loss=0.133, cr_loss=0.3543, over 17002.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1249, cr_loss=0.3423, over 3288800.73 frames. ], batch size: 51, lr: 3.14e-03, grad_scale: 8.0
2024-09-25 05:47:09,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=676214.0, ans=0.125
2024-09-25 05:47:12,449 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.277e+02 1.363e+02 1.416e+02 2.105e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-25 05:47:25,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1.whitening_limit, batch_count=676260.6666666666, ans=10.0
2024-09-25 05:47:26,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=676260.6666666666, ans=0.0
2024-09-25 05:47:47,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=676354.0, ans=0.125
2024-09-25 05:48:02,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=676354.0, ans=0.2
2024-09-25 05:48:04,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.96 vs. limit=6.0
2024-09-25 05:48:10,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=676400.6666666666, ans=0.125
2024-09-25 05:48:21,650 INFO [train.py:1198] (1/4) Epoch 38, batch 800, loss[loss=0.2312, ctc_loss=0.1513, cr_loss=0.3992, over 17214.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1249, cr_loss=0.3421, over 3300185.22 frames. ], batch size: 55, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:48:30,547 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.37 vs. limit=10.0
2024-09-25 05:48:31,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=676447.3333333334, ans=0.125
2024-09-25 05:49:27,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=676634.0, ans=0.0
2024-09-25 05:49:31,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=676634.0, ans=0.0
2024-09-25 05:49:37,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=676634.0, ans=0.025
2024-09-25 05:49:49,052 INFO [train.py:1198] (1/4) Epoch 38, batch 850, loss[loss=0.2646, ctc_loss=0.1787, cr_loss=0.4299, over 11723.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1255, cr_loss=0.3433, over 3313342.18 frames. ], batch size: 123, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:50:01,637 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.281e+02 1.360e+02 1.434e+02 2.186e+02, threshold=2.720e+02, percent-clipped=0.0
2024-09-25 05:50:19,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=676774.0, ans=0.125
2024-09-25 05:50:30,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=676774.0, ans=0.0
2024-09-25 05:50:32,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=676774.0, ans=0.0
2024-09-25 05:50:38,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=676820.6666666666, ans=0.125
2024-09-25 05:50:53,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=676867.3333333334, ans=0.125
2024-09-25 05:51:05,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=676867.3333333334, ans=0.0
2024-09-25 05:51:08,594 INFO [train.py:1198] (1/4) Epoch 38, batch 900, loss[loss=0.1654, ctc_loss=0.1044, cr_loss=0.3051, over 17118.00 frames. ], tot_loss[loss=0.194, ctc_loss=0.1254, cr_loss=0.3431, over 3332146.81 frames. ], batch size: 40, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:51:40,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.15 vs. limit=15.0
2024-09-25 05:51:49,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=677007.3333333334, ans=0.2
2024-09-25 05:51:50,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=677007.3333333334, ans=0.125
2024-09-25 05:52:13,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=677100.6666666666, ans=0.125
2024-09-25 05:52:29,101 INFO [train.py:1198] (1/4) Epoch 38, batch 950, loss[loss=0.2015, ctc_loss=0.1318, cr_loss=0.3483, over 17028.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1251, cr_loss=0.343, over 3330056.06 frames. ], batch size: 51, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:52:36,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.92 vs. limit=15.0
2024-09-25 05:52:41,914 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.327e+02 1.410e+02 1.537e+02 3.330e+02, threshold=2.819e+02, percent-clipped=2.0
2024-09-25 05:53:19,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=677287.3333333334, ans=0.1
2024-09-25 05:53:52,325 INFO [train.py:1198] (1/4) Epoch 38, batch 1000, loss[loss=0.1911, ctc_loss=0.1249, cr_loss=0.3312, over 17221.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1246, cr_loss=0.3419, over 3336909.73 frames. ], batch size: 55, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:53:54,605 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.40 vs. limit=15.0
2024-09-25 05:54:15,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=677427.3333333334, ans=0.125
2024-09-25 05:54:17,898 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.68 vs. limit=15.0
2024-09-25 05:54:39,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=677474.0, ans=0.125
2024-09-25 05:55:12,988 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.99 vs. limit=15.0
2024-09-25 05:55:20,559 INFO [train.py:1198] (1/4) Epoch 38, batch 1050, loss[loss=0.2145, ctc_loss=0.1418, cr_loss=0.3634, over 16999.00 frames. ], tot_loss[loss=0.1929, ctc_loss=0.1245, cr_loss=0.3418, over 3339037.78 frames. ], batch size: 56, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:55:33,503 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.267e+02 1.356e+02 1.451e+02 1.928e+02, threshold=2.711e+02, percent-clipped=0.0
2024-09-25 05:55:52,044 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.80 vs. limit=10.0
2024-09-25 05:55:56,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=677707.3333333334, ans=0.0
2024-09-25 05:56:40,343 INFO [train.py:1198] (1/4) Epoch 38, batch 1100, loss[loss=0.2118, ctc_loss=0.1369, cr_loss=0.3742, over 17211.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.1249, cr_loss=0.343, over 3347615.78 frames. ], batch size: 55, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:56:44,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.90 vs. limit=15.0
2024-09-25 05:56:45,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=677847.3333333334, ans=0.125
2024-09-25 05:56:50,526 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.60 vs. limit=22.5
2024-09-25 05:57:01,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=677894.0, ans=0.0
2024-09-25 05:57:25,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=677940.6666666666, ans=0.125
2024-09-25 05:57:27,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=677987.3333333334, ans=0.125
2024-09-25 05:57:28,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=677987.3333333334, ans=0.1
2024-09-25 05:57:33,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=677987.3333333334, ans=0.125
2024-09-25 05:57:47,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=678034.0, ans=0.125
2024-09-25 05:58:02,651 INFO [train.py:1198] (1/4) Epoch 38, batch 1150, loss[loss=0.1413, ctc_loss=0.0891, cr_loss=0.2609, over 17183.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1246, cr_loss=0.342, over 3355881.30 frames. ], batch size: 41, lr: 3.14e-03, grad_scale: 16.0
2024-09-25 05:58:02,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=678080.6666666666, ans=0.5
2024-09-25 05:58:15,128 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.254e+02 1.322e+02 1.438e+02 2.414e+02, threshold=2.644e+02, percent-clipped=0.0
2024-09-25 05:58:31,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=678127.3333333334, ans=0.0
2024-09-25 05:59:08,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=678267.3333333334, ans=0.125
2024-09-25 05:59:17,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=678267.3333333334, ans=0.2
2024-09-25 05:59:25,215 INFO [train.py:1198] (1/4) Epoch 38, batch 1200, loss[loss=0.1836, ctc_loss=0.1162, cr_loss=0.3372, over 17090.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3417, over 3359619.89 frames. ], batch size: 49, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 05:59:30,900 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.96 vs. limit=22.5
2024-09-25 05:59:57,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=678360.6666666666, ans=0.125
2024-09-25 06:00:05,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=678407.3333333334, ans=0.125
2024-09-25 06:00:15,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=678407.3333333334, ans=0.0
2024-09-25 06:00:31,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_abs, batch_count=678454.0, ans=0.5
2024-09-25 06:00:50,406 INFO [train.py:1198] (1/4) Epoch 38, batch 1250, loss[loss=0.1872, ctc_loss=0.1207, cr_loss=0.3326, over 17209.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1246, cr_loss=0.3429, over 3361277.95 frames. ], batch size: 47, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 06:00:52,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=678547.3333333334, ans=0.025
2024-09-25 06:00:57,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=678547.3333333334, ans=0.125
2024-09-25 06:01:03,955 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.29 vs. limit=15.0
2024-09-25 06:01:04,695 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.283e+02 1.378e+02 1.489e+02 1.932e+02, threshold=2.757e+02, percent-clipped=0.0
2024-09-25 06:01:16,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=678594.0, ans=0.1
2024-09-25 06:01:34,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=678640.6666666666, ans=0.125
2024-09-25 06:01:40,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=678687.3333333334, ans=0.0
2024-09-25 06:01:45,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=678687.3333333334, ans=0.0
2024-09-25 06:01:46,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=678687.3333333334, ans=0.025
2024-09-25 06:01:51,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=678687.3333333334, ans=0.2
2024-09-25 06:01:58,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=678734.0, ans=0.025
2024-09-25 06:02:11,137 INFO [train.py:1198] (1/4) Epoch 38, batch 1300, loss[loss=0.2107, ctc_loss=0.1381, cr_loss=0.3629, over 17005.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.125, cr_loss=0.3436, over 3363228.36 frames. ], batch size: 52, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 06:02:29,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=678827.3333333334, ans=0.07
2024-09-25 06:02:34,252 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.90 vs. limit=15.0
2024-09-25 06:03:10,467 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.68 vs. limit=12.0
2024-09-25 06:03:27,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=678967.3333333334, ans=0.125
2024-09-25 06:03:33,486 INFO [train.py:1198] (1/4) Epoch 38, batch 1350, loss[loss=0.1791, ctc_loss=0.1146, cr_loss=0.3226, over 17171.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1244, cr_loss=0.3417, over 3350127.29 frames. ], batch size: 45, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 06:03:44,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=4.07 vs. limit=15.0
2024-09-25 06:03:44,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=679014.0, ans=0.035
2024-09-25 06:03:47,755 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.262e+02 1.346e+02 1.430e+02 2.071e+02, threshold=2.693e+02, percent-clipped=0.0
2024-09-25 06:03:51,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=679060.6666666666, ans=0.125
2024-09-25 06:04:09,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=679107.3333333334, ans=0.125
2024-09-25 06:05:01,173 INFO [train.py:1198] (1/4) Epoch 38, batch 1400, loss[loss=0.1567, ctc_loss=0.1001, cr_loss=0.2831, over 17185.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1239, cr_loss=0.3402, over 3350938.78 frames. ], batch size: 41, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 06:05:12,798 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.49 vs. limit=15.0
2024-09-25 06:05:15,683 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=679294.0, ans=0.0
2024-09-25 06:05:19,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=679294.0, ans=0.125
2024-09-25 06:05:30,609 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.98 vs. limit=15.0
2024-09-25 06:05:31,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=679340.6666666666, ans=0.0
2024-09-25 06:05:49,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=679387.3333333334, ans=0.0
2024-09-25 06:06:03,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=679434.0, ans=0.2
2024-09-25 06:06:05,617 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.79 vs. limit=10.0
2024-09-25 06:06:07,441 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.38 vs. limit=15.0
2024-09-25 06:06:10,213 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.89 vs. limit=10.0
2024-09-25 06:06:21,082 INFO [train.py:1198] (1/4) Epoch 38, batch 1450, loss[loss=0.1961, ctc_loss=0.1249, cr_loss=0.3556, over 17209.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1234, cr_loss=0.3403, over 3357307.35 frames. ], batch size: 47, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 06:06:35,619 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.250e+02 1.326e+02 1.407e+02 2.354e+02, threshold=2.651e+02, percent-clipped=0.0
2024-09-25 06:06:37,524 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=679527.3333333334, ans=0.125
2024-09-25 06:06:45,332 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=679527.3333333334, ans=0.125
2024-09-25 06:06:48,997 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.66 vs. limit=10.0
2024-09-25 06:06:53,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=679574.0, ans=0.125
2024-09-25 06:07:09,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=679620.6666666666, ans=0.025
2024-09-25 06:07:36,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=679667.3333333334, ans=0.1
2024-09-25 06:07:41,100 INFO [train.py:1198] (1/4) Epoch 38, batch 1500, loss[loss=0.2046, ctc_loss=0.1338, cr_loss=0.3542, over 17306.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1237, cr_loss=0.3399, over 3337107.37 frames. ], batch size: 51, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 06:07:41,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=679714.0, ans=0.1
2024-09-25 06:07:52,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=679714.0, ans=0.0
2024-09-25 06:08:03,299 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=679760.6666666666, ans=0.0
2024-09-25 06:08:09,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=679760.6666666666, ans=0.125
2024-09-25 06:08:15,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=679807.3333333334, ans=0.2
2024-09-25 06:08:49,671 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 06:09:06,158 INFO [train.py:1198] (1/4) Epoch 38, batch 1550, loss[loss=0.1965, ctc_loss=0.1263, cr_loss=0.351, over 17147.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1236, cr_loss=0.3393, over 3331380.87 frames. ], batch size: 48, lr: 3.13e-03, grad_scale: 16.0
2024-09-25 06:09:20,508 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.263e+02 1.343e+02 1.440e+02 2.044e+02, threshold=2.685e+02, percent-clipped=0.0
2024-09-25 06:09:22,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=679994.0, ans=0.125
2024-09-25 06:09:31,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=679994.0, ans=0.0
2024-09-25 06:09:51,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=680040.6666666666, ans=0.07
2024-09-25 06:09:53,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=680040.6666666666, ans=0.0
2024-09-25 06:10:12,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=680087.3333333334, ans=0.125
2024-09-25 06:10:31,468 INFO [train.py:1198] (1/4) Epoch 38, batch 1600, loss[loss=0.2087, ctc_loss=0.1349, cr_loss=0.3693, over 17143.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1234, cr_loss=0.3389, over 3337673.25 frames. ], batch size: 48, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:10:36,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=680180.6666666666, ans=0.125
2024-09-25 06:10:59,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=680227.3333333334, ans=0.0
2024-09-25 06:11:13,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=680274.0, ans=0.0
2024-09-25 06:11:16,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=680274.0, ans=0.1
2024-09-25 06:11:24,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=680320.6666666666, ans=0.0
2024-09-25 06:11:51,472 INFO [train.py:1198] (1/4) Epoch 38, batch 1650, loss[loss=0.1784, ctc_loss=0.1122, cr_loss=0.3313, over 17080.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1237, cr_loss=0.3402, over 3344946.95 frames. ], batch size: 40, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:12:05,735 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.273e+02 1.346e+02 1.505e+02 2.146e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-25 06:12:15,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=680460.6666666666, ans=0.5
2024-09-25 06:13:02,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=680600.6666666666, ans=0.125
2024-09-25 06:13:13,539 INFO [train.py:1198] (1/4) Epoch 38, batch 1700, loss[loss=0.1784, ctc_loss=0.1134, cr_loss=0.3251, over 17110.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1244, cr_loss=0.3418, over 3348467.88 frames. ], batch size: 40, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:13:15,576 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 06:13:44,757 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.40 vs. limit=6.0
2024-09-25 06:13:55,634 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=7.93 vs. limit=15.0
2024-09-25 06:14:02,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=680787.3333333334, ans=0.1
2024-09-25 06:14:02,793 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 06:14:12,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=680787.3333333334, ans=0.125
2024-09-25 06:14:38,370 INFO [train.py:1198] (1/4) Epoch 38, batch 1750, loss[loss=0.207, ctc_loss=0.1349, cr_loss=0.3607, over 17351.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.1242, cr_loss=0.3409, over 3357404.78 frames. ], batch size: 48, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:14:54,713 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=7.64 vs. limit=15.0
2024-09-25 06:14:55,387 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.278e+02 1.373e+02 1.469e+02 4.120e+02, threshold=2.745e+02, percent-clipped=1.0
2024-09-25 06:15:18,756 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.18 vs. limit=22.5
2024-09-25 06:15:52,037 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.32 vs. limit=15.0
2024-09-25 06:16:01,066 INFO [train.py:1198] (1/4) Epoch 38, batch 1800, loss[loss=0.1964, ctc_loss=0.1244, cr_loss=0.36, over 17291.00 frames. ], tot_loss[loss=0.1929, ctc_loss=0.1245, cr_loss=0.3417, over 3359763.90 frames. ], batch size: 51, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:16:19,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=681160.6666666666, ans=0.1
2024-09-25 06:16:20,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=681160.6666666666, ans=0.125
2024-09-25 06:16:27,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=681160.6666666666, ans=0.125
2024-09-25 06:16:30,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=681160.6666666666, ans=0.125
2024-09-25 06:16:46,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=681207.3333333334, ans=0.025
2024-09-25 06:16:56,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=681254.0, ans=0.125
2024-09-25 06:17:21,940 INFO [train.py:1198] (1/4) Epoch 38, batch 1850, loss[loss=0.2074, ctc_loss=0.1317, cr_loss=0.3787, over 16667.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1237, cr_loss=0.3403, over 3367404.62 frames. ], batch size: 61, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:17:36,439 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.263e+02 1.331e+02 1.457e+02 2.352e+02, threshold=2.661e+02, percent-clipped=0.0
2024-09-25 06:17:38,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=681394.0, ans=0.125
2024-09-25 06:17:46,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=681394.0, ans=0.0
2024-09-25 06:17:58,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=681440.6666666666, ans=0.5
2024-09-25 06:18:14,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=681487.3333333334, ans=0.025
2024-09-25 06:18:27,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=681534.0, ans=0.125
2024-09-25 06:18:32,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=681534.0, ans=0.0
2024-09-25 06:18:36,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=681534.0, ans=0.1
2024-09-25 06:18:44,347 INFO [train.py:1198] (1/4) Epoch 38, batch 1900, loss[loss=0.2333, ctc_loss=0.1518, cr_loss=0.4076, over 17006.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1234, cr_loss=0.3402, over 3368395.46 frames. ], batch size: 56, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:18:56,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=681580.6666666666, ans=0.0
2024-09-25 06:19:15,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=681627.3333333334, ans=0.125
2024-09-25 06:19:25,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=681674.0, ans=0.0
2024-09-25 06:20:01,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=681767.3333333334, ans=0.1
2024-09-25 06:20:04,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=681767.3333333334, ans=0.2
2024-09-25 06:20:08,018 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.59 vs. limit=15.0
2024-09-25 06:20:11,785 INFO [train.py:1198] (1/4) Epoch 38, batch 1950, loss[loss=0.2212, ctc_loss=0.1494, cr_loss=0.3588, over 11794.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1232, cr_loss=0.3398, over 3356533.53 frames. ], batch size: 123, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:20:19,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=681814.0, ans=0.0
2024-09-25 06:20:23,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=681814.0, ans=0.0
2024-09-25 06:20:27,501 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.276e+02 1.376e+02 1.498e+02 2.117e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-25 06:20:27,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=681860.6666666666, ans=0.2
2024-09-25 06:20:27,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=681860.6666666666, ans=0.5
2024-09-25 06:20:43,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=681907.3333333334, ans=0.125
2024-09-25 06:20:45,999 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.88 vs. limit=15.0
2024-09-25 06:21:31,235 INFO [train.py:1198] (1/4) Epoch 38, batch 2000, loss[loss=0.1887, ctc_loss=0.1207, cr_loss=0.3403, over 17008.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1233, cr_loss=0.3393, over 3352956.94 frames. ], batch size: 51, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:21:55,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=682094.0, ans=0.2
2024-09-25 06:21:57,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=682094.0, ans=0.0
2024-09-25 06:22:14,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=682140.6666666666, ans=0.0
2024-09-25 06:22:24,646 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.90 vs. limit=15.0
2024-09-25 06:22:51,391 INFO [train.py:1198] (1/4) Epoch 38, batch 2050, loss[loss=0.183, ctc_loss=0.1175, cr_loss=0.3274, over 17075.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1229, cr_loss=0.3387, over 3358302.81 frames. ], batch size: 46, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:22:54,212 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.66 vs. limit=15.0
2024-09-25 06:23:09,956 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.278e+02 1.341e+02 1.480e+02 2.182e+02, threshold=2.683e+02, percent-clipped=0.0
2024-09-25 06:23:42,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=682420.6666666666, ans=0.1
2024-09-25 06:23:46,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=682420.6666666666, ans=0.0
2024-09-25 06:24:12,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=682467.3333333334, ans=0.0
2024-09-25 06:24:16,375 INFO [train.py:1198] (1/4) Epoch 38, batch 2100, loss[loss=0.1902, ctc_loss=0.125, cr_loss=0.3263, over 17092.00 frames. ], tot_loss[loss=0.1901, ctc_loss=0.1225, cr_loss=0.3382, over 3364957.35 frames. ], batch size: 49, lr: 3.13e-03, grad_scale: 32.0
2024-09-25 06:24:35,357 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=682560.6666666666, ans=0.2
2024-09-25 06:24:46,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=682560.6666666666, ans=0.125
2024-09-25 06:24:47,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=682560.6666666666, ans=0.05
2024-09-25 06:25:00,285 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.04 vs. limit=22.5
2024-09-25 06:25:18,085 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.39 vs. limit=15.0
2024-09-25 06:25:36,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=682700.6666666666, ans=0.125
2024-09-25 06:25:41,096 INFO [train.py:1198] (1/4) Epoch 38, batch 2150, loss[loss=0.1491, ctc_loss=0.0941, cr_loss=0.2752, over 16265.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1229, cr_loss=0.3382, over 3360295.14 frames. ], batch size: 36, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:25:51,309 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=682747.3333333334, ans=0.0
2024-09-25 06:25:59,160 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.297e+02 1.376e+02 1.525e+02 2.502e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-25 06:26:01,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=682794.0, ans=0.0
2024-09-25 06:26:37,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=682887.3333333334, ans=0.0
2024-09-25 06:26:39,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=682887.3333333334, ans=0.125
2024-09-25 06:26:48,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=682934.0, ans=0.125
2024-09-25 06:26:52,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=682934.0, ans=0.2
2024-09-25 06:27:01,625 INFO [train.py:1198] (1/4) Epoch 38, batch 2200, loss[loss=0.1973, ctc_loss=0.1277, cr_loss=0.3483, over 17023.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1233, cr_loss=0.3396, over 3369064.56 frames. ], batch size: 56, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:27:05,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=682980.6666666666, ans=0.125
2024-09-25 06:27:43,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=683074.0, ans=0.125
2024-09-25 06:27:51,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=683120.6666666666, ans=0.125
2024-09-25 06:28:08,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=683167.3333333334, ans=0.0
2024-09-25 06:28:13,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=683167.3333333334, ans=0.125
2024-09-25 06:28:20,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=683167.3333333334, ans=0.1
2024-09-25 06:28:24,720 INFO [train.py:1198] (1/4) Epoch 38, batch 2250, loss[loss=0.1933, ctc_loss=0.1244, cr_loss=0.3449, over 16595.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1238, cr_loss=0.3408, over 3371751.50 frames. ], batch size: 66, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:28:41,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=683260.6666666666, ans=0.5
2024-09-25 06:28:42,346 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.283e+02 1.354e+02 1.470e+02 2.386e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-25 06:28:45,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=683260.6666666666, ans=0.125
2024-09-25 06:29:09,377 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=8.01 vs. limit=15.0
2024-09-25 06:29:43,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=683400.6666666666, ans=0.0
2024-09-25 06:29:45,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=683400.6666666666, ans=0.0
2024-09-25 06:29:49,855 INFO [train.py:1198] (1/4) Epoch 38, batch 2300, loss[loss=0.1911, ctc_loss=0.1236, cr_loss=0.3379, over 16736.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1237, cr_loss=0.3414, over 3377832.12 frames. ], batch size: 61, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:30:02,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=683447.3333333334, ans=0.0
2024-09-25 06:30:18,380 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 06:30:39,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=683587.3333333334, ans=0.125
2024-09-25 06:30:45,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=683587.3333333334, ans=0.0
2024-09-25 06:30:46,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=683587.3333333334, ans=0.0
2024-09-25 06:30:47,503 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.31 vs. limit=15.0
2024-09-25 06:30:56,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=683634.0, ans=0.2
2024-09-25 06:31:12,144 INFO [train.py:1198] (1/4) Epoch 38, batch 2350, loss[loss=0.1912, ctc_loss=0.1219, cr_loss=0.3464, over 17046.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1233, cr_loss=0.3405, over 3375473.27 frames. ], batch size: 52, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:31:12,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=683680.6666666666, ans=0.5
2024-09-25 06:31:15,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=683680.6666666666, ans=0.125
2024-09-25 06:31:15,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=683680.6666666666, ans=0.2
2024-09-25 06:31:16,414 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.91 vs. limit=12.0
2024-09-25 06:31:29,751 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.295e+02 1.360e+02 1.434e+02 2.304e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-25 06:32:24,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten.whitening_limit, batch_count=683867.3333333334, ans=15.0
2024-09-25 06:32:31,858 INFO [train.py:1198] (1/4) Epoch 38, batch 2400, loss[loss=0.2029, ctc_loss=0.1297, cr_loss=0.3662, over 17182.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1237, cr_loss=0.3409, over 3372427.31 frames. ], batch size: 45, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:32:33,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=683914.0, ans=0.2
2024-09-25 06:33:22,428 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=684054.0, ans=0.0
2024-09-25 06:33:24,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=684054.0, ans=0.2
2024-09-25 06:33:54,234 INFO [train.py:1198] (1/4) Epoch 38, batch 2450, loss[loss=0.2241, ctc_loss=0.1521, cr_loss=0.3598, over 11282.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.124, cr_loss=0.3411, over 3362551.49 frames. ], batch size: 123, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:34:11,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=684194.0, ans=0.04949747468305833
2024-09-25 06:34:14,632 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.306e+02 1.378e+02 1.472e+02 2.938e+02, threshold=2.756e+02, percent-clipped=1.0
2024-09-25 06:34:16,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=684194.0, ans=0.0
2024-09-25 06:34:24,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=684194.0, ans=0.125
2024-09-25 06:35:17,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=684334.0, ans=0.125
2024-09-25 06:35:22,453 INFO [train.py:1198] (1/4) Epoch 38, batch 2500, loss[loss=0.2183, ctc_loss=0.1452, cr_loss=0.3655, over 16545.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.124, cr_loss=0.3416, over 3365270.31 frames. ], batch size: 66, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:35:29,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=684380.6666666666, ans=0.0
2024-09-25 06:35:48,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=684427.3333333334, ans=0.0
2024-09-25 06:35:48,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=684427.3333333334, ans=0.025
2024-09-25 06:35:55,292 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.02 vs. limit=15.0
2024-09-25 06:36:05,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=684474.0, ans=0.125
2024-09-25 06:36:17,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=684520.6666666666, ans=0.125
2024-09-25 06:36:23,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=684520.6666666666, ans=0.0
2024-09-25 06:36:26,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=684567.3333333334, ans=0.09899494936611666
2024-09-25 06:36:31,690 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.74 vs. limit=15.0
2024-09-25 06:36:42,200 INFO [train.py:1198] (1/4) Epoch 38, batch 2550, loss[loss=0.2072, ctc_loss=0.1339, cr_loss=0.3665, over 17323.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1237, cr_loss=0.3412, over 3376058.15 frames. ], batch size: 49, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:36:42,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=684614.0, ans=0.125
2024-09-25 06:37:00,104 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.274e+02 1.355e+02 1.436e+02 2.221e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-25 06:37:10,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=684660.6666666666, ans=0.2
2024-09-25 06:37:11,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=684660.6666666666, ans=0.125
2024-09-25 06:37:30,134 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.44 vs. limit=22.5
2024-09-25 06:37:40,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=684754.0, ans=0.0
2024-09-25 06:37:49,099 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=684800.6666666666, ans=0.125
2024-09-25 06:37:50,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=684800.6666666666, ans=0.125
2024-09-25 06:37:51,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=11.13 vs. limit=22.5
2024-09-25 06:38:05,957 INFO [train.py:1198] (1/4) Epoch 38, batch 2600, loss[loss=0.1443, ctc_loss=0.09057, cr_loss=0.2687, over 17268.00 frames. ], tot_loss[loss=0.1929, ctc_loss=0.1245, cr_loss=0.3419, over 3369639.69 frames. ], batch size: 42, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:38:07,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=684847.3333333334, ans=0.125
2024-09-25 06:38:14,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=684847.3333333334, ans=10.0
2024-09-25 06:38:14,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.94 vs. limit=15.0
2024-09-25 06:38:20,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=684894.0, ans=0.125
2024-09-25 06:39:05,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=684987.3333333334, ans=0.125
2024-09-25 06:39:05,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=684987.3333333334, ans=0.2
2024-09-25 06:39:26,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=685034.0, ans=0.2
2024-09-25 06:39:31,217 INFO [train.py:1198] (1/4) Epoch 38, batch 2650, loss[loss=0.2417, ctc_loss=0.1565, cr_loss=0.4259, over 16982.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1237, cr_loss=0.3403, over 3373352.82 frames. ], batch size: 53, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:39:36,260 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 06:39:39,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=685080.6666666666, ans=0.025
2024-09-25 06:39:48,718 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.311e+02 1.384e+02 1.483e+02 1.840e+02, threshold=2.769e+02, percent-clipped=0.0
2024-09-25 06:40:07,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=685174.0, ans=10.0
2024-09-25 06:40:15,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=685174.0, ans=0.125
2024-09-25 06:40:39,474 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.78 vs. limit=15.0
2024-09-25 06:40:53,183 INFO [train.py:1198] (1/4) Epoch 38, batch 2700, loss[loss=0.2089, ctc_loss=0.1361, cr_loss=0.3636, over 17083.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1242, cr_loss=0.3411, over 3368869.44 frames. ], batch size: 49, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:41:13,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=685360.6666666666, ans=0.2
2024-09-25 06:41:16,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=685360.6666666666, ans=0.125
2024-09-25 06:42:12,893 INFO [train.py:1198] (1/4) Epoch 38, batch 2750, loss[loss=0.1651, ctc_loss=0.1049, cr_loss=0.3011, over 17304.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1239, cr_loss=0.3407, over 3369332.15 frames. ], batch size: 42, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:42:32,105 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.260e+02 1.351e+02 1.429e+02 2.193e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-25 06:42:57,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=685640.6666666666, ans=0.125
2024-09-25 06:43:24,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=685734.0, ans=0.125
2024-09-25 06:43:27,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=685734.0, ans=0.0
2024-09-25 06:43:35,231 INFO [train.py:1198] (1/4) Epoch 38, batch 2800, loss[loss=0.1786, ctc_loss=0.1141, cr_loss=0.3222, over 17297.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1236, cr_loss=0.3397, over 3365923.73 frames. ], batch size: 46, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:43:40,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=685780.6666666666, ans=0.0
2024-09-25 06:44:08,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=685874.0, ans=0.1
2024-09-25 06:44:10,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=685874.0, ans=0.125
2024-09-25 06:44:38,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=685920.6666666666, ans=0.125
2024-09-25 06:45:03,154 INFO [train.py:1198] (1/4) Epoch 38, batch 2850, loss[loss=0.1936, ctc_loss=0.1229, cr_loss=0.3535, over 17018.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1234, cr_loss=0.3397, over 3372319.33 frames. ], batch size: 44, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:45:22,310 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.271e+02 1.362e+02 1.479e+02 2.279e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-25 06:45:30,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=686060.6666666666, ans=0.025
2024-09-25 06:46:23,067 INFO [train.py:1198] (1/4) Epoch 38, batch 2900, loss[loss=0.1737, ctc_loss=0.1108, cr_loss=0.3149, over 17022.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1237, cr_loss=0.3406, over 3382185.38 frames. ], batch size: 44, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:46:23,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=686247.3333333334, ans=0.0
2024-09-25 06:46:28,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=686247.3333333334, ans=0.2
2024-09-25 06:46:39,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=686294.0, ans=0.0
2024-09-25 06:46:43,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=686294.0, ans=0.0
2024-09-25 06:46:46,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=686294.0, ans=0.125
2024-09-25 06:47:15,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=686387.3333333334, ans=0.125
2024-09-25 06:47:30,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=686434.0, ans=0.125
2024-09-25 06:47:30,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=686434.0, ans=0.125
2024-09-25 06:47:42,602 INFO [train.py:1198] (1/4) Epoch 38, batch 2950, loss[loss=0.1761, ctc_loss=0.1112, cr_loss=0.3243, over 17011.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1238, cr_loss=0.3407, over 3374123.92 frames. ], batch size: 39, lr: 3.12e-03, grad_scale: 32.0
2024-09-25 06:47:44,568 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 06:47:54,252 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.16 vs. limit=15.0
2024-09-25 06:47:56,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=686480.6666666666, ans=0.1
2024-09-25 06:48:04,470 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.305e+02 1.376e+02 1.477e+02 2.268e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-25 06:48:20,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=686574.0, ans=0.0
2024-09-25 06:48:20,391 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=686574.0, ans=0.125
2024-09-25 06:48:55,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=686667.3333333334, ans=0.95
2024-09-25 06:49:01,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=686667.3333333334, ans=0.04949747468305833
2024-09-25 06:49:07,363 INFO [train.py:1198] (1/4) Epoch 38, batch 3000, loss[loss=0.1742, ctc_loss=0.111, cr_loss=0.3162, over 16696.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.123, cr_loss=0.3395, over 3374935.75 frames. ], batch size: 37, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:49:07,364 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 06:49:18,578 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([5.4825, 4.8868, 4.9211, 5.2049], device='cuda:1')
2024-09-25 06:49:21,343 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.1.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.4169, 4.0841, 3.6259, 4.1723], device='cuda:1')
2024-09-25 06:49:22,916 INFO [train.py:1230] (1/4) Epoch 38, validation: loss=0.03571, ctc_loss=0.03571, cr_loss=9.665e-15, over 944034.00 frames. 
2024-09-25 06:49:22,917 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 06:49:29,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=686714.0, ans=0.125
2024-09-25 06:49:35,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=686714.0, ans=0.1
2024-09-25 06:49:49,156 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.17 vs. limit=15.0
2024-09-25 06:49:53,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=686807.3333333334, ans=0.125
2024-09-25 06:50:12,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=686854.0, ans=0.125
2024-09-25 06:50:17,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=686854.0, ans=0.5
2024-09-25 06:50:30,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=686900.6666666666, ans=0.1
2024-09-25 06:50:44,315 INFO [train.py:1198] (1/4) Epoch 38, batch 3050, loss[loss=0.1863, ctc_loss=0.1163, cr_loss=0.35, over 17311.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1228, cr_loss=0.339, over 3374246.37 frames. ], batch size: 49, lr: 3.12e-03, grad_scale: 16.0
2024-09-25 06:50:45,202 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.36 vs. limit=15.0
2024-09-25 06:50:48,142 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.51 vs. limit=22.5
2024-09-25 06:50:49,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=686947.3333333334, ans=0.0
2024-09-25 06:51:02,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=686994.0, ans=0.125
2024-09-25 06:51:04,197 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.282e+02 1.358e+02 1.470e+02 1.835e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 06:51:55,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=687134.0, ans=0.125
2024-09-25 06:52:01,989 INFO [train.py:1198] (1/4) Epoch 38, batch 3100, loss[loss=0.2199, ctc_loss=0.1439, cr_loss=0.3797, over 17006.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.123, cr_loss=0.3396, over 3372591.83 frames. ], batch size: 53, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 06:52:15,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=687180.6666666666, ans=0.0
2024-09-25 06:52:18,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=687227.3333333334, ans=0.1
2024-09-25 06:52:44,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=687274.0, ans=0.125
2024-09-25 06:53:20,025 INFO [train.py:1198] (1/4) Epoch 38, batch 3150, loss[loss=0.2039, ctc_loss=0.1315, cr_loss=0.3619, over 17220.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1239, cr_loss=0.3411, over 3372407.85 frames. ], batch size: 50, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 06:53:21,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=687414.0, ans=0.1
2024-09-25 06:53:25,392 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.09 vs. limit=22.5
2024-09-25 06:53:37,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=687460.6666666666, ans=0.125
2024-09-25 06:53:39,515 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.01 vs. limit=12.0
2024-09-25 06:53:40,302 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.270e+02 1.356e+02 1.474e+02 1.773e+02, threshold=2.711e+02, percent-clipped=0.0
2024-09-25 06:54:00,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=687507.3333333334, ans=0.125
2024-09-25 06:54:31,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=687600.6666666666, ans=0.125
2024-09-25 06:54:37,787 INFO [train.py:1198] (1/4) Epoch 38, batch 3200, loss[loss=0.1744, ctc_loss=0.111, cr_loss=0.317, over 17199.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1229, cr_loss=0.3387, over 3364428.78 frames. ], batch size: 41, lr: 3.11e-03, grad_scale: 32.0
2024-09-25 06:54:47,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=687647.3333333334, ans=0.125
2024-09-25 06:55:07,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=687740.6666666666, ans=0.125
2024-09-25 06:55:17,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=687740.6666666666, ans=0.125
2024-09-25 06:55:38,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=687834.0, ans=0.125
2024-09-25 06:55:56,023 INFO [train.py:1198] (1/4) Epoch 38, batch 3250, loss[loss=0.1897, ctc_loss=0.1227, cr_loss=0.3348, over 17216.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1233, cr_loss=0.3401, over 3361736.30 frames. ], batch size: 47, lr: 3.11e-03, grad_scale: 32.0
2024-09-25 06:56:01,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=687880.6666666666, ans=0.1
2024-09-25 06:56:17,749 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.275e+02 1.365e+02 1.473e+02 2.154e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-25 06:56:49,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=688020.6666666666, ans=0.0
2024-09-25 06:56:54,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=688020.6666666666, ans=0.125
2024-09-25 06:57:01,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=688067.3333333334, ans=0.125
2024-09-25 06:57:02,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=688067.3333333334, ans=0.5
2024-09-25 06:57:10,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=688067.3333333334, ans=0.0
2024-09-25 06:57:16,214 INFO [train.py:1198] (1/4) Epoch 38, batch 3300, loss[loss=0.2029, ctc_loss=0.1314, cr_loss=0.3579, over 17038.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1239, cr_loss=0.3409, over 3358283.66 frames. ], batch size: 52, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 06:57:21,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=688114.0, ans=0.1
2024-09-25 06:57:27,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=688114.0, ans=0.5
2024-09-25 06:57:28,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=688114.0, ans=0.0
2024-09-25 06:57:39,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=688160.6666666666, ans=0.1
2024-09-25 06:57:53,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=688207.3333333334, ans=0.05
2024-09-25 06:57:58,313 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 06:58:06,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=688254.0, ans=0.025
2024-09-25 06:58:24,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=688300.6666666666, ans=0.125
2024-09-25 06:58:33,935 INFO [train.py:1198] (1/4) Epoch 38, batch 3350, loss[loss=0.2005, ctc_loss=0.1311, cr_loss=0.3472, over 17032.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1242, cr_loss=0.3421, over 3365646.21 frames. ], batch size: 52, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 06:58:55,606 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.280e+02 1.359e+02 1.507e+02 2.410e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-25 06:59:19,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=688440.6666666666, ans=0.1
2024-09-25 06:59:34,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=688487.3333333334, ans=0.0
2024-09-25 06:59:56,003 INFO [train.py:1198] (1/4) Epoch 38, batch 3400, loss[loss=0.2111, ctc_loss=0.1363, cr_loss=0.3743, over 17210.00 frames. ], tot_loss[loss=0.1933, ctc_loss=0.1247, cr_loss=0.3427, over 3353975.34 frames. ], batch size: 47, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:00:38,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=688674.0, ans=0.125
2024-09-25 07:01:16,601 INFO [train.py:1198] (1/4) Epoch 38, batch 3450, loss[loss=0.1849, ctc_loss=0.1193, cr_loss=0.3277, over 17218.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1243, cr_loss=0.3415, over 3361801.95 frames. ], batch size: 47, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:01:39,810 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=8.97 vs. limit=15.0
2024-09-25 07:01:40,121 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.283e+02 1.390e+02 1.486e+02 2.473e+02, threshold=2.781e+02, percent-clipped=0.0
2024-09-25 07:01:56,561 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.08 vs. limit=22.5
2024-09-25 07:02:02,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=688954.0, ans=0.0
2024-09-25 07:02:02,829 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=8.86 vs. limit=15.0
2024-09-25 07:02:04,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=688954.0, ans=0.125
2024-09-25 07:02:24,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=689000.6666666666, ans=0.125
2024-09-25 07:02:29,404 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.24 vs. limit=15.0
2024-09-25 07:02:35,140 INFO [train.py:1198] (1/4) Epoch 38, batch 3500, loss[loss=0.1966, ctc_loss=0.1248, cr_loss=0.3589, over 17295.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1248, cr_loss=0.3424, over 3367552.39 frames. ], batch size: 46, lr: 3.11e-03, grad_scale: 8.0
2024-09-25 07:02:41,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=689047.3333333334, ans=0.125
2024-09-25 07:03:23,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=689187.3333333334, ans=0.125
2024-09-25 07:03:35,445 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.18 vs. limit=15.0
2024-09-25 07:03:53,031 INFO [train.py:1198] (1/4) Epoch 38, batch 3550, loss[loss=0.2243, ctc_loss=0.1453, cr_loss=0.3948, over 16993.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1249, cr_loss=0.3431, over 3368502.16 frames. ], batch size: 56, lr: 3.11e-03, grad_scale: 8.0
2024-09-25 07:04:05,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=689280.6666666666, ans=0.125
2024-09-25 07:04:16,680 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.303e+02 1.383e+02 1.456e+02 2.390e+02, threshold=2.765e+02, percent-clipped=0.0
2024-09-25 07:04:19,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=11.23 vs. limit=15.0
2024-09-25 07:04:54,902 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.89 vs. limit=15.0
2024-09-25 07:05:06,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=689467.3333333334, ans=0.125
2024-09-25 07:05:08,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=689467.3333333334, ans=0.125
2024-09-25 07:05:11,163 INFO [train.py:1198] (1/4) Epoch 38, batch 3600, loss[loss=0.1925, ctc_loss=0.1234, cr_loss=0.3457, over 17306.00 frames. ], tot_loss[loss=0.1929, ctc_loss=0.1245, cr_loss=0.3422, over 3361538.73 frames. ], batch size: 46, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:05:30,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=689560.6666666666, ans=0.125
2024-09-25 07:05:36,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=689560.6666666666, ans=0.0
2024-09-25 07:05:49,430 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.19 vs. limit=22.5
2024-09-25 07:06:29,868 INFO [train.py:1198] (1/4) Epoch 38, batch 3650, loss[loss=0.1964, ctc_loss=0.1288, cr_loss=0.3381, over 16879.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1249, cr_loss=0.3436, over 3365011.49 frames. ], batch size: 58, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:06:41,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=689747.3333333334, ans=0.1
2024-09-25 07:06:55,349 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.248e+02 1.330e+02 1.434e+02 2.019e+02, threshold=2.659e+02, percent-clipped=0.0
2024-09-25 07:07:03,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=689840.6666666666, ans=0.0
2024-09-25 07:07:04,057 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.35 vs. limit=15.0
2024-09-25 07:07:13,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=689840.6666666666, ans=0.125
2024-09-25 07:07:50,475 INFO [train.py:1198] (1/4) Epoch 38, batch 3700, loss[loss=0.2306, ctc_loss=0.1485, cr_loss=0.4105, over 17226.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.1239, cr_loss=0.3415, over 3364688.39 frames. ], batch size: 55, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:08:17,558 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.32 vs. limit=10.0
2024-09-25 07:08:24,696 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.05 vs. limit=15.0
2024-09-25 07:08:53,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=690167.3333333334, ans=0.125
2024-09-25 07:09:01,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=690167.3333333334, ans=0.0
2024-09-25 07:09:10,774 INFO [train.py:1198] (1/4) Epoch 38, batch 3750, loss[loss=0.1854, ctc_loss=0.1204, cr_loss=0.3249, over 17036.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1243, cr_loss=0.3421, over 3358679.97 frames. ], batch size: 44, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:09:15,917 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=690214.0, ans=0.125
2024-09-25 07:09:15,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=690214.0, ans=0.2
2024-09-25 07:09:17,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=690214.0, ans=0.09899494936611666
2024-09-25 07:09:19,523 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=23.21 vs. limit=22.5
2024-09-25 07:09:34,561 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.279e+02 1.362e+02 1.449e+02 2.293e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-25 07:09:40,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_abs, batch_count=690307.3333333334, ans=0.5
2024-09-25 07:10:02,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=690354.0, ans=0.125
2024-09-25 07:10:30,980 INFO [train.py:1198] (1/4) Epoch 38, batch 3800, loss[loss=0.2129, ctc_loss=0.1347, cr_loss=0.391, over 16947.00 frames. ], tot_loss[loss=0.1933, ctc_loss=0.1249, cr_loss=0.3423, over 3332921.79 frames. ], batch size: 58, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:10:40,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=690447.3333333334, ans=0.025
2024-09-25 07:10:42,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=690447.3333333334, ans=0.125
2024-09-25 07:10:49,752 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=690494.0, ans=0.0
2024-09-25 07:10:49,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=690494.0, ans=0.0
2024-09-25 07:11:02,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=690540.6666666666, ans=0.1
2024-09-25 07:11:22,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=690587.3333333334, ans=0.1
2024-09-25 07:11:24,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=690587.3333333334, ans=0.0
2024-09-25 07:11:34,382 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.03 vs. limit=12.0
2024-09-25 07:11:36,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=690634.0, ans=0.125
2024-09-25 07:11:41,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=690634.0, ans=0.0
2024-09-25 07:11:48,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=690634.0, ans=0.125
2024-09-25 07:11:52,446 INFO [train.py:1198] (1/4) Epoch 38, batch 3850, loss[loss=0.2252, ctc_loss=0.1475, cr_loss=0.3885, over 14981.00 frames. ], tot_loss[loss=0.1962, ctc_loss=0.1273, cr_loss=0.3446, over 3239841.42 frames. ], batch size: 89, lr: 3.11e-03, grad_scale: 16.0
2024-09-25 07:11:59,711 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=8.68 vs. limit=22.5
2024-09-25 07:12:00,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=690680.6666666666, ans=0.125
2024-09-25 07:12:03,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=690680.6666666666, ans=0.0
2024-09-25 07:12:04,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=8.26 vs. limit=15.0
2024-09-25 07:12:15,186 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.374e+02 1.486e+02 1.640e+02 2.118e+02, threshold=2.972e+02, percent-clipped=0.0
2024-09-25 07:12:32,272 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 07:12:50,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=690820.6666666666, ans=10.0
2024-09-25 07:12:50,919 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.29 vs. limit=22.5
2024-09-25 07:13:50,137 INFO [train.py:1198] (1/4) Epoch 39, batch 0, loss[loss=0.1931, ctc_loss=0.1236, cr_loss=0.3474, over 17081.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1236, cr_loss=0.3474, over 17081.00 frames. ], batch size: 49, lr: 3.07e-03, grad_scale: 32.0
2024-09-25 07:13:50,138 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 07:13:57,667 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.2.self_attn_weights, attn_weights_entropy = tensor([2.9408, 3.3980, 3.2250, 3.6440, 3.2067, 3.1025, 3.7095, 3.9333],
       device='cuda:1')
2024-09-25 07:14:06,124 INFO [train.py:1230] (1/4) Epoch 39, validation: loss=0.03529, ctc_loss=0.03529, cr_loss=1.033e-14, over 944034.00 frames. 
2024-09-25 07:14:06,124 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 07:14:25,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=690942.0, ans=0.125
2024-09-25 07:14:44,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=690988.6666666666, ans=0.125
2024-09-25 07:14:54,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=690988.6666666666, ans=0.125
2024-09-25 07:14:57,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=691035.3333333334, ans=0.07
2024-09-25 07:15:02,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=691035.3333333334, ans=0.125
2024-09-25 07:15:25,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=691082.0, ans=0.0
2024-09-25 07:15:28,967 INFO [train.py:1198] (1/4) Epoch 39, batch 50, loss[loss=0.1901, ctc_loss=0.1247, cr_loss=0.3269, over 17254.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1234, cr_loss=0.3418, over 755803.58 frames. ], batch size: 44, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:15:47,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=691175.3333333334, ans=0.125
2024-09-25 07:15:47,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=691175.3333333334, ans=0.09899494936611666
2024-09-25 07:15:59,581 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.279e+02 1.417e+02 1.630e+02 3.403e+02, threshold=2.834e+02, percent-clipped=1.0
2024-09-25 07:16:01,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=691222.0, ans=0.025
2024-09-25 07:16:02,171 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.60 vs. limit=15.0
2024-09-25 07:16:22,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=691268.6666666666, ans=0.125
2024-09-25 07:16:52,218 INFO [train.py:1198] (1/4) Epoch 39, batch 100, loss[loss=0.2036, ctc_loss=0.1328, cr_loss=0.3544, over 16785.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1231, cr_loss=0.3401, over 1331495.26 frames. ], batch size: 61, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:17:15,619 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.43 vs. limit=15.0
2024-09-25 07:17:40,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=691502.0, ans=0.0
2024-09-25 07:18:05,140 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.34 vs. limit=15.0
2024-09-25 07:18:10,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=691595.3333333334, ans=0.0
2024-09-25 07:18:12,292 INFO [train.py:1198] (1/4) Epoch 39, batch 150, loss[loss=0.1473, ctc_loss=0.09337, cr_loss=0.2695, over 17277.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1215, cr_loss=0.3368, over 1779989.54 frames. ], batch size: 42, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:18:19,277 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 07:18:39,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=691642.0, ans=0.1
2024-09-25 07:18:45,368 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.277e+02 1.390e+02 1.504e+02 2.454e+02, threshold=2.780e+02, percent-clipped=0.0
2024-09-25 07:19:01,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.67 vs. limit=22.5
2024-09-25 07:19:28,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=691782.0, ans=0.125
2024-09-25 07:19:40,576 INFO [train.py:1198] (1/4) Epoch 39, batch 200, loss[loss=0.165, ctc_loss=0.1018, cr_loss=0.3161, over 17183.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1219, cr_loss=0.3379, over 2134866.98 frames. ], batch size: 41, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:19:42,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=691828.6666666666, ans=0.0
2024-09-25 07:20:00,848 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.01 vs. limit=15.0
2024-09-25 07:20:08,626 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.10 vs. limit=15.0
2024-09-25 07:20:19,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=691922.0, ans=0.125
2024-09-25 07:20:32,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=691968.6666666666, ans=0.05
2024-09-25 07:20:43,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=692015.3333333334, ans=0.2
2024-09-25 07:20:53,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.46 vs. limit=6.0
2024-09-25 07:21:00,250 INFO [train.py:1198] (1/4) Epoch 39, batch 250, loss[loss=0.211, ctc_loss=0.1375, cr_loss=0.3674, over 16847.00 frames. ], tot_loss[loss=0.1894, ctc_loss=0.1219, cr_loss=0.3374, over 2400765.85 frames. ], batch size: 58, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:21:02,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=692062.0, ans=0.1
2024-09-25 07:21:33,649 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.285e+02 1.349e+02 1.463e+02 2.685e+02, threshold=2.698e+02, percent-clipped=0.0
2024-09-25 07:21:42,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=692155.3333333334, ans=0.125
2024-09-25 07:21:43,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=692155.3333333334, ans=0.125
2024-09-25 07:21:44,272 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.07 vs. limit=12.0
2024-09-25 07:21:46,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=692155.3333333334, ans=0.125
2024-09-25 07:21:53,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=692202.0, ans=0.1
2024-09-25 07:22:08,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=692248.6666666666, ans=0.125
2024-09-25 07:22:20,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=692248.6666666666, ans=0.125
2024-09-25 07:22:22,920 INFO [train.py:1198] (1/4) Epoch 39, batch 300, loss[loss=0.1861, ctc_loss=0.1186, cr_loss=0.3374, over 17211.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1216, cr_loss=0.3373, over 2619055.46 frames. ], batch size: 47, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:22:34,951 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.95 vs. limit=15.0
2024-09-25 07:23:07,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=692388.6666666666, ans=0.125
2024-09-25 07:23:32,628 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=7.53 vs. limit=15.0
2024-09-25 07:23:33,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=692482.0, ans=0.125
2024-09-25 07:23:45,871 INFO [train.py:1198] (1/4) Epoch 39, batch 350, loss[loss=0.2068, ctc_loss=0.1327, cr_loss=0.3704, over 17028.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.122, cr_loss=0.3377, over 2785644.67 frames. ], batch size: 51, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:24:19,333 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.72 vs. limit=12.0
2024-09-25 07:24:21,676 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.269e+02 1.341e+02 1.429e+02 1.987e+02, threshold=2.681e+02, percent-clipped=0.0
2024-09-25 07:24:22,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=692622.0, ans=0.5
2024-09-25 07:24:26,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=692622.0, ans=0.125
2024-09-25 07:24:35,403 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.04 vs. limit=22.5
2024-09-25 07:25:10,801 INFO [train.py:1198] (1/4) Epoch 39, batch 400, loss[loss=0.1788, ctc_loss=0.1119, cr_loss=0.3343, over 16943.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.122, cr_loss=0.3374, over 2915914.45 frames. ], batch size: 42, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:25:27,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=692808.6666666666, ans=0.0
2024-09-25 07:26:05,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=692902.0, ans=0.125
2024-09-25 07:26:10,907 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.52 vs. limit=15.0
2024-09-25 07:26:30,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=692948.6666666666, ans=0.0
2024-09-25 07:26:33,631 INFO [train.py:1198] (1/4) Epoch 39, batch 450, loss[loss=0.1926, ctc_loss=0.1243, cr_loss=0.3418, over 17225.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1226, cr_loss=0.3379, over 3012061.55 frames. ], batch size: 47, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:26:35,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=692995.3333333334, ans=0.125
2024-09-25 07:26:45,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=692995.3333333334, ans=0.1
2024-09-25 07:26:46,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=692995.3333333334, ans=0.125
2024-09-25 07:26:51,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=693042.0, ans=0.125
2024-09-25 07:26:56,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=693042.0, ans=0.125
2024-09-25 07:26:56,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=693042.0, ans=0.125
2024-09-25 07:26:57,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=693042.0, ans=0.2
2024-09-25 07:27:03,899 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.286e+02 1.365e+02 1.440e+02 1.919e+02, threshold=2.731e+02, percent-clipped=0.0
2024-09-25 07:27:05,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=693088.6666666666, ans=0.125
2024-09-25 07:27:22,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=693135.3333333334, ans=0.1
2024-09-25 07:27:29,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=693135.3333333334, ans=0.125
2024-09-25 07:27:53,307 INFO [train.py:1198] (1/4) Epoch 39, batch 500, loss[loss=0.1991, ctc_loss=0.1274, cr_loss=0.3587, over 17236.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1235, cr_loss=0.3391, over 3077054.05 frames. ], batch size: 50, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:28:00,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=693228.6666666666, ans=0.125
2024-09-25 07:28:06,985 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.63 vs. limit=15.0
2024-09-25 07:28:08,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=693275.3333333334, ans=0.125
2024-09-25 07:28:14,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=693275.3333333334, ans=0.125
2024-09-25 07:28:16,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=693275.3333333334, ans=0.0
2024-09-25 07:28:24,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=693322.0, ans=0.1
2024-09-25 07:28:40,077 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.47 vs. limit=15.0
2024-09-25 07:28:47,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=693368.6666666666, ans=0.0
2024-09-25 07:29:19,089 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.88 vs. limit=12.0
2024-09-25 07:29:21,520 INFO [train.py:1198] (1/4) Epoch 39, batch 550, loss[loss=0.2093, ctc_loss=0.1351, cr_loss=0.3711, over 16509.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1246, cr_loss=0.3418, over 3132034.62 frames. ], batch size: 66, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:29:23,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=693462.0, ans=0.125
2024-09-25 07:29:30,780 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=6.25 vs. limit=12.0
2024-09-25 07:29:39,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=693508.6666666666, ans=0.125
2024-09-25 07:29:52,388 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.262e+02 1.347e+02 1.440e+02 2.072e+02, threshold=2.694e+02, percent-clipped=0.0
2024-09-25 07:29:57,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=693555.3333333334, ans=0.2
2024-09-25 07:30:34,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=693648.6666666666, ans=0.0
2024-09-25 07:30:42,146 INFO [train.py:1198] (1/4) Epoch 39, batch 600, loss[loss=0.1731, ctc_loss=0.11, cr_loss=0.3152, over 17077.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1244, cr_loss=0.3419, over 3182731.00 frames. ], batch size: 49, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:30:50,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=693695.3333333334, ans=0.0
2024-09-25 07:30:57,755 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.96 vs. limit=15.0
2024-09-25 07:30:58,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=693742.0, ans=0.125
2024-09-25 07:31:14,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=693788.6666666666, ans=0.0
2024-09-25 07:31:40,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=693835.3333333334, ans=0.0
2024-09-25 07:31:42,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=693835.3333333334, ans=0.1
2024-09-25 07:31:49,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=693882.0, ans=0.09899494936611666
2024-09-25 07:32:04,862 INFO [train.py:1198] (1/4) Epoch 39, batch 650, loss[loss=0.1843, ctc_loss=0.1168, cr_loss=0.3371, over 17036.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1247, cr_loss=0.3426, over 3231077.87 frames. ], batch size: 51, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:32:30,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=693975.3333333334, ans=0.0
2024-09-25 07:32:32,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=693975.3333333334, ans=0.2
2024-09-25 07:32:35,162 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.253e+02 1.371e+02 1.476e+02 2.121e+02, threshold=2.742e+02, percent-clipped=0.0
2024-09-25 07:32:40,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=694022.0, ans=0.125
2024-09-25 07:32:54,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=694068.6666666666, ans=0.2
2024-09-25 07:33:07,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=694115.3333333334, ans=0.0
2024-09-25 07:33:15,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=694115.3333333334, ans=0.125
2024-09-25 07:33:15,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=694115.3333333334, ans=0.05
2024-09-25 07:33:24,771 INFO [train.py:1198] (1/4) Epoch 39, batch 700, loss[loss=0.185, ctc_loss=0.1166, cr_loss=0.3419, over 17249.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1246, cr_loss=0.3424, over 3263751.99 frames. ], batch size: 44, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:33:39,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=694162.0, ans=0.2
2024-09-25 07:34:14,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=694255.3333333334, ans=0.1
2024-09-25 07:34:17,901 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 07:34:32,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.whiten.whitening_limit, batch_count=694302.0, ans=15.0
2024-09-25 07:34:43,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=694348.6666666666, ans=0.2
2024-09-25 07:34:52,576 INFO [train.py:1198] (1/4) Epoch 39, batch 750, loss[loss=0.215, ctc_loss=0.1408, cr_loss=0.3709, over 17022.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1249, cr_loss=0.3425, over 3284681.41 frames. ], batch size: 52, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:34:54,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=694395.3333333334, ans=0.125
2024-09-25 07:34:58,723 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.05 vs. limit=12.0
2024-09-25 07:34:59,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=694395.3333333334, ans=0.2
2024-09-25 07:35:08,120 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=17.05 vs. limit=22.5
2024-09-25 07:35:21,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=694442.0, ans=0.1
2024-09-25 07:35:23,322 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.271e+02 1.368e+02 1.469e+02 1.814e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 07:35:30,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=694488.6666666666, ans=0.125
2024-09-25 07:35:35,291 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.94 vs. limit=22.5
2024-09-25 07:35:39,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=694535.3333333334, ans=0.125
2024-09-25 07:35:45,216 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.33 vs. limit=22.5
2024-09-25 07:35:46,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=694535.3333333334, ans=0.125
2024-09-25 07:36:03,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=694582.0, ans=0.1
2024-09-25 07:36:13,111 INFO [train.py:1198] (1/4) Epoch 39, batch 800, loss[loss=0.1614, ctc_loss=0.1035, cr_loss=0.2892, over 17366.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.125, cr_loss=0.3428, over 3301115.28 frames. ], batch size: 48, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:36:19,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=694628.6666666666, ans=0.125
2024-09-25 07:37:05,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=694768.6666666666, ans=0.125
2024-09-25 07:37:35,917 INFO [train.py:1198] (1/4) Epoch 39, batch 850, loss[loss=0.1927, ctc_loss=0.124, cr_loss=0.3435, over 17254.00 frames. ], tot_loss[loss=0.1933, ctc_loss=0.1248, cr_loss=0.3427, over 3310667.64 frames. ], batch size: 44, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:37:58,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=694908.6666666666, ans=0.125
2024-09-25 07:38:06,360 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.258e+02 1.327e+02 1.456e+02 1.924e+02, threshold=2.653e+02, percent-clipped=0.0
2024-09-25 07:38:17,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=694955.3333333334, ans=0.2
2024-09-25 07:38:33,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=695002.0, ans=10.0
2024-09-25 07:38:50,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=695048.6666666666, ans=0.5
2024-09-25 07:38:50,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=695048.6666666666, ans=0.0
2024-09-25 07:39:01,743 INFO [train.py:1198] (1/4) Epoch 39, batch 900, loss[loss=0.2053, ctc_loss=0.1333, cr_loss=0.3599, over 16974.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1243, cr_loss=0.3413, over 3318706.55 frames. ], batch size: 58, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:39:02,447 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.26 vs. limit=15.0
2024-09-25 07:39:05,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=695095.3333333334, ans=0.125
2024-09-25 07:39:11,001 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=695095.3333333334, ans=0.1
2024-09-25 07:39:12,844 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.77 vs. limit=22.5
2024-09-25 07:39:14,573 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.89 vs. limit=6.0
2024-09-25 07:39:18,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=695142.0, ans=0.0
2024-09-25 07:39:22,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=695142.0, ans=0.1
2024-09-25 07:39:44,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=695188.6666666666, ans=0.2
2024-09-25 07:39:46,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=695188.6666666666, ans=0.0
2024-09-25 07:39:49,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=695188.6666666666, ans=0.125
2024-09-25 07:39:49,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=695188.6666666666, ans=0.0
2024-09-25 07:40:00,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=695235.3333333334, ans=0.125
2024-09-25 07:40:24,418 INFO [train.py:1198] (1/4) Epoch 39, batch 950, loss[loss=0.1913, ctc_loss=0.1222, cr_loss=0.3458, over 17181.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1236, cr_loss=0.3404, over 3331432.99 frames. ], batch size: 45, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:40:55,205 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.302e+02 1.382e+02 1.480e+02 1.852e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-25 07:40:56,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=5.28 vs. limit=15.0
2024-09-25 07:41:03,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=695422.0, ans=0.1
2024-09-25 07:41:08,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=695422.0, ans=0.125
2024-09-25 07:41:28,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=695468.6666666666, ans=0.025
2024-09-25 07:41:31,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=695515.3333333334, ans=0.125
2024-09-25 07:41:39,840 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 07:41:47,516 INFO [train.py:1198] (1/4) Epoch 39, batch 1000, loss[loss=0.22, ctc_loss=0.1411, cr_loss=0.3947, over 17206.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1241, cr_loss=0.3419, over 3344233.64 frames. ], batch size: 47, lr: 3.06e-03, grad_scale: 32.0
2024-09-25 07:41:51,960 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=8.29 vs. limit=15.0
2024-09-25 07:41:55,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=695562.0, ans=0.125
2024-09-25 07:42:07,559 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.28 vs. limit=22.5
2024-09-25 07:42:28,179 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.66 vs. limit=15.0
2024-09-25 07:42:43,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=695702.0, ans=0.125
2024-09-25 07:42:53,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=695748.6666666666, ans=0.125
2024-09-25 07:43:07,518 INFO [train.py:1198] (1/4) Epoch 39, batch 1050, loss[loss=0.1789, ctc_loss=0.1158, cr_loss=0.3152, over 16995.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3419, over 3344296.45 frames. ], batch size: 51, lr: 3.05e-03, grad_scale: 32.0
2024-09-25 07:43:07,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=695795.3333333334, ans=0.125
2024-09-25 07:43:22,686 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.42 vs. limit=12.0
2024-09-25 07:43:37,198 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 07:43:40,129 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.305e+02 1.388e+02 1.496e+02 1.693e+02, threshold=2.775e+02, percent-clipped=0.0
2024-09-25 07:43:40,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=695888.6666666666, ans=0.0
2024-09-25 07:44:02,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=695935.3333333334, ans=0.2
2024-09-25 07:44:07,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer_ff2.min_abs, batch_count=695935.3333333334, ans=0.1
2024-09-25 07:44:22,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=695982.0, ans=0.0
2024-09-25 07:44:23,797 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=695982.0, ans=0.125
2024-09-25 07:44:34,844 INFO [train.py:1198] (1/4) Epoch 39, batch 1100, loss[loss=0.163, ctc_loss=0.1026, cr_loss=0.3019, over 17028.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1238, cr_loss=0.341, over 3355649.08 frames. ], batch size: 39, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:44:38,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=696028.6666666666, ans=0.125
2024-09-25 07:44:43,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=696028.6666666666, ans=0.125
2024-09-25 07:44:46,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=696028.6666666666, ans=0.025
2024-09-25 07:44:56,296 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.21 vs. limit=22.5
2024-09-25 07:45:03,884 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=696075.3333333334, ans=0.125
2024-09-25 07:45:03,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=696075.3333333334, ans=0.1
2024-09-25 07:45:11,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=696122.0, ans=0.125
2024-09-25 07:45:43,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=696215.3333333334, ans=0.0
2024-09-25 07:45:54,571 INFO [train.py:1198] (1/4) Epoch 39, batch 1150, loss[loss=0.1784, ctc_loss=0.1164, cr_loss=0.3103, over 17161.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1244, cr_loss=0.3418, over 3349727.23 frames. ], batch size: 45, lr: 3.05e-03, grad_scale: 8.0
2024-09-25 07:46:30,825 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.295e+02 1.353e+02 1.453e+02 1.736e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-25 07:46:42,905 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.29 vs. limit=15.0
2024-09-25 07:46:54,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=696402.0, ans=0.0
2024-09-25 07:47:16,819 INFO [train.py:1198] (1/4) Epoch 39, batch 1200, loss[loss=0.23, ctc_loss=0.1487, cr_loss=0.4065, over 16940.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1242, cr_loss=0.3412, over 3339529.08 frames. ], batch size: 58, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:47:41,459 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.85 vs. limit=15.0
2024-09-25 07:47:55,572 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=5.56 vs. limit=15.0
2024-09-25 07:48:00,551 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.48 vs. limit=15.0
2024-09-25 07:48:39,496 INFO [train.py:1198] (1/4) Epoch 39, batch 1250, loss[loss=0.1984, ctc_loss=0.1297, cr_loss=0.3436, over 17313.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1251, cr_loss=0.3422, over 3341100.78 frames. ], batch size: 51, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:48:54,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=696728.6666666666, ans=0.125
2024-09-25 07:48:58,546 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.74 vs. limit=22.5
2024-09-25 07:49:05,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=696775.3333333334, ans=0.125
2024-09-25 07:49:05,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=696775.3333333334, ans=0.125
2024-09-25 07:49:17,453 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.275e+02 1.351e+02 1.497e+02 2.020e+02, threshold=2.702e+02, percent-clipped=0.0
2024-09-25 07:49:28,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=696822.0, ans=0.1
2024-09-25 07:50:03,595 INFO [train.py:1198] (1/4) Epoch 39, batch 1300, loss[loss=0.2014, ctc_loss=0.1272, cr_loss=0.3708, over 17085.00 frames. ], tot_loss[loss=0.1942, ctc_loss=0.1255, cr_loss=0.3436, over 3346443.76 frames. ], batch size: 49, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:50:04,274 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.74 vs. limit=12.0
2024-09-25 07:50:17,226 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.77 vs. limit=15.0
2024-09-25 07:50:31,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=697008.6666666666, ans=0.0
2024-09-25 07:50:49,323 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.39 vs. limit=12.0
2024-09-25 07:50:59,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=697102.0, ans=0.125
2024-09-25 07:51:26,312 INFO [train.py:1198] (1/4) Epoch 39, batch 1350, loss[loss=0.1651, ctc_loss=0.1055, cr_loss=0.2979, over 17098.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.1242, cr_loss=0.3406, over 3346078.38 frames. ], batch size: 43, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:51:31,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=697195.3333333334, ans=0.025
2024-09-25 07:51:59,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=697288.6666666666, ans=0.1
2024-09-25 07:52:00,575 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.264e+02 1.325e+02 1.430e+02 2.601e+02, threshold=2.650e+02, percent-clipped=0.0
2024-09-25 07:52:02,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=697288.6666666666, ans=0.0
2024-09-25 07:52:20,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=697335.3333333334, ans=0.2
2024-09-25 07:52:42,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=697382.0, ans=0.025
2024-09-25 07:52:44,912 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.29 vs. limit=15.0
2024-09-25 07:52:47,367 INFO [train.py:1198] (1/4) Epoch 39, batch 1400, loss[loss=0.1988, ctc_loss=0.1271, cr_loss=0.3585, over 17289.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1238, cr_loss=0.3399, over 3345885.50 frames. ], batch size: 46, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:53:10,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=697475.3333333334, ans=0.125
2024-09-25 07:53:21,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=697522.0, ans=0.125
2024-09-25 07:53:29,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=697522.0, ans=0.125
2024-09-25 07:53:37,655 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.63 vs. limit=6.0
2024-09-25 07:53:41,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=697568.6666666666, ans=0.125
2024-09-25 07:54:11,253 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.31 vs. limit=22.5
2024-09-25 07:54:15,111 INFO [train.py:1198] (1/4) Epoch 39, batch 1450, loss[loss=0.1941, ctc_loss=0.1249, cr_loss=0.346, over 17227.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.1241, cr_loss=0.3404, over 3341123.79 frames. ], batch size: 50, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:54:48,362 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.285e+02 1.356e+02 1.493e+02 2.927e+02, threshold=2.712e+02, percent-clipped=2.0
2024-09-25 07:55:34,690 INFO [train.py:1198] (1/4) Epoch 39, batch 1500, loss[loss=0.1487, ctc_loss=0.09323, cr_loss=0.2773, over 17129.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1237, cr_loss=0.3399, over 3353127.54 frames. ], batch size: 40, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:55:41,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=697895.3333333334, ans=0.125
2024-09-25 07:56:38,253 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 07:56:57,035 INFO [train.py:1198] (1/4) Epoch 39, batch 1550, loss[loss=0.2283, ctc_loss=0.1472, cr_loss=0.4059, over 17002.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1237, cr_loss=0.3401, over 3341954.78 frames. ], batch size: 53, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 07:57:30,846 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.291e+02 1.368e+02 1.486e+02 2.492e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 07:57:35,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=698222.0, ans=0.125
2024-09-25 07:58:04,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=698315.3333333334, ans=0.125
2024-09-25 07:58:17,122 INFO [train.py:1198] (1/4) Epoch 39, batch 1600, loss[loss=0.2, ctc_loss=0.1299, cr_loss=0.3504, over 17340.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1239, cr_loss=0.3407, over 3350124.90 frames. ], batch size: 48, lr: 3.05e-03, grad_scale: 32.0
2024-09-25 07:58:32,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.87 vs. limit=22.5
2024-09-25 07:58:32,241 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.31 vs. limit=22.5
2024-09-25 07:59:14,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=698502.0, ans=0.125
2024-09-25 07:59:27,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=698548.6666666666, ans=0.125
2024-09-25 07:59:42,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=698595.3333333334, ans=0.1
2024-09-25 07:59:44,295 INFO [train.py:1198] (1/4) Epoch 39, batch 1650, loss[loss=0.1635, ctc_loss=0.1028, cr_loss=0.3039, over 17092.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1237, cr_loss=0.3409, over 3345747.21 frames. ], batch size: 43, lr: 3.05e-03, grad_scale: 32.0
2024-09-25 08:00:07,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=698642.0, ans=0.125
2024-09-25 08:00:13,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=698642.0, ans=0.2
2024-09-25 08:00:17,963 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.287e+02 1.365e+02 1.434e+02 2.064e+02, threshold=2.730e+02, percent-clipped=0.0
2024-09-25 08:00:24,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=698688.6666666666, ans=0.125
2024-09-25 08:00:53,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=698782.0, ans=0.1
2024-09-25 08:01:04,261 INFO [train.py:1198] (1/4) Epoch 39, batch 1700, loss[loss=0.1787, ctc_loss=0.1144, cr_loss=0.3214, over 16992.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1242, cr_loss=0.3409, over 3334306.49 frames. ], batch size: 51, lr: 3.05e-03, grad_scale: 32.0
2024-09-25 08:01:26,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=698875.3333333334, ans=0.07
2024-09-25 08:01:53,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=698968.6666666666, ans=0.95
2024-09-25 08:02:23,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=699015.3333333334, ans=0.2
2024-09-25 08:02:25,992 INFO [train.py:1198] (1/4) Epoch 39, batch 1750, loss[loss=0.1956, ctc_loss=0.1273, cr_loss=0.3417, over 17225.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1242, cr_loss=0.3409, over 3339652.12 frames. ], batch size: 50, lr: 3.05e-03, grad_scale: 32.0
2024-09-25 08:02:45,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.18 vs. limit=15.0
2024-09-25 08:03:01,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.281e+02 1.369e+02 1.457e+02 2.012e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-25 08:03:16,162 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.84 vs. limit=15.0
2024-09-25 08:03:17,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=699202.0, ans=0.125
2024-09-25 08:03:37,140 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.87 vs. limit=12.0
2024-09-25 08:03:53,686 INFO [train.py:1198] (1/4) Epoch 39, batch 1800, loss[loss=0.2065, ctc_loss=0.1331, cr_loss=0.3671, over 17096.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3416, over 3345198.44 frames. ], batch size: 49, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 08:04:06,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.min_positive, batch_count=699295.3333333334, ans=0.05
2024-09-25 08:04:08,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=699342.0, ans=0.125
2024-09-25 08:04:19,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=699342.0, ans=0.125
2024-09-25 08:04:30,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=699388.6666666666, ans=0.1
2024-09-25 08:04:54,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=699435.3333333334, ans=0.125
2024-09-25 08:05:01,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=699482.0, ans=0.1
2024-09-25 08:05:04,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=699482.0, ans=0.125
2024-09-25 08:05:13,795 INFO [train.py:1198] (1/4) Epoch 39, batch 1850, loss[loss=0.2021, ctc_loss=0.1308, cr_loss=0.3563, over 17185.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.124, cr_loss=0.3413, over 3348856.47 frames. ], batch size: 55, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 08:05:44,476 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=699622.0, ans=0.0
2024-09-25 08:05:48,762 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.264e+02 1.367e+02 1.511e+02 2.303e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-25 08:06:08,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=699668.6666666666, ans=0.2
2024-09-25 08:06:22,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=699715.3333333334, ans=0.04949747468305833
2024-09-25 08:06:29,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.27 vs. limit=15.0
2024-09-25 08:06:36,625 INFO [train.py:1198] (1/4) Epoch 39, batch 1900, loss[loss=0.2266, ctc_loss=0.1483, cr_loss=0.3916, over 17022.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1243, cr_loss=0.3418, over 3346850.68 frames. ], batch size: 56, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 08:06:37,652 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.07 vs. limit=15.0
2024-09-25 08:07:07,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=699855.3333333334, ans=0.1
2024-09-25 08:07:08,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=699855.3333333334, ans=0.1
2024-09-25 08:07:21,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=699855.3333333334, ans=0.125
2024-09-25 08:07:47,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=699948.6666666666, ans=0.0
2024-09-25 08:07:53,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=699948.6666666666, ans=0.1
2024-09-25 08:07:56,980 INFO [train.py:1198] (1/4) Epoch 39, batch 1950, loss[loss=0.2087, ctc_loss=0.1336, cr_loss=0.3752, over 16998.00 frames. ], tot_loss[loss=0.194, ctc_loss=0.1252, cr_loss=0.3439, over 3349658.53 frames. ], batch size: 53, lr: 3.05e-03, grad_scale: 16.0
2024-09-25 08:08:17,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=700042.0, ans=0.07
2024-09-25 08:08:35,377 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.35 vs. limit=12.0
2024-09-25 08:08:37,869 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.289e+02 1.367e+02 1.538e+02 1.984e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-25 08:08:52,952 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=9.77 vs. limit=15.0
2024-09-25 08:09:20,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=700182.0, ans=0.125
2024-09-25 08:09:22,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=700182.0, ans=0.1
2024-09-25 08:09:24,325 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.56 vs. limit=22.5
2024-09-25 08:09:25,321 INFO [train.py:1198] (1/4) Epoch 39, batch 2000, loss[loss=0.2267, ctc_loss=0.1507, cr_loss=0.3803, over 15320.00 frames. ], tot_loss[loss=0.1937, ctc_loss=0.1251, cr_loss=0.343, over 3344848.54 frames. ], batch size: 89, lr: 3.04e-03, grad_scale: 32.0
2024-09-25 08:09:25,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=700228.6666666666, ans=0.0
2024-09-25 08:09:30,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=700228.6666666666, ans=0.125
2024-09-25 08:09:36,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=700228.6666666666, ans=0.0
2024-09-25 08:09:36,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=700228.6666666666, ans=0.125
2024-09-25 08:09:42,027 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.93 vs. limit=15.0
2024-09-25 08:09:45,760 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.36 vs. limit=22.5
2024-09-25 08:09:51,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=700275.3333333334, ans=0.125
2024-09-25 08:10:17,310 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.85 vs. limit=15.0
2024-09-25 08:10:36,713 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.88 vs. limit=15.0
2024-09-25 08:10:37,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=700415.3333333334, ans=0.125
2024-09-25 08:10:39,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=700415.3333333334, ans=0.125
2024-09-25 08:10:45,500 INFO [train.py:1198] (1/4) Epoch 39, batch 2050, loss[loss=0.1695, ctc_loss=0.1059, cr_loss=0.3183, over 17167.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1247, cr_loss=0.3424, over 3339039.82 frames. ], batch size: 45, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:10:49,350 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.29 vs. limit=15.0
2024-09-25 08:10:50,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=700462.0, ans=0.0
2024-09-25 08:10:54,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=700462.0, ans=0.2
2024-09-25 08:11:00,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=700508.6666666666, ans=0.125
2024-09-25 08:11:15,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=700508.6666666666, ans=0.125
2024-09-25 08:11:24,179 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.09 vs. limit=15.0
2024-09-25 08:11:25,132 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.300e+02 1.377e+02 1.457e+02 2.836e+02, threshold=2.753e+02, percent-clipped=1.0
2024-09-25 08:11:33,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=700555.3333333334, ans=0.1
2024-09-25 08:11:39,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=700602.0, ans=0.125
2024-09-25 08:11:44,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=700602.0, ans=0.125
2024-09-25 08:11:55,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=700648.6666666666, ans=0.0
2024-09-25 08:12:08,228 INFO [train.py:1198] (1/4) Epoch 39, batch 2100, loss[loss=0.1804, ctc_loss=0.117, cr_loss=0.3167, over 17254.00 frames. ], tot_loss[loss=0.1934, ctc_loss=0.1248, cr_loss=0.343, over 3350810.33 frames. ], batch size: 44, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:12:15,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=700695.3333333334, ans=0.125
2024-09-25 08:12:27,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=700742.0, ans=0.2
2024-09-25 08:12:49,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=700788.6666666666, ans=0.125
2024-09-25 08:12:54,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=700835.3333333334, ans=0.125
2024-09-25 08:13:01,663 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.72 vs. limit=12.0
2024-09-25 08:13:04,990 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.99 vs. limit=22.5
2024-09-25 08:13:30,505 INFO [train.py:1198] (1/4) Epoch 39, batch 2150, loss[loss=0.1857, ctc_loss=0.1203, cr_loss=0.3274, over 17014.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1249, cr_loss=0.3433, over 3351613.20 frames. ], batch size: 44, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:14:10,019 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.289e+02 1.363e+02 1.447e+02 2.047e+02, threshold=2.727e+02, percent-clipped=0.0
2024-09-25 08:14:10,901 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.15 vs. limit=6.0
2024-09-25 08:14:16,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=701022.0, ans=10.0
2024-09-25 08:14:41,084 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.49 vs. limit=15.0
2024-09-25 08:14:51,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=701162.0, ans=0.0
2024-09-25 08:14:53,283 INFO [train.py:1198] (1/4) Epoch 39, batch 2200, loss[loss=0.2143, ctc_loss=0.1409, cr_loss=0.3671, over 16585.00 frames. ], tot_loss[loss=0.1929, ctc_loss=0.1246, cr_loss=0.3413, over 3349851.58 frames. ], batch size: 66, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:15:04,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=701162.0, ans=0.2
2024-09-25 08:15:06,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=701162.0, ans=0.125
2024-09-25 08:15:17,925 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 08:15:32,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=701255.3333333334, ans=0.2
2024-09-25 08:15:51,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=701302.0, ans=0.035
2024-09-25 08:15:56,374 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.78 vs. limit=15.0
2024-09-25 08:16:16,047 INFO [train.py:1198] (1/4) Epoch 39, batch 2250, loss[loss=0.1641, ctc_loss=0.1048, cr_loss=0.2966, over 16983.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1248, cr_loss=0.3417, over 3360312.39 frames. ], batch size: 42, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:16:52,927 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.277e+02 1.348e+02 1.481e+02 2.538e+02, threshold=2.695e+02, percent-clipped=0.0
2024-09-25 08:16:53,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=701488.6666666666, ans=0.125
2024-09-25 08:17:02,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=701535.3333333334, ans=0.0
2024-09-25 08:17:10,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=701535.3333333334, ans=0.0
2024-09-25 08:17:10,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=701535.3333333334, ans=0.125
2024-09-25 08:17:12,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=701535.3333333334, ans=0.125
2024-09-25 08:17:13,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=701535.3333333334, ans=0.025
2024-09-25 08:17:35,950 INFO [train.py:1198] (1/4) Epoch 39, batch 2300, loss[loss=0.2231, ctc_loss=0.1458, cr_loss=0.3863, over 17232.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1249, cr_loss=0.3415, over 3351382.94 frames. ], batch size: 55, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:18:00,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=701675.3333333334, ans=0.125
2024-09-25 08:18:14,815 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.13 vs. limit=15.0
2024-09-25 08:18:15,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=701722.0, ans=0.125
2024-09-25 08:18:19,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=701722.0, ans=0.125
2024-09-25 08:18:26,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=701722.0, ans=0.125
2024-09-25 08:18:49,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=701815.3333333334, ans=0.05
2024-09-25 08:19:03,993 INFO [train.py:1198] (1/4) Epoch 39, batch 2350, loss[loss=0.2301, ctc_loss=0.1502, cr_loss=0.3995, over 17012.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1249, cr_loss=0.3417, over 3355863.14 frames. ], batch size: 56, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:19:10,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=701862.0, ans=0.125
2024-09-25 08:19:20,903 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.18 vs. limit=15.0
2024-09-25 08:19:40,705 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.317e+02 1.394e+02 1.473e+02 1.777e+02, threshold=2.787e+02, percent-clipped=0.0
2024-09-25 08:20:19,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=702048.6666666666, ans=0.0
2024-09-25 08:20:23,819 INFO [train.py:1198] (1/4) Epoch 39, batch 2400, loss[loss=0.2036, ctc_loss=0.1307, cr_loss=0.3647, over 16749.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3411, over 3350365.02 frames. ], batch size: 61, lr: 3.04e-03, grad_scale: 32.0
2024-09-25 08:20:24,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=702095.3333333334, ans=0.0
2024-09-25 08:20:29,308 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.48 vs. limit=15.0
2024-09-25 08:20:33,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=702095.3333333334, ans=0.125
2024-09-25 08:20:46,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=702142.0, ans=0.1
2024-09-25 08:20:47,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=702142.0, ans=0.0
2024-09-25 08:21:13,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.58 vs. limit=15.0
2024-09-25 08:21:14,918 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.74 vs. limit=12.0
2024-09-25 08:21:20,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=702235.3333333334, ans=0.125
2024-09-25 08:21:24,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=702235.3333333334, ans=0.0
2024-09-25 08:21:26,167 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=7.95 vs. limit=15.0
2024-09-25 08:21:28,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=702282.0, ans=0.125
2024-09-25 08:21:36,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=702282.0, ans=0.125
2024-09-25 08:21:36,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=702282.0, ans=0.1
2024-09-25 08:21:45,938 INFO [train.py:1198] (1/4) Epoch 39, batch 2450, loss[loss=0.2053, ctc_loss=0.1316, cr_loss=0.3685, over 17217.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1246, cr_loss=0.3419, over 3344953.74 frames. ], batch size: 50, lr: 3.04e-03, grad_scale: 32.0
2024-09-25 08:21:52,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=702328.6666666666, ans=0.125
2024-09-25 08:21:52,974 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.02 vs. limit=10.0
2024-09-25 08:21:56,064 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.03 vs. limit=6.0
2024-09-25 08:22:00,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=702375.3333333334, ans=0.0
2024-09-25 08:22:14,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=702375.3333333334, ans=0.0
2024-09-25 08:22:18,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=702422.0, ans=0.1
2024-09-25 08:22:24,052 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.314e+02 1.404e+02 1.496e+02 1.831e+02, threshold=2.808e+02, percent-clipped=0.0
2024-09-25 08:23:08,347 INFO [train.py:1198] (1/4) Epoch 39, batch 2500, loss[loss=0.1472, ctc_loss=0.09388, cr_loss=0.2668, over 17267.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1239, cr_loss=0.3406, over 3356088.63 frames. ], batch size: 42, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:23:11,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=702562.0, ans=0.0
2024-09-25 08:23:22,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=702562.0, ans=0.125
2024-09-25 08:23:39,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=702608.6666666666, ans=0.0
2024-09-25 08:23:47,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=702655.3333333334, ans=0.125
2024-09-25 08:24:00,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=702702.0, ans=0.125
2024-09-25 08:24:01,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=702702.0, ans=0.125
2024-09-25 08:24:13,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=702702.0, ans=0.125
2024-09-25 08:24:33,559 INFO [train.py:1198] (1/4) Epoch 39, batch 2550, loss[loss=0.1796, ctc_loss=0.115, cr_loss=0.3231, over 16955.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1234, cr_loss=0.3405, over 3359480.38 frames. ], batch size: 42, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:24:39,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=702795.3333333334, ans=0.1
2024-09-25 08:24:42,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=702795.3333333334, ans=0.2
2024-09-25 08:25:02,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=702842.0, ans=0.125
2024-09-25 08:25:12,023 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.300e+02 1.397e+02 1.510e+02 1.872e+02, threshold=2.794e+02, percent-clipped=0.0
2024-09-25 08:25:23,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=702935.3333333334, ans=0.125
2024-09-25 08:25:25,513 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 08:25:56,179 INFO [train.py:1198] (1/4) Epoch 39, batch 2600, loss[loss=0.1872, ctc_loss=0.1218, cr_loss=0.3272, over 17195.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1234, cr_loss=0.3409, over 3361784.60 frames. ], batch size: 47, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:26:07,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=703028.6666666666, ans=0.125
2024-09-25 08:26:27,367 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.67 vs. limit=15.0
2024-09-25 08:26:44,590 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.49 vs. limit=15.0
2024-09-25 08:26:57,733 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.38 vs. limit=15.0
2024-09-25 08:27:15,879 INFO [train.py:1198] (1/4) Epoch 39, batch 2650, loss[loss=0.2054, ctc_loss=0.1311, cr_loss=0.3715, over 17288.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1235, cr_loss=0.3409, over 3362434.04 frames. ], batch size: 46, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:27:41,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=703308.6666666666, ans=0.0
2024-09-25 08:27:53,523 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.290e+02 1.367e+02 1.451e+02 1.893e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-25 08:27:59,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=703355.3333333334, ans=0.05
2024-09-25 08:28:10,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=703402.0, ans=0.025
2024-09-25 08:28:25,948 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=703448.6666666666, ans=0.125
2024-09-25 08:28:43,413 INFO [train.py:1198] (1/4) Epoch 39, batch 2700, loss[loss=0.2313, ctc_loss=0.1483, cr_loss=0.4149, over 16996.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1237, cr_loss=0.3412, over 3368251.48 frames. ], batch size: 53, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:28:49,155 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.07 vs. limit=15.0
2024-09-25 08:29:02,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=703542.0, ans=0.2
2024-09-25 08:29:05,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=703542.0, ans=0.125
2024-09-25 08:29:05,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=703542.0, ans=0.1
2024-09-25 08:29:18,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=703588.6666666666, ans=0.1
2024-09-25 08:29:23,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.99 vs. limit=15.0
2024-09-25 08:29:26,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=703588.6666666666, ans=0.0
2024-09-25 08:29:38,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.28 vs. limit=15.0
2024-09-25 08:29:39,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=703635.3333333334, ans=0.125
2024-09-25 08:29:51,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=703682.0, ans=0.125
2024-09-25 08:30:02,892 INFO [train.py:1198] (1/4) Epoch 39, batch 2750, loss[loss=0.1923, ctc_loss=0.1248, cr_loss=0.3377, over 17199.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1232, cr_loss=0.3406, over 3372988.33 frames. ], batch size: 55, lr: 3.04e-03, grad_scale: 16.0
2024-09-25 08:30:13,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=703728.6666666666, ans=0.0
2024-09-25 08:30:17,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=703775.3333333334, ans=0.125
2024-09-25 08:30:22,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=703775.3333333334, ans=0.025
2024-09-25 08:30:37,536 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.43 vs. limit=22.5
2024-09-25 08:30:38,956 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.59 vs. limit=15.0
2024-09-25 08:30:41,401 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.292e+02 1.404e+02 1.488e+02 2.567e+02, threshold=2.807e+02, percent-clipped=0.0
2024-09-25 08:30:41,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=703822.0, ans=0.1
2024-09-25 08:31:19,526 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=5.177e-03
2024-09-25 08:31:25,415 INFO [train.py:1198] (1/4) Epoch 39, batch 2800, loss[loss=0.1823, ctc_loss=0.1177, cr_loss=0.323, over 16879.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.124, cr_loss=0.3421, over 3371021.92 frames. ], batch size: 58, lr: 3.04e-03, grad_scale: 32.0
2024-09-25 08:31:27,619 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.87 vs. limit=10.0
2024-09-25 08:32:18,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=704102.0, ans=0.0
2024-09-25 08:32:20,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=704102.0, ans=0.125
2024-09-25 08:32:45,469 INFO [train.py:1198] (1/4) Epoch 39, batch 2850, loss[loss=0.1716, ctc_loss=0.108, cr_loss=0.318, over 17223.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1238, cr_loss=0.3416, over 3370510.84 frames. ], batch size: 50, lr: 3.04e-03, grad_scale: 32.0
2024-09-25 08:33:24,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=704288.6666666666, ans=0.025
2024-09-25 08:33:30,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=704288.6666666666, ans=0.125
2024-09-25 08:33:31,755 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.313e+02 1.374e+02 1.493e+02 1.951e+02, threshold=2.749e+02, percent-clipped=0.0
2024-09-25 08:33:33,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=704288.6666666666, ans=0.05
2024-09-25 08:34:02,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=704382.0, ans=0.125
2024-09-25 08:34:09,679 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.34 vs. limit=15.0
2024-09-25 08:34:13,711 INFO [train.py:1198] (1/4) Epoch 39, batch 2900, loss[loss=0.2503, ctc_loss=0.1675, cr_loss=0.4143, over 15097.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1242, cr_loss=0.3419, over 3364287.19 frames. ], batch size: 89, lr: 3.04e-03, grad_scale: 32.0
2024-09-25 08:34:13,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=704428.6666666666, ans=0.125
2024-09-25 08:35:19,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=704615.3333333334, ans=0.125
2024-09-25 08:35:33,809 INFO [train.py:1198] (1/4) Epoch 39, batch 2950, loss[loss=0.2143, ctc_loss=0.1414, cr_loss=0.3645, over 15966.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1243, cr_loss=0.342, over 3374132.63 frames. ], batch size: 74, lr: 3.04e-03, grad_scale: 32.0
2024-09-25 08:35:59,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=704708.6666666666, ans=0.0
2024-09-25 08:35:59,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=704708.6666666666, ans=0.0
2024-09-25 08:36:06,144 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.31 vs. limit=15.0
2024-09-25 08:36:07,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=704755.3333333334, ans=0.125
2024-09-25 08:36:15,121 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.295e+02 1.375e+02 1.468e+02 1.743e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-25 08:36:21,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=704755.3333333334, ans=0.125
2024-09-25 08:36:23,456 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.09 vs. limit=10.0
2024-09-25 08:36:32,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=704802.0, ans=0.125
2024-09-25 08:36:42,816 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.45 vs. limit=15.0
2024-09-25 08:36:46,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=704848.6666666666, ans=0.0
2024-09-25 08:36:51,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=704848.6666666666, ans=0.2
2024-09-25 08:36:55,830 INFO [train.py:1198] (1/4) Epoch 39, batch 3000, loss[loss=0.1821, ctc_loss=0.1169, cr_loss=0.3258, over 17296.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1247, cr_loss=0.3425, over 3362965.74 frames. ], batch size: 46, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:36:55,831 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 08:37:11,152 INFO [train.py:1230] (1/4) Epoch 39, validation: loss=0.03549, ctc_loss=0.03549, cr_loss=9.367e-15, over 944034.00 frames. 
2024-09-25 08:37:11,153 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 08:37:35,295 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=15.63 vs. limit=22.5
2024-09-25 08:38:09,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=705035.3333333334, ans=0.0
2024-09-25 08:38:12,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=705082.0, ans=0.025
2024-09-25 08:38:13,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=705082.0, ans=0.0
2024-09-25 08:38:16,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=705082.0, ans=0.125
2024-09-25 08:38:18,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=705082.0, ans=0.125
2024-09-25 08:38:28,207 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.82 vs. limit=12.0
2024-09-25 08:38:29,029 INFO [train.py:1198] (1/4) Epoch 39, batch 3050, loss[loss=0.1694, ctc_loss=0.106, cr_loss=0.3169, over 16969.00 frames. ], tot_loss[loss=0.193, ctc_loss=0.1245, cr_loss=0.3425, over 3364594.34 frames. ], batch size: 42, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:38:40,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=705128.6666666666, ans=0.2
2024-09-25 08:38:41,823 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 08:38:48,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=705175.3333333334, ans=0.125
2024-09-25 08:38:49,735 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.19 vs. limit=15.0
2024-09-25 08:39:01,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=705222.0, ans=0.0
2024-09-25 08:39:08,845 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.272e+02 1.353e+02 1.453e+02 2.990e+02, threshold=2.707e+02, percent-clipped=1.0
2024-09-25 08:39:36,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=705315.3333333334, ans=0.125
2024-09-25 08:39:54,025 INFO [train.py:1198] (1/4) Epoch 39, batch 3100, loss[loss=0.1766, ctc_loss=0.1145, cr_loss=0.3102, over 17015.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1237, cr_loss=0.3413, over 3372799.85 frames. ], batch size: 51, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:41:11,987 INFO [train.py:1198] (1/4) Epoch 39, batch 3150, loss[loss=0.168, ctc_loss=0.1058, cr_loss=0.3107, over 16708.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1235, cr_loss=0.3408, over 3373127.16 frames. ], batch size: 37, lr: 3.03e-03, grad_scale: 16.0
2024-09-25 08:41:15,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=705595.3333333334, ans=0.05
2024-09-25 08:41:48,965 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.11 vs. limit=22.5
2024-09-25 08:41:50,876 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.273e+02 1.360e+02 1.459e+02 1.912e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-25 08:41:57,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=705735.3333333334, ans=0.1
2024-09-25 08:41:57,670 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.43 vs. limit=15.0
2024-09-25 08:42:05,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=705735.3333333334, ans=0.025
2024-09-25 08:42:09,962 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 08:42:17,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=705782.0, ans=0.125
2024-09-25 08:42:20,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=705782.0, ans=0.125
2024-09-25 08:42:29,861 INFO [train.py:1198] (1/4) Epoch 39, batch 3200, loss[loss=0.1987, ctc_loss=0.1275, cr_loss=0.3556, over 17043.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.123, cr_loss=0.34, over 3376105.32 frames. ], batch size: 52, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:43:01,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=705922.0, ans=0.0
2024-09-25 08:43:06,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=705922.0, ans=0.04949747468305833
2024-09-25 08:43:23,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=705968.6666666666, ans=0.0
2024-09-25 08:43:48,013 INFO [train.py:1198] (1/4) Epoch 39, batch 3250, loss[loss=0.1803, ctc_loss=0.1156, cr_loss=0.3237, over 17034.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1232, cr_loss=0.34, over 3373606.69 frames. ], batch size: 51, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:44:08,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=706108.6666666666, ans=0.125
2024-09-25 08:44:24,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=706155.3333333334, ans=0.0
2024-09-25 08:44:26,959 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.281e+02 1.355e+02 1.471e+02 2.202e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-25 08:44:28,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=706155.3333333334, ans=0.05
2024-09-25 08:44:35,131 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=706202.0, ans=0.0
2024-09-25 08:44:50,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=706248.6666666666, ans=0.0
2024-09-25 08:45:01,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=706248.6666666666, ans=0.125
2024-09-25 08:45:05,829 INFO [train.py:1198] (1/4) Epoch 39, batch 3300, loss[loss=0.173, ctc_loss=0.1081, cr_loss=0.3244, over 17348.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.124, cr_loss=0.3409, over 3354470.10 frames. ], batch size: 48, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:45:14,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.28 vs. limit=22.5
2024-09-25 08:45:21,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=706342.0, ans=0.125
2024-09-25 08:45:30,171 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.79 vs. limit=15.0
2024-09-25 08:45:45,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=706388.6666666666, ans=0.125
2024-09-25 08:46:26,218 INFO [train.py:1198] (1/4) Epoch 39, batch 3350, loss[loss=0.1883, ctc_loss=0.1199, cr_loss=0.3422, over 17156.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1239, cr_loss=0.3408, over 3358286.77 frames. ], batch size: 48, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:47:05,602 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.259e+02 1.360e+02 1.476e+02 1.978e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-25 08:47:09,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=706622.0, ans=0.1
2024-09-25 08:47:10,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=706622.0, ans=0.125
2024-09-25 08:47:27,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=706715.3333333334, ans=0.025
2024-09-25 08:47:40,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=706715.3333333334, ans=0.125
2024-09-25 08:47:42,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=706715.3333333334, ans=0.125
2024-09-25 08:47:44,981 INFO [train.py:1198] (1/4) Epoch 39, batch 3400, loss[loss=0.1816, ctc_loss=0.1194, cr_loss=0.311, over 16797.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1237, cr_loss=0.3399, over 3350559.98 frames. ], batch size: 61, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:48:13,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=706808.6666666666, ans=0.0
2024-09-25 08:48:50,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=706948.6666666666, ans=0.125
2024-09-25 08:48:57,631 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.49 vs. limit=10.0
2024-09-25 08:49:00,521 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.69 vs. limit=15.0
2024-09-25 08:49:03,008 INFO [train.py:1198] (1/4) Epoch 39, batch 3450, loss[loss=0.2036, ctc_loss=0.134, cr_loss=0.3482, over 17293.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.124, cr_loss=0.3409, over 3349009.92 frames. ], batch size: 51, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:49:17,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=707042.0, ans=0.0
2024-09-25 08:49:29,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=707042.0, ans=0.0
2024-09-25 08:49:34,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=707042.0, ans=0.125
2024-09-25 08:49:36,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.out_whiten.whitening_limit, batch_count=707042.0, ans=8.0
2024-09-25 08:49:48,211 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.321e+02 1.412e+02 1.498e+02 2.585e+02, threshold=2.824e+02, percent-clipped=0.0
2024-09-25 08:49:50,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=707088.6666666666, ans=0.0
2024-09-25 08:49:51,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=707088.6666666666, ans=0.125
2024-09-25 08:50:14,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=707182.0, ans=0.125
2024-09-25 08:50:18,294 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.17 vs. limit=15.0
2024-09-25 08:50:26,704 INFO [train.py:1198] (1/4) Epoch 39, batch 3500, loss[loss=0.2089, ctc_loss=0.1372, cr_loss=0.3585, over 16552.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1237, cr_loss=0.3407, over 3353067.36 frames. ], batch size: 66, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:51:44,613 INFO [train.py:1198] (1/4) Epoch 39, batch 3550, loss[loss=0.1795, ctc_loss=0.1158, cr_loss=0.3185, over 17022.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1233, cr_loss=0.3396, over 3345306.68 frames. ], batch size: 44, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:51:54,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=707462.0, ans=0.0
2024-09-25 08:52:03,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=707508.6666666666, ans=0.125
2024-09-25 08:52:12,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1.whitening_limit, batch_count=707508.6666666666, ans=10.0
2024-09-25 08:52:18,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.whiten.whitening_limit, batch_count=707555.3333333334, ans=12.0
2024-09-25 08:52:23,764 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.264e+02 1.345e+02 1.429e+02 1.997e+02, threshold=2.690e+02, percent-clipped=0.0
2024-09-25 08:52:25,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=707555.3333333334, ans=0.125
2024-09-25 08:52:32,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=707602.0, ans=0.0
2024-09-25 08:52:32,552 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.96 vs. limit=15.0
2024-09-25 08:52:39,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=707602.0, ans=0.0
2024-09-25 08:53:02,932 INFO [train.py:1198] (1/4) Epoch 39, batch 3600, loss[loss=0.208, ctc_loss=0.1338, cr_loss=0.3713, over 16880.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1239, cr_loss=0.3407, over 3354937.86 frames. ], batch size: 58, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:53:04,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=707695.3333333334, ans=0.125
2024-09-25 08:53:20,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=707742.0, ans=0.125
2024-09-25 08:54:03,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=707882.0, ans=0.125
2024-09-25 08:54:08,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.min_positive, batch_count=707882.0, ans=0.05
2024-09-25 08:54:20,462 INFO [train.py:1198] (1/4) Epoch 39, batch 3650, loss[loss=0.1548, ctc_loss=0.09557, cr_loss=0.2963, over 17237.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1237, cr_loss=0.3406, over 3366558.37 frames. ], batch size: 47, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:54:35,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=707975.3333333334, ans=0.0
2024-09-25 08:54:59,908 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.284e+02 1.368e+02 1.461e+02 2.127e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 08:55:04,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=708022.0, ans=0.025
2024-09-25 08:55:15,138 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.74 vs. limit=10.0
2024-09-25 08:55:34,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=708115.3333333334, ans=0.1
2024-09-25 08:55:40,513 INFO [train.py:1198] (1/4) Epoch 39, batch 3700, loss[loss=0.2284, ctc_loss=0.1558, cr_loss=0.3634, over 11675.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1247, cr_loss=0.342, over 3338275.84 frames. ], batch size: 123, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:55:48,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=708162.0, ans=0.1
2024-09-25 08:56:06,374 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.55 vs. limit=15.0
2024-09-25 08:56:28,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=708302.0, ans=0.0
2024-09-25 08:56:59,402 INFO [train.py:1198] (1/4) Epoch 39, batch 3750, loss[loss=0.2095, ctc_loss=0.1358, cr_loss=0.3686, over 16883.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.1241, cr_loss=0.3409, over 3329484.82 frames. ], batch size: 58, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:57:24,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=708442.0, ans=0.125
2024-09-25 08:57:38,679 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.270e+02 1.361e+02 1.496e+02 2.109e+02, threshold=2.722e+02, percent-clipped=0.0
2024-09-25 08:58:18,291 INFO [train.py:1198] (1/4) Epoch 39, batch 3800, loss[loss=0.152, ctc_loss=0.09405, cr_loss=0.2895, over 17167.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.1241, cr_loss=0.3406, over 3324700.09 frames. ], batch size: 41, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:58:29,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=708628.6666666666, ans=0.125
2024-09-25 08:58:31,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=708628.6666666666, ans=0.125
2024-09-25 08:59:18,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=708768.6666666666, ans=0.0
2024-09-25 08:59:39,441 INFO [train.py:1198] (1/4) Epoch 39, batch 3850, loss[loss=0.1593, ctc_loss=0.09864, cr_loss=0.3035, over 16270.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1225, cr_loss=0.3373, over 3299586.17 frames. ], batch size: 36, lr: 3.03e-03, grad_scale: 32.0
2024-09-25 08:59:47,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=708862.0, ans=0.1
2024-09-25 09:00:13,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=10.20 vs. limit=15.0
2024-09-25 09:00:18,549 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.306e+02 1.417e+02 1.619e+02 3.168e+02, threshold=2.835e+02, percent-clipped=2.0
2024-09-25 09:01:40,860 INFO [train.py:1198] (1/4) Epoch 40, batch 0, loss[loss=0.2244, ctc_loss=0.1425, cr_loss=0.4095, over 16992.00 frames. ], tot_loss[loss=0.2244, ctc_loss=0.1425, cr_loss=0.4095, over 16992.00 frames. ], batch size: 53, lr: 2.99e-03, grad_scale: 32.0
2024-09-25 09:01:40,861 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 09:01:56,598 INFO [train.py:1230] (1/4) Epoch 40, validation: loss=0.03491, ctc_loss=0.03491, cr_loss=1.007e-14, over 944034.00 frames. 
2024-09-25 09:01:56,599 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 09:02:00,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=709076.6666666666, ans=0.0
2024-09-25 09:02:05,725 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.32 vs. limit=12.0
2024-09-25 09:02:46,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=709216.6666666666, ans=0.0
2024-09-25 09:03:15,888 INFO [train.py:1198] (1/4) Epoch 40, batch 50, loss[loss=0.2054, ctc_loss=0.1328, cr_loss=0.363, over 16860.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1203, cr_loss=0.335, over 751815.98 frames. ], batch size: 58, lr: 2.99e-03, grad_scale: 32.0
2024-09-25 09:03:30,123 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 09:03:33,864 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.21 vs. limit=22.5
2024-09-25 09:03:41,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=709356.6666666666, ans=0.025
2024-09-25 09:03:49,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.68 vs. limit=15.0
2024-09-25 09:04:10,882 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.285e+02 1.412e+02 1.570e+02 2.190e+02, threshold=2.824e+02, percent-clipped=0.0
2024-09-25 09:04:25,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=709450.0, ans=0.125
2024-09-25 09:04:44,462 INFO [train.py:1198] (1/4) Epoch 40, batch 100, loss[loss=0.1874, ctc_loss=0.1173, cr_loss=0.3503, over 17309.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1193, cr_loss=0.3329, over 1330882.48 frames. ], batch size: 46, lr: 2.99e-03, grad_scale: 32.0
2024-09-25 09:04:49,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=709543.3333333334, ans=0.0
2024-09-25 09:05:21,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=709636.6666666666, ans=0.0
2024-09-25 09:05:22,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=709636.6666666666, ans=0.125
2024-09-25 09:05:24,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=709636.6666666666, ans=0.0
2024-09-25 09:05:25,686 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=709636.6666666666, ans=0.125
2024-09-25 09:05:33,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=709683.3333333334, ans=0.125
2024-09-25 09:05:38,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=709683.3333333334, ans=0.1
2024-09-25 09:05:38,765 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=5.06 vs. limit=15.0
2024-09-25 09:06:05,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=709776.6666666666, ans=0.0
2024-09-25 09:06:06,998 INFO [train.py:1198] (1/4) Epoch 40, batch 150, loss[loss=0.1823, ctc_loss=0.1163, cr_loss=0.33, over 17209.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1199, cr_loss=0.3343, over 1784304.41 frames. ], batch size: 47, lr: 2.99e-03, grad_scale: 32.0
2024-09-25 09:06:10,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=709776.6666666666, ans=0.125
2024-09-25 09:06:10,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=709776.6666666666, ans=0.1
2024-09-25 09:06:12,486 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.87 vs. limit=6.0
2024-09-25 09:06:18,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=709776.6666666666, ans=0.0
2024-09-25 09:06:56,159 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.270e+02 1.336e+02 1.420e+02 2.555e+02, threshold=2.672e+02, percent-clipped=0.0
2024-09-25 09:06:56,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=709916.6666666666, ans=0.125
2024-09-25 09:07:01,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=709916.6666666666, ans=0.0
2024-09-25 09:07:29,562 INFO [train.py:1198] (1/4) Epoch 40, batch 200, loss[loss=0.2002, ctc_loss=0.1305, cr_loss=0.3488, over 16736.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1204, cr_loss=0.3352, over 2139960.30 frames. ], batch size: 61, lr: 2.99e-03, grad_scale: 32.0
2024-09-25 09:07:36,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=710010.0, ans=0.0
2024-09-25 09:07:53,099 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.71 vs. limit=15.0
2024-09-25 09:08:03,244 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 09:08:50,009 INFO [train.py:1198] (1/4) Epoch 40, batch 250, loss[loss=0.2108, ctc_loss=0.1382, cr_loss=0.3631, over 17038.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1217, cr_loss=0.3377, over 2413592.78 frames. ], batch size: 52, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:08:50,982 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.83 vs. limit=10.0
2024-09-25 09:09:17,539 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 09:09:19,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=710290.0, ans=0.1
2024-09-25 09:09:39,756 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.275e+02 1.359e+02 1.486e+02 2.398e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-25 09:09:51,691 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.29 vs. limit=10.0
2024-09-25 09:10:16,461 INFO [train.py:1198] (1/4) Epoch 40, batch 300, loss[loss=0.2232, ctc_loss=0.1446, cr_loss=0.3931, over 17318.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1228, cr_loss=0.3394, over 2618590.63 frames. ], batch size: 51, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:10:38,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=710523.3333333334, ans=0.125
2024-09-25 09:10:38,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=710523.3333333334, ans=0.05
2024-09-25 09:11:04,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=710616.6666666666, ans=0.0
2024-09-25 09:11:36,228 INFO [train.py:1198] (1/4) Epoch 40, batch 350, loss[loss=0.2184, ctc_loss=0.1427, cr_loss=0.3785, over 16919.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.124, cr_loss=0.341, over 2774083.06 frames. ], batch size: 58, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:11:36,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=710710.0, ans=0.1
2024-09-25 09:12:27,035 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.301e+02 1.389e+02 1.518e+02 2.541e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-25 09:12:52,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=710896.6666666666, ans=10.0
2024-09-25 09:12:59,094 INFO [train.py:1198] (1/4) Epoch 40, batch 400, loss[loss=0.2284, ctc_loss=0.1501, cr_loss=0.3919, over 15032.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1229, cr_loss=0.3392, over 2903913.35 frames. ], batch size: 89, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:13:16,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=710990.0, ans=0.0
2024-09-25 09:13:26,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=710990.0, ans=0.0
2024-09-25 09:13:33,599 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=7.15 vs. limit=15.0
2024-09-25 09:13:45,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=711083.3333333334, ans=0.125
2024-09-25 09:13:48,016 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=4.92 vs. limit=15.0
2024-09-25 09:13:49,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=711083.3333333334, ans=0.2
2024-09-25 09:13:57,719 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.51 vs. limit=15.0
2024-09-25 09:13:58,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=711083.3333333334, ans=0.125
2024-09-25 09:14:01,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=711083.3333333334, ans=0.125
2024-09-25 09:14:03,939 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.89 vs. limit=15.0
2024-09-25 09:14:09,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=711130.0, ans=0.125
2024-09-25 09:14:22,072 INFO [train.py:1198] (1/4) Epoch 40, batch 450, loss[loss=0.1969, ctc_loss=0.126, cr_loss=0.3542, over 17016.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1233, cr_loss=0.3397, over 2997323.78 frames. ], batch size: 51, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:14:33,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=711176.6666666666, ans=0.0
2024-09-25 09:15:12,493 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.283e+02 1.337e+02 1.424e+02 2.250e+02, threshold=2.674e+02, percent-clipped=0.0
2024-09-25 09:15:16,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.23 vs. limit=22.5
2024-09-25 09:15:28,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=711363.3333333334, ans=0.125
2024-09-25 09:15:31,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=711363.3333333334, ans=0.1
2024-09-25 09:15:38,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=711363.3333333334, ans=0.125
2024-09-25 09:15:44,620 INFO [train.py:1198] (1/4) Epoch 40, batch 500, loss[loss=0.2098, ctc_loss=0.1354, cr_loss=0.372, over 16421.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1234, cr_loss=0.3401, over 3085954.20 frames. ], batch size: 66, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:16:07,966 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.19 vs. limit=22.5
2024-09-25 09:16:14,105 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.80 vs. limit=10.0
2024-09-25 09:16:19,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=711503.3333333334, ans=0.125
2024-09-25 09:16:23,525 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.90 vs. limit=15.0
2024-09-25 09:16:36,747 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=7.87 vs. limit=15.0
2024-09-25 09:16:53,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=711596.6666666666, ans=0.2
2024-09-25 09:17:02,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=711596.6666666666, ans=0.125
2024-09-25 09:17:07,179 INFO [train.py:1198] (1/4) Epoch 40, batch 550, loss[loss=0.1835, ctc_loss=0.1174, cr_loss=0.3305, over 17047.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.123, cr_loss=0.3401, over 3155459.26 frames. ], batch size: 46, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:17:26,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=711690.0, ans=0.1
2024-09-25 09:17:57,148 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.279e+02 1.358e+02 1.490e+02 2.059e+02, threshold=2.716e+02, percent-clipped=0.0
2024-09-25 09:18:20,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=711830.0, ans=0.125
2024-09-25 09:18:28,035 INFO [train.py:1198] (1/4) Epoch 40, batch 600, loss[loss=0.1606, ctc_loss=0.102, cr_loss=0.293, over 17198.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1225, cr_loss=0.3391, over 3200622.27 frames. ], batch size: 41, lr: 2.98e-03, grad_scale: 16.0
2024-09-25 09:18:37,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=711876.6666666666, ans=0.0
2024-09-25 09:19:16,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=711970.0, ans=0.0
2024-09-25 09:19:29,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=712016.6666666666, ans=0.125
2024-09-25 09:19:46,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=712063.3333333334, ans=0.125
2024-09-25 09:19:55,899 INFO [train.py:1198] (1/4) Epoch 40, batch 650, loss[loss=0.1876, ctc_loss=0.1211, cr_loss=0.3325, over 17003.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.1222, cr_loss=0.3387, over 3231259.74 frames. ], batch size: 39, lr: 2.98e-03, grad_scale: 16.0
2024-09-25 09:20:02,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=712110.0, ans=0.125
2024-09-25 09:20:07,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=712110.0, ans=0.1
2024-09-25 09:20:22,691 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.00 vs. limit=6.0
2024-09-25 09:20:22,988 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.74 vs. limit=15.0
2024-09-25 09:20:35,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.49 vs. limit=15.0
2024-09-25 09:20:45,845 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.262e+02 1.364e+02 1.466e+02 1.763e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-25 09:20:51,661 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.75 vs. limit=6.0
2024-09-25 09:21:16,165 INFO [train.py:1198] (1/4) Epoch 40, batch 700, loss[loss=0.1993, ctc_loss=0.1289, cr_loss=0.352, over 17021.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1222, cr_loss=0.3395, over 3265447.68 frames. ], batch size: 52, lr: 2.98e-03, grad_scale: 16.0
2024-09-25 09:21:30,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=712390.0, ans=0.0
2024-09-25 09:21:32,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=712390.0, ans=0.125
2024-09-25 09:21:32,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=712390.0, ans=0.0
2024-09-25 09:21:37,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.01 vs. limit=15.0
2024-09-25 09:21:54,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=712436.6666666666, ans=0.2
2024-09-25 09:22:15,201 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.64 vs. limit=22.5
2024-09-25 09:22:19,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=712483.3333333334, ans=0.125
2024-09-25 09:22:38,079 INFO [train.py:1198] (1/4) Epoch 40, batch 750, loss[loss=0.1775, ctc_loss=0.1145, cr_loss=0.3146, over 17094.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.1229, cr_loss=0.3402, over 3281951.43 frames. ], batch size: 43, lr: 2.98e-03, grad_scale: 16.0
2024-09-25 09:22:39,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=712576.6666666666, ans=0.0
2024-09-25 09:22:40,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=712576.6666666666, ans=0.125
2024-09-25 09:22:43,412 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=10.01 vs. limit=15.0
2024-09-25 09:23:02,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=712623.3333333334, ans=0.025
2024-09-25 09:23:21,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=712670.0, ans=0.1
2024-09-25 09:23:27,401 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.294e+02 1.361e+02 1.444e+02 2.754e+02, threshold=2.722e+02, percent-clipped=1.0
2024-09-25 09:23:35,647 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=712716.6666666666, ans=0.125
2024-09-25 09:24:00,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=712763.3333333334, ans=0.125
2024-09-25 09:24:03,443 INFO [train.py:1198] (1/4) Epoch 40, batch 800, loss[loss=0.1964, ctc_loss=0.1227, cr_loss=0.3681, over 17155.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1231, cr_loss=0.3405, over 3297646.00 frames. ], batch size: 48, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:24:36,903 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=712903.3333333334, ans=0.1
2024-09-25 09:24:45,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.62 vs. limit=15.0
2024-09-25 09:24:53,849 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.58 vs. limit=22.5
2024-09-25 09:25:10,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=712996.6666666666, ans=0.035
2024-09-25 09:25:26,601 INFO [train.py:1198] (1/4) Epoch 40, batch 850, loss[loss=0.176, ctc_loss=0.1141, cr_loss=0.3098, over 17296.00 frames. ], tot_loss[loss=0.1908, ctc_loss=0.1228, cr_loss=0.34, over 3307398.44 frames. ], batch size: 49, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:25:48,713 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.47 vs. limit=10.0
2024-09-25 09:25:56,437 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.43 vs. limit=10.0
2024-09-25 09:26:05,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=713136.6666666666, ans=0.1
2024-09-25 09:26:16,238 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.263e+02 1.352e+02 1.405e+02 2.259e+02, threshold=2.704e+02, percent-clipped=0.0
2024-09-25 09:26:26,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=713183.3333333334, ans=0.09899494936611666
2024-09-25 09:26:49,140 INFO [train.py:1198] (1/4) Epoch 40, batch 900, loss[loss=0.1899, ctc_loss=0.1225, cr_loss=0.3371, over 17301.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1235, cr_loss=0.3414, over 3318975.90 frames. ], batch size: 49, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:26:54,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=713276.6666666666, ans=0.125
2024-09-25 09:26:57,913 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.25 vs. limit=22.5
2024-09-25 09:27:05,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=713323.3333333334, ans=0.025
2024-09-25 09:27:26,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=713370.0, ans=0.025
2024-09-25 09:27:43,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.18 vs. limit=15.0
2024-09-25 09:27:44,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=713416.6666666666, ans=0.125
2024-09-25 09:27:48,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=713416.6666666666, ans=0.125
2024-09-25 09:27:52,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=713463.3333333334, ans=0.125
2024-09-25 09:28:09,204 INFO [train.py:1198] (1/4) Epoch 40, batch 950, loss[loss=0.2098, ctc_loss=0.1366, cr_loss=0.3659, over 16826.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1233, cr_loss=0.3409, over 3334312.50 frames. ], batch size: 61, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:29:04,047 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.318e+02 1.382e+02 1.485e+02 2.096e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-25 09:29:09,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=713650.0, ans=0.2
2024-09-25 09:29:32,939 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=713696.6666666666, ans=0.125
2024-09-25 09:29:37,543 INFO [train.py:1198] (1/4) Epoch 40, batch 1000, loss[loss=0.1713, ctc_loss=0.1099, cr_loss=0.3066, over 17174.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1236, cr_loss=0.34, over 3315426.10 frames. ], batch size: 41, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:29:41,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=713743.3333333334, ans=0.0
2024-09-25 09:29:44,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=713743.3333333334, ans=0.125
2024-09-25 09:29:46,211 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.16 vs. limit=15.0
2024-09-25 09:30:23,405 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=11.97 vs. limit=15.0
2024-09-25 09:30:25,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=713883.3333333334, ans=0.025
2024-09-25 09:30:27,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=713883.3333333334, ans=0.0
2024-09-25 09:30:56,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=713976.6666666666, ans=0.0
2024-09-25 09:30:57,911 INFO [train.py:1198] (1/4) Epoch 40, batch 1050, loss[loss=0.1548, ctc_loss=0.09753, cr_loss=0.2862, over 17119.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3415, over 3324761.23 frames. ], batch size: 40, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:31:04,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=713976.6666666666, ans=0.0
2024-09-25 09:31:37,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=714070.0, ans=0.125
2024-09-25 09:31:41,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=714070.0, ans=0.125
2024-09-25 09:31:48,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=714116.6666666666, ans=0.0
2024-09-25 09:31:51,952 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.281e+02 1.362e+02 1.480e+02 1.897e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-25 09:32:20,604 INFO [train.py:1198] (1/4) Epoch 40, batch 1100, loss[loss=0.1528, ctc_loss=0.09679, cr_loss=0.2798, over 17078.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1237, cr_loss=0.3407, over 3337495.96 frames. ], batch size: 39, lr: 2.98e-03, grad_scale: 16.0
2024-09-25 09:33:02,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=714303.3333333334, ans=0.125
2024-09-25 09:33:18,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=714350.0, ans=0.125
2024-09-25 09:33:43,020 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=6.15 vs. limit=15.0
2024-09-25 09:33:43,636 INFO [train.py:1198] (1/4) Epoch 40, batch 1150, loss[loss=0.1891, ctc_loss=0.1245, cr_loss=0.3233, over 15940.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1232, cr_loss=0.3399, over 3351531.48 frames. ], batch size: 74, lr: 2.98e-03, grad_scale: 16.0
2024-09-25 09:34:13,979 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 09:34:38,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=714583.3333333334, ans=0.125
2024-09-25 09:34:40,141 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.266e+02 1.369e+02 1.564e+02 2.152e+02, threshold=2.737e+02, percent-clipped=0.0
2024-09-25 09:34:43,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=714583.3333333334, ans=0.025
2024-09-25 09:35:08,859 INFO [train.py:1198] (1/4) Epoch 40, batch 1200, loss[loss=0.1762, ctc_loss=0.112, cr_loss=0.3214, over 17211.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.123, cr_loss=0.3397, over 3355453.92 frames. ], batch size: 41, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:35:13,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=714676.6666666666, ans=0.125
2024-09-25 09:35:47,692 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=714770.0, ans=0.1
2024-09-25 09:36:05,850 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.96 vs. limit=12.0
2024-09-25 09:36:15,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=714863.3333333334, ans=0.0
2024-09-25 09:36:18,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=714863.3333333334, ans=0.035
2024-09-25 09:36:29,056 INFO [train.py:1198] (1/4) Epoch 40, batch 1250, loss[loss=0.2131, ctc_loss=0.138, cr_loss=0.3755, over 17148.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1233, cr_loss=0.3402, over 3354578.91 frames. ], batch size: 48, lr: 2.98e-03, grad_scale: 32.0
2024-09-25 09:37:00,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=714956.6666666666, ans=0.125
2024-09-25 09:37:02,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=715003.3333333334, ans=0.125
2024-09-25 09:37:12,329 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.93 vs. limit=15.0
2024-09-25 09:37:14,018 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.92 vs. limit=22.5
2024-09-25 09:37:22,756 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.288e+02 1.344e+02 1.458e+02 2.122e+02, threshold=2.687e+02, percent-clipped=0.0
2024-09-25 09:37:38,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=715096.6666666666, ans=0.2
2024-09-25 09:37:51,161 INFO [train.py:1198] (1/4) Epoch 40, batch 1300, loss[loss=0.1706, ctc_loss=0.1083, cr_loss=0.3112, over 17087.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1225, cr_loss=0.3388, over 3355128.64 frames. ], batch size: 43, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:38:08,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=715190.0, ans=0.125
2024-09-25 09:38:21,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=715236.6666666666, ans=0.0
2024-09-25 09:38:23,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=715236.6666666666, ans=0.2
2024-09-25 09:38:24,869 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=715236.6666666666, ans=0.125
2024-09-25 09:38:54,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=715283.3333333334, ans=10.0
2024-09-25 09:38:57,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=715283.3333333334, ans=0.125
2024-09-25 09:39:04,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=715330.0, ans=0.0
2024-09-25 09:39:19,070 INFO [train.py:1198] (1/4) Epoch 40, batch 1350, loss[loss=0.1517, ctc_loss=0.09655, cr_loss=0.2759, over 17191.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1227, cr_loss=0.3392, over 3353299.39 frames. ], batch size: 41, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:39:24,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=715376.6666666666, ans=0.125
2024-09-25 09:39:25,891 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 09:40:10,085 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.269e+02 1.368e+02 1.487e+02 1.942e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 09:40:10,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=715516.6666666666, ans=0.2
2024-09-25 09:40:39,122 INFO [train.py:1198] (1/4) Epoch 40, batch 1400, loss[loss=0.221, ctc_loss=0.1512, cr_loss=0.349, over 11932.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.123, cr_loss=0.3399, over 3344530.45 frames. ], batch size: 123, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:40:49,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=715610.0, ans=0.125
2024-09-25 09:41:11,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=715703.3333333334, ans=0.125
2024-09-25 09:41:31,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=715750.0, ans=0.125
2024-09-25 09:41:49,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=715796.6666666666, ans=0.125
2024-09-25 09:41:56,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=715796.6666666666, ans=0.125
2024-09-25 09:42:00,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=715843.3333333334, ans=0.0
2024-09-25 09:42:02,343 INFO [train.py:1198] (1/4) Epoch 40, batch 1450, loss[loss=0.1803, ctc_loss=0.1139, cr_loss=0.3318, over 17060.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.122, cr_loss=0.3382, over 3352190.65 frames. ], batch size: 39, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:42:04,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=715843.3333333334, ans=0.2
2024-09-25 09:42:07,424 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 09:42:10,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=715843.3333333334, ans=0.125
2024-09-25 09:42:13,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=715843.3333333334, ans=0.1
2024-09-25 09:42:25,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=715890.0, ans=0.0
2024-09-25 09:42:28,161 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=715890.0, ans=0.125
2024-09-25 09:42:53,418 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.248e+02 1.311e+02 1.404e+02 2.777e+02, threshold=2.622e+02, percent-clipped=1.0
2024-09-25 09:43:03,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=715983.3333333334, ans=15.0
2024-09-25 09:43:04,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=716030.0, ans=0.1
2024-09-25 09:43:17,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=716030.0, ans=0.0
2024-09-25 09:43:21,941 INFO [train.py:1198] (1/4) Epoch 40, batch 1500, loss[loss=0.2185, ctc_loss=0.1393, cr_loss=0.3958, over 17223.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1228, cr_loss=0.3395, over 3350558.24 frames. ], batch size: 47, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:43:49,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=716123.3333333334, ans=0.125
2024-09-25 09:43:51,518 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.67 vs. limit=12.0
2024-09-25 09:44:40,733 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.76 vs. limit=22.5
2024-09-25 09:44:46,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=716263.3333333334, ans=0.125
2024-09-25 09:44:49,480 INFO [train.py:1198] (1/4) Epoch 40, batch 1550, loss[loss=0.1955, ctc_loss=0.1259, cr_loss=0.348, over 17162.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1224, cr_loss=0.3394, over 3362732.86 frames. ], batch size: 45, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:45:06,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=716356.6666666666, ans=0.125
2024-09-25 09:45:13,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=716356.6666666666, ans=0.125
2024-09-25 09:45:18,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=716356.6666666666, ans=0.125
2024-09-25 09:45:25,251 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.20 vs. limit=22.5
2024-09-25 09:45:31,428 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=716403.3333333334, ans=0.0
2024-09-25 09:45:31,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=716403.3333333334, ans=0.2
2024-09-25 09:45:40,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=716450.0, ans=0.125
2024-09-25 09:45:42,232 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.286e+02 1.352e+02 1.488e+02 2.645e+02, threshold=2.703e+02, percent-clipped=1.0
2024-09-25 09:45:45,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=716450.0, ans=0.09899494936611666
2024-09-25 09:45:47,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=716450.0, ans=0.125
2024-09-25 09:46:09,313 INFO [train.py:1198] (1/4) Epoch 40, batch 1600, loss[loss=0.1984, ctc_loss=0.1269, cr_loss=0.3575, over 15922.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1223, cr_loss=0.3393, over 3354578.10 frames. ], batch size: 74, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:46:22,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=716543.3333333334, ans=0.125
2024-09-25 09:46:43,434 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=17.95 vs. limit=22.5
2024-09-25 09:47:32,134 INFO [train.py:1198] (1/4) Epoch 40, batch 1650, loss[loss=0.1773, ctc_loss=0.1133, cr_loss=0.3202, over 17221.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1224, cr_loss=0.3394, over 3350393.01 frames. ], batch size: 50, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:47:54,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=716823.3333333334, ans=0.0
2024-09-25 09:48:27,241 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.303e+02 1.369e+02 1.457e+02 1.993e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-25 09:48:56,878 INFO [train.py:1198] (1/4) Epoch 40, batch 1700, loss[loss=0.1957, ctc_loss=0.1253, cr_loss=0.3518, over 17219.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1231, cr_loss=0.3408, over 3356739.53 frames. ], batch size: 50, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:49:05,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=717010.0, ans=0.125
2024-09-25 09:49:09,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=717010.0, ans=0.125
2024-09-25 09:49:17,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=717056.6666666666, ans=0.0
2024-09-25 09:49:33,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=717103.3333333334, ans=0.1
2024-09-25 09:49:49,060 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=717150.0, ans=0.2
2024-09-25 09:50:07,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=717196.6666666666, ans=0.0
2024-09-25 09:50:18,902 INFO [train.py:1198] (1/4) Epoch 40, batch 1750, loss[loss=0.2093, ctc_loss=0.135, cr_loss=0.3716, over 15073.00 frames. ], tot_loss[loss=0.1905, ctc_loss=0.1226, cr_loss=0.3395, over 3356329.23 frames. ], batch size: 89, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:50:35,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=717290.0, ans=0.0
2024-09-25 09:50:35,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=5.15 vs. limit=15.0
2024-09-25 09:50:40,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=717290.0, ans=0.025
2024-09-25 09:51:02,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=717336.6666666666, ans=0.0
2024-09-25 09:51:04,599 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.99 vs. limit=15.0
2024-09-25 09:51:11,756 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.261e+02 1.332e+02 1.446e+02 2.217e+02, threshold=2.663e+02, percent-clipped=0.0
2024-09-25 09:51:12,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=717383.3333333334, ans=0.07
2024-09-25 09:51:16,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=717383.3333333334, ans=0.125
2024-09-25 09:51:17,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=717383.3333333334, ans=0.125
2024-09-25 09:51:31,098 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.78 vs. limit=15.0
2024-09-25 09:51:37,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=717430.0, ans=0.0
2024-09-25 09:51:41,753 INFO [train.py:1198] (1/4) Epoch 40, batch 1800, loss[loss=0.2031, ctc_loss=0.133, cr_loss=0.3504, over 17220.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1219, cr_loss=0.3388, over 3366283.07 frames. ], batch size: 50, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:52:09,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=717523.3333333334, ans=0.0
2024-09-25 09:52:12,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=717570.0, ans=0.025
2024-09-25 09:53:01,965 INFO [train.py:1198] (1/4) Epoch 40, batch 1850, loss[loss=0.1962, ctc_loss=0.1272, cr_loss=0.3453, over 17165.00 frames. ], tot_loss[loss=0.1901, ctc_loss=0.1223, cr_loss=0.339, over 3368637.38 frames. ], batch size: 45, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:53:03,950 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 09:53:03,985 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=717710.0, ans=0.0
2024-09-25 09:53:10,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=717710.0, ans=0.0
2024-09-25 09:53:15,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=717710.0, ans=10.0
2024-09-25 09:53:27,475 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.38 vs. limit=22.5
2024-09-25 09:53:50,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=717803.3333333334, ans=0.125
2024-09-25 09:53:52,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=717803.3333333334, ans=0.0
2024-09-25 09:53:57,921 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.93 vs. limit=8.0
2024-09-25 09:54:00,035 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.246e+02 1.343e+02 1.431e+02 1.891e+02, threshold=2.686e+02, percent-clipped=0.0
2024-09-25 09:54:20,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=717896.6666666666, ans=0.2
2024-09-25 09:54:20,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=717896.6666666666, ans=0.0
2024-09-25 09:54:23,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=717896.6666666666, ans=0.0
2024-09-25 09:54:27,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=717896.6666666666, ans=0.125
2024-09-25 09:54:29,717 INFO [train.py:1198] (1/4) Epoch 40, batch 1900, loss[loss=0.1925, ctc_loss=0.1219, cr_loss=0.3526, over 17046.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1231, cr_loss=0.341, over 3363875.91 frames. ], batch size: 52, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:54:33,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=717943.3333333334, ans=0.125
2024-09-25 09:54:36,866 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.36 vs. limit=10.0
2024-09-25 09:54:38,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=717943.3333333334, ans=0.125
2024-09-25 09:55:05,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=718036.6666666666, ans=0.125
2024-09-25 09:55:20,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=718083.3333333334, ans=0.0
2024-09-25 09:55:24,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=718083.3333333334, ans=0.0
2024-09-25 09:55:32,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=718130.0, ans=0.0
2024-09-25 09:55:40,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=718130.0, ans=0.125
2024-09-25 09:55:41,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=718130.0, ans=0.125
2024-09-25 09:55:49,712 INFO [train.py:1198] (1/4) Epoch 40, batch 1950, loss[loss=0.198, ctc_loss=0.1288, cr_loss=0.3458, over 17023.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.123, cr_loss=0.3409, over 3366997.92 frames. ], batch size: 56, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:56:32,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=718270.0, ans=0.125
2024-09-25 09:56:45,247 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.303e+02 1.365e+02 1.463e+02 2.159e+02, threshold=2.730e+02, percent-clipped=0.0
2024-09-25 09:56:51,016 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=5.50 vs. limit=15.0
2024-09-25 09:57:01,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=718363.3333333334, ans=0.2
2024-09-25 09:57:12,466 INFO [train.py:1198] (1/4) Epoch 40, batch 2000, loss[loss=0.1818, ctc_loss=0.1147, cr_loss=0.3354, over 17153.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1222, cr_loss=0.3398, over 3376340.61 frames. ], batch size: 45, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:57:22,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=718410.0, ans=0.0
2024-09-25 09:57:26,533 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.81 vs. limit=10.0
2024-09-25 09:57:33,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=718456.6666666666, ans=0.125
2024-09-25 09:57:43,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=718503.3333333334, ans=0.2
2024-09-25 09:57:57,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=718503.3333333334, ans=0.2
2024-09-25 09:58:32,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=718596.6666666666, ans=0.125
2024-09-25 09:58:37,833 INFO [train.py:1198] (1/4) Epoch 40, batch 2050, loss[loss=0.1764, ctc_loss=0.111, cr_loss=0.3272, over 17252.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1215, cr_loss=0.3381, over 3372043.11 frames. ], batch size: 44, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 09:58:52,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=718690.0, ans=0.125
2024-09-25 09:59:04,214 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=718690.0, ans=15.0
2024-09-25 09:59:05,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=718690.0, ans=0.125
2024-09-25 09:59:25,403 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=718736.6666666666, ans=0.1
2024-09-25 09:59:33,095 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.276e+02 1.355e+02 1.490e+02 2.224e+02, threshold=2.711e+02, percent-clipped=0.0
2024-09-25 09:59:43,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.17 vs. limit=12.0
2024-09-25 09:59:52,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=718830.0, ans=0.125
2024-09-25 10:00:00,330 INFO [train.py:1198] (1/4) Epoch 40, batch 2100, loss[loss=0.1538, ctc_loss=0.09588, cr_loss=0.2897, over 17138.00 frames. ], tot_loss[loss=0.1894, ctc_loss=0.1218, cr_loss=0.3383, over 3368317.98 frames. ], batch size: 40, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 10:00:07,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=718876.6666666666, ans=0.0
2024-09-25 10:00:13,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=718876.6666666666, ans=0.1
2024-09-25 10:00:18,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=718923.3333333334, ans=0.0
2024-09-25 10:00:34,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=718970.0, ans=0.125
2024-09-25 10:01:20,395 INFO [train.py:1198] (1/4) Epoch 40, batch 2150, loss[loss=0.2282, ctc_loss=0.1479, cr_loss=0.4016, over 16942.00 frames. ], tot_loss[loss=0.1901, ctc_loss=0.1222, cr_loss=0.3394, over 3369607.42 frames. ], batch size: 58, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 10:01:27,910 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=719110.0, ans=0.125
2024-09-25 10:01:28,239 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.06 vs. limit=15.0
2024-09-25 10:01:45,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=719156.6666666666, ans=0.125
2024-09-25 10:02:16,124 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.284e+02 1.361e+02 1.483e+02 2.210e+02, threshold=2.722e+02, percent-clipped=0.0
2024-09-25 10:02:43,177 INFO [train.py:1198] (1/4) Epoch 40, batch 2200, loss[loss=0.1974, ctc_loss=0.1274, cr_loss=0.3501, over 17022.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1222, cr_loss=0.3388, over 3357300.00 frames. ], batch size: 51, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 10:03:02,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=719390.0, ans=0.125
2024-09-25 10:03:32,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=719483.3333333334, ans=0.125
2024-09-25 10:03:52,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=719530.0, ans=0.1
2024-09-25 10:03:59,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=719530.0, ans=0.0
2024-09-25 10:04:08,273 INFO [train.py:1198] (1/4) Epoch 40, batch 2250, loss[loss=0.1888, ctc_loss=0.1203, cr_loss=0.3424, over 17167.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1228, cr_loss=0.3398, over 3355841.10 frames. ], batch size: 45, lr: 2.97e-03, grad_scale: 32.0
2024-09-25 10:04:43,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=719670.0, ans=0.2
2024-09-25 10:05:04,026 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.272e+02 1.370e+02 1.440e+02 2.779e+02, threshold=2.741e+02, percent-clipped=1.0
2024-09-25 10:05:20,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=719763.3333333334, ans=0.125
2024-09-25 10:05:26,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=719763.3333333334, ans=0.125
2024-09-25 10:05:31,318 INFO [train.py:1198] (1/4) Epoch 40, batch 2300, loss[loss=0.2386, ctc_loss=0.162, cr_loss=0.3826, over 11836.00 frames. ], tot_loss[loss=0.1908, ctc_loss=0.1229, cr_loss=0.3394, over 3350617.43 frames. ], batch size: 124, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:06:31,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=719950.0, ans=0.125
2024-09-25 10:06:54,045 INFO [train.py:1198] (1/4) Epoch 40, batch 2350, loss[loss=0.1978, ctc_loss=0.1279, cr_loss=0.3492, over 17207.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1231, cr_loss=0.3398, over 3346600.84 frames. ], batch size: 55, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:07:15,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=720090.0, ans=0.125
2024-09-25 10:07:46,592 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.272e+02 1.331e+02 1.456e+02 1.927e+02, threshold=2.662e+02, percent-clipped=0.0
2024-09-25 10:07:58,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=720230.0, ans=0.2
2024-09-25 10:08:16,523 INFO [train.py:1198] (1/4) Epoch 40, batch 2400, loss[loss=0.1967, ctc_loss=0.1267, cr_loss=0.3497, over 17031.00 frames. ], tot_loss[loss=0.1928, ctc_loss=0.1244, cr_loss=0.3422, over 3341691.96 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:08:23,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=720276.6666666666, ans=0.125
2024-09-25 10:08:41,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=720323.3333333334, ans=0.1
2024-09-25 10:08:48,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=720323.3333333334, ans=0.1
2024-09-25 10:09:24,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=720463.3333333334, ans=0.2
2024-09-25 10:09:27,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=720463.3333333334, ans=0.125
2024-09-25 10:09:41,532 INFO [train.py:1198] (1/4) Epoch 40, batch 2450, loss[loss=0.2212, ctc_loss=0.143, cr_loss=0.3911, over 16825.00 frames. ], tot_loss[loss=0.1935, ctc_loss=0.1248, cr_loss=0.3433, over 3342607.18 frames. ], batch size: 61, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:10:03,046 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 10:10:11,756 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.37 vs. limit=22.5
2024-09-25 10:10:20,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=720603.3333333334, ans=0.1
2024-09-25 10:10:23,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=720603.3333333334, ans=0.125
2024-09-25 10:10:34,963 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.276e+02 1.362e+02 1.464e+02 1.936e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-25 10:10:41,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=720650.0, ans=0.0
2024-09-25 10:10:54,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=720696.6666666666, ans=0.125
2024-09-25 10:11:02,047 INFO [train.py:1198] (1/4) Epoch 40, batch 2500, loss[loss=0.2113, ctc_loss=0.1395, cr_loss=0.3591, over 17219.00 frames. ], tot_loss[loss=0.1932, ctc_loss=0.1246, cr_loss=0.343, over 3346966.23 frames. ], batch size: 55, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:11:32,947 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.06 vs. limit=15.0
2024-09-25 10:11:48,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=8.41 vs. limit=12.0
2024-09-25 10:11:53,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.79 vs. limit=15.0
2024-09-25 10:11:56,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=720883.3333333334, ans=0.125
2024-09-25 10:11:59,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=720883.3333333334, ans=0.2
2024-09-25 10:11:59,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=2.99 vs. limit=15.0
2024-09-25 10:12:00,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.39 vs. limit=15.0
2024-09-25 10:12:17,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=720930.0, ans=10.0
2024-09-25 10:12:20,975 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.46 vs. limit=15.0
2024-09-25 10:12:23,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=720976.6666666666, ans=0.5
2024-09-25 10:12:24,938 INFO [train.py:1198] (1/4) Epoch 40, batch 2550, loss[loss=0.1727, ctc_loss=0.1122, cr_loss=0.3028, over 17319.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1241, cr_loss=0.3425, over 3349354.27 frames. ], batch size: 51, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:12:41,942 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.71 vs. limit=6.0
2024-09-25 10:13:20,274 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.334e+02 1.455e+02 1.593e+02 2.101e+02, threshold=2.910e+02, percent-clipped=0.0
2024-09-25 10:13:48,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=721210.0, ans=0.2
2024-09-25 10:13:50,327 INFO [train.py:1198] (1/4) Epoch 40, batch 2600, loss[loss=0.2026, ctc_loss=0.1322, cr_loss=0.3522, over 17221.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3418, over 3347620.00 frames. ], batch size: 47, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:14:04,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=721210.0, ans=0.125
2024-09-25 10:14:37,135 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=5.10 vs. limit=15.0
2024-09-25 10:14:41,858 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.02 vs. limit=12.0
2024-09-25 10:14:49,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=721350.0, ans=0.0
2024-09-25 10:15:13,246 INFO [train.py:1198] (1/4) Epoch 40, batch 2650, loss[loss=0.2103, ctc_loss=0.1364, cr_loss=0.3696, over 17003.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1241, cr_loss=0.3419, over 3349843.97 frames. ], batch size: 53, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:15:25,206 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 10:15:30,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=721490.0, ans=0.0
2024-09-25 10:15:50,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=721536.6666666666, ans=0.125
2024-09-25 10:15:57,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=721536.6666666666, ans=0.0
2024-09-25 10:16:06,673 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.318e+02 1.393e+02 1.493e+02 1.834e+02, threshold=2.785e+02, percent-clipped=0.0
2024-09-25 10:16:36,435 INFO [train.py:1198] (1/4) Epoch 40, batch 2700, loss[loss=0.1766, ctc_loss=0.1119, cr_loss=0.3239, over 17096.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.1231, cr_loss=0.3393, over 3352371.96 frames. ], batch size: 43, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:16:59,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=721723.3333333334, ans=0.125
2024-09-25 10:17:02,592 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 10:17:05,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=721723.3333333334, ans=0.0
2024-09-25 10:17:39,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=721863.3333333334, ans=0.025
2024-09-25 10:17:48,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=721863.3333333334, ans=0.025
2024-09-25 10:17:56,551 INFO [train.py:1198] (1/4) Epoch 40, batch 2750, loss[loss=0.1817, ctc_loss=0.1171, cr_loss=0.3231, over 17027.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1237, cr_loss=0.3405, over 3353257.70 frames. ], batch size: 51, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:18:23,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=721956.6666666666, ans=0.125
2024-09-25 10:18:35,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=722003.3333333334, ans=0.125
2024-09-25 10:18:41,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=722003.3333333334, ans=0.2
2024-09-25 10:18:44,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=722003.3333333334, ans=0.125
2024-09-25 10:18:54,279 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.293e+02 1.370e+02 1.533e+02 1.979e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-25 10:19:20,242 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.66 vs. limit=15.0
2024-09-25 10:19:24,032 INFO [train.py:1198] (1/4) Epoch 40, batch 2800, loss[loss=0.1567, ctc_loss=0.09935, cr_loss=0.2865, over 17102.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1223, cr_loss=0.3382, over 3365353.18 frames. ], batch size: 43, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:19:24,274 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=722143.3333333334, ans=0.1
2024-09-25 10:19:30,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=722143.3333333334, ans=0.1
2024-09-25 10:19:34,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=722143.3333333334, ans=22.5
2024-09-25 10:19:47,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=6.23 vs. limit=15.0
2024-09-25 10:19:59,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=722236.6666666666, ans=0.125
2024-09-25 10:20:04,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=722236.6666666666, ans=0.0
2024-09-25 10:20:44,360 INFO [train.py:1198] (1/4) Epoch 40, batch 2850, loss[loss=0.1967, ctc_loss=0.1304, cr_loss=0.3316, over 17116.00 frames. ], tot_loss[loss=0.1898, ctc_loss=0.1222, cr_loss=0.3381, over 3357692.00 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:20:53,215 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=8.21 vs. limit=12.0
2024-09-25 10:20:59,702 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.26 vs. limit=12.0
2024-09-25 10:21:05,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=722423.3333333334, ans=0.2
2024-09-25 10:21:12,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=722423.3333333334, ans=0.0
2024-09-25 10:21:13,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=722423.3333333334, ans=0.125
2024-09-25 10:21:40,028 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.304e+02 1.350e+02 1.486e+02 2.860e+02, threshold=2.699e+02, percent-clipped=1.0
2024-09-25 10:21:49,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=722563.3333333334, ans=0.125
2024-09-25 10:21:56,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.93 vs. limit=15.0
2024-09-25 10:22:07,585 INFO [train.py:1198] (1/4) Epoch 40, batch 2900, loss[loss=0.1675, ctc_loss=0.1046, cr_loss=0.3144, over 16933.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1217, cr_loss=0.3375, over 3360761.04 frames. ], batch size: 42, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:22:38,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=722703.3333333334, ans=0.125
2024-09-25 10:22:58,305 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.41 vs. limit=15.0
2024-09-25 10:23:09,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=722750.0, ans=0.125
2024-09-25 10:23:16,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=722796.6666666666, ans=0.125
2024-09-25 10:23:28,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=722796.6666666666, ans=0.125
2024-09-25 10:23:32,769 INFO [train.py:1198] (1/4) Epoch 40, batch 2950, loss[loss=0.1611, ctc_loss=0.1022, cr_loss=0.2942, over 16959.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1226, cr_loss=0.3389, over 3357178.27 frames. ], batch size: 42, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:23:33,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=722843.3333333334, ans=0.125
2024-09-25 10:24:01,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=722890.0, ans=0.025
2024-09-25 10:24:26,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=722983.3333333334, ans=0.2
2024-09-25 10:24:27,664 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.302e+02 1.375e+02 1.472e+02 2.387e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-25 10:24:53,998 INFO [train.py:1198] (1/4) Epoch 40, batch 3000, loss[loss=0.1613, ctc_loss=0.09881, cr_loss=0.3124, over 16953.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1237, cr_loss=0.3413, over 3357498.53 frames. ], batch size: 42, lr: 2.96e-03, grad_scale: 32.0
2024-09-25 10:24:53,999 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 10:25:09,261 INFO [train.py:1230] (1/4) Epoch 40, validation: loss=0.03571, ctc_loss=0.03571, cr_loss=9.785e-15, over 944034.00 frames. 
2024-09-25 10:25:09,262 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 10:25:51,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=723170.0, ans=0.0
2024-09-25 10:26:22,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=723263.3333333334, ans=0.0
2024-09-25 10:26:27,243 INFO [train.py:1198] (1/4) Epoch 40, batch 3050, loss[loss=0.2102, ctc_loss=0.1407, cr_loss=0.3474, over 16099.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.124, cr_loss=0.3413, over 3343382.23 frames. ], batch size: 74, lr: 2.96e-03, grad_scale: 16.0
2024-09-25 10:26:37,678 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.30 vs. limit=15.0
2024-09-25 10:26:48,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=723356.6666666666, ans=0.125
2024-09-25 10:26:55,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=723356.6666666666, ans=0.125
2024-09-25 10:26:55,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=723356.6666666666, ans=0.1
2024-09-25 10:26:57,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=723403.3333333334, ans=0.125
2024-09-25 10:27:20,627 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.265e+02 1.358e+02 1.442e+02 1.711e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 10:27:25,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=723450.0, ans=0.125
2024-09-25 10:27:42,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=723496.6666666666, ans=0.0
2024-09-25 10:27:45,678 INFO [train.py:1198] (1/4) Epoch 40, batch 3100, loss[loss=0.1775, ctc_loss=0.1128, cr_loss=0.3238, over 17092.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.1239, cr_loss=0.3418, over 3355836.39 frames. ], batch size: 43, lr: 2.96e-03, grad_scale: 16.0
2024-09-25 10:27:53,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=723543.3333333334, ans=0.0
2024-09-25 10:27:59,174 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.56 vs. limit=10.0
2024-09-25 10:28:01,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=723590.0, ans=0.04949747468305833
2024-09-25 10:28:43,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=723683.3333333334, ans=0.0
2024-09-25 10:28:55,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.54 vs. limit=22.5
2024-09-25 10:29:06,037 INFO [train.py:1198] (1/4) Epoch 40, batch 3150, loss[loss=0.2104, ctc_loss=0.1376, cr_loss=0.3641, over 16662.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1238, cr_loss=0.3409, over 3351882.17 frames. ], batch size: 61, lr: 2.96e-03, grad_scale: 16.0
2024-09-25 10:29:17,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=723776.6666666666, ans=0.2
2024-09-25 10:29:19,529 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.90 vs. limit=22.5
2024-09-25 10:29:29,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=723823.3333333334, ans=0.0
2024-09-25 10:29:31,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=723823.3333333334, ans=0.1
2024-09-25 10:29:51,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=723916.6666666666, ans=0.125
2024-09-25 10:29:55,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten.whitening_limit, batch_count=723916.6666666666, ans=15.0
2024-09-25 10:30:00,971 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.278e+02 1.373e+02 1.497e+02 1.845e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-25 10:30:22,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=723963.3333333334, ans=0.95
2024-09-25 10:30:24,821 INFO [train.py:1198] (1/4) Epoch 40, batch 3200, loss[loss=0.1841, ctc_loss=0.1176, cr_loss=0.3325, over 17156.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1245, cr_loss=0.3428, over 3354114.99 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 16.0
2024-09-25 10:30:53,854 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.74 vs. limit=12.0
2024-09-25 10:31:01,638 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.93 vs. limit=6.0
2024-09-25 10:31:03,386 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.84 vs. limit=15.0
2024-09-25 10:31:22,044 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.46 vs. limit=15.0
2024-09-25 10:31:24,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=724150.0, ans=0.1
2024-09-25 10:31:27,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=724196.6666666666, ans=0.125
2024-09-25 10:31:43,139 INFO [train.py:1198] (1/4) Epoch 40, batch 3250, loss[loss=0.2533, ctc_loss=0.1728, cr_loss=0.4026, over 11494.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1244, cr_loss=0.3418, over 3338401.52 frames. ], batch size: 123, lr: 2.96e-03, grad_scale: 16.0
2024-09-25 10:32:21,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=724336.6666666666, ans=0.1
2024-09-25 10:32:39,474 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 10:32:42,130 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.319e+02 1.415e+02 1.534e+02 3.607e+02, threshold=2.830e+02, percent-clipped=1.0
2024-09-25 10:32:56,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=724430.0, ans=0.0
2024-09-25 10:33:05,492 INFO [train.py:1198] (1/4) Epoch 40, batch 3300, loss[loss=0.2016, ctc_loss=0.1316, cr_loss=0.35, over 15065.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1238, cr_loss=0.3407, over 3347067.53 frames. ], batch size: 88, lr: 2.96e-03, grad_scale: 16.0
2024-09-25 10:33:15,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=724476.6666666666, ans=0.09899494936611666
2024-09-25 10:33:39,321 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.49 vs. limit=15.0
2024-09-25 10:33:42,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=724570.0, ans=0.125
2024-09-25 10:33:58,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=724616.6666666666, ans=0.1
2024-09-25 10:34:22,777 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.70 vs. limit=15.0
2024-09-25 10:34:25,461 INFO [train.py:1198] (1/4) Epoch 40, batch 3350, loss[loss=0.1807, ctc_loss=0.1166, cr_loss=0.3205, over 17334.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1231, cr_loss=0.3399, over 3353329.85 frames. ], batch size: 52, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:34:31,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=724710.0, ans=0.125
2024-09-25 10:34:36,888 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 10:35:19,910 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.291e+02 1.364e+02 1.477e+02 1.997e+02, threshold=2.727e+02, percent-clipped=0.0
2024-09-25 10:35:37,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=724896.6666666666, ans=0.0
2024-09-25 10:35:43,179 INFO [train.py:1198] (1/4) Epoch 40, batch 3400, loss[loss=0.206, ctc_loss=0.1335, cr_loss=0.3624, over 17144.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1235, cr_loss=0.3401, over 3349597.38 frames. ], batch size: 48, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:35:45,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.38 vs. limit=15.0
2024-09-25 10:36:03,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=724990.0, ans=0.1
2024-09-25 10:36:05,844 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.83 vs. limit=15.0
2024-09-25 10:36:06,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=724990.0, ans=0.09899494936611666
2024-09-25 10:36:07,124 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=19.77 vs. limit=22.5
2024-09-25 10:36:16,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=725036.6666666666, ans=0.2
2024-09-25 10:36:55,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=725130.0, ans=0.125
2024-09-25 10:37:01,219 INFO [train.py:1198] (1/4) Epoch 40, batch 3450, loss[loss=0.1849, ctc_loss=0.1202, cr_loss=0.3235, over 17099.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1238, cr_loss=0.3402, over 3347611.76 frames. ], batch size: 43, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:37:06,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.65 vs. limit=10.0
2024-09-25 10:37:14,795 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.91 vs. limit=6.0
2024-09-25 10:37:19,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=725223.3333333334, ans=0.0
2024-09-25 10:37:22,661 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.54 vs. limit=15.0
2024-09-25 10:37:22,794 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.41 vs. limit=22.5
2024-09-25 10:37:25,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=725223.3333333334, ans=0.125
2024-09-25 10:37:53,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=725316.6666666666, ans=0.125
2024-09-25 10:37:56,649 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.302e+02 1.377e+02 1.496e+02 2.659e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-25 10:38:22,231 INFO [train.py:1198] (1/4) Epoch 40, batch 3500, loss[loss=0.1888, ctc_loss=0.1198, cr_loss=0.3448, over 16319.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.124, cr_loss=0.3405, over 3354234.43 frames. ], batch size: 36, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:38:30,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=725410.0, ans=0.2
2024-09-25 10:38:31,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=725410.0, ans=0.025
2024-09-25 10:38:33,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=725410.0, ans=0.1
2024-09-25 10:39:15,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=725550.0, ans=0.0
2024-09-25 10:39:23,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=725596.6666666666, ans=0.0
2024-09-25 10:39:26,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=725596.6666666666, ans=0.5
2024-09-25 10:39:27,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=725596.6666666666, ans=0.2
2024-09-25 10:39:39,941 INFO [train.py:1198] (1/4) Epoch 40, batch 3550, loss[loss=0.2251, ctc_loss=0.1504, cr_loss=0.3736, over 15159.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1238, cr_loss=0.3407, over 3354811.40 frames. ], batch size: 89, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:39:46,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=725643.3333333334, ans=0.05
2024-09-25 10:39:57,429 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=725690.0, ans=0.0
2024-09-25 10:39:57,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.54 vs. limit=15.0
2024-09-25 10:40:02,734 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.18 vs. limit=15.0
2024-09-25 10:40:08,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=725690.0, ans=0.07
2024-09-25 10:40:13,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=725736.6666666666, ans=0.0
2024-09-25 10:40:28,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=725783.3333333334, ans=0.09899494936611666
2024-09-25 10:40:34,532 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.284e+02 1.346e+02 1.448e+02 2.075e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-25 10:40:55,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=725830.0, ans=0.125
2024-09-25 10:40:58,150 INFO [train.py:1198] (1/4) Epoch 40, batch 3600, loss[loss=0.1842, ctc_loss=0.1158, cr_loss=0.3421, over 17261.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1233, cr_loss=0.3405, over 3356854.11 frames. ], batch size: 44, lr: 2.95e-03, grad_scale: 32.0
2024-09-25 10:40:58,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=725876.6666666666, ans=0.125
2024-09-25 10:41:42,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=725970.0, ans=0.125
2024-09-25 10:41:47,797 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.57 vs. limit=15.0
2024-09-25 10:42:20,144 INFO [train.py:1198] (1/4) Epoch 40, batch 3650, loss[loss=0.2231, ctc_loss=0.1488, cr_loss=0.3715, over 16400.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1224, cr_loss=0.3385, over 3355771.23 frames. ], batch size: 66, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:42:24,416 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=12.43 vs. limit=15.0
2024-09-25 10:42:39,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=726156.6666666666, ans=0.1
2024-09-25 10:42:47,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=726156.6666666666, ans=0.125
2024-09-25 10:43:07,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=726203.3333333334, ans=0.125
2024-09-25 10:43:19,107 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.298e+02 1.359e+02 1.454e+02 1.962e+02, threshold=2.718e+02, percent-clipped=0.0
2024-09-25 10:43:32,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=726296.6666666666, ans=0.125
2024-09-25 10:43:38,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=726296.6666666666, ans=0.0
2024-09-25 10:43:41,336 INFO [train.py:1198] (1/4) Epoch 40, batch 3700, loss[loss=0.2086, ctc_loss=0.1352, cr_loss=0.3674, over 17033.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1233, cr_loss=0.3402, over 3337357.39 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:43:54,639 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.68 vs. limit=12.0
2024-09-25 10:44:59,749 INFO [train.py:1198] (1/4) Epoch 40, batch 3750, loss[loss=0.1921, ctc_loss=0.1219, cr_loss=0.351, over 17022.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.124, cr_loss=0.3416, over 3343136.12 frames. ], batch size: 44, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:45:00,868 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.92 vs. limit=12.0
2024-09-25 10:45:15,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=726623.3333333334, ans=0.07
2024-09-25 10:45:18,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=726623.3333333334, ans=0.04949747468305833
2024-09-25 10:45:23,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=726623.3333333334, ans=0.125
2024-09-25 10:45:55,153 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.300e+02 1.354e+02 1.476e+02 2.861e+02, threshold=2.708e+02, percent-clipped=2.0
2024-09-25 10:46:00,804 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=5.10 vs. limit=15.0
2024-09-25 10:46:07,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=726763.3333333334, ans=0.0
2024-09-25 10:46:16,510 INFO [train.py:1198] (1/4) Epoch 40, batch 3800, loss[loss=0.2039, ctc_loss=0.1323, cr_loss=0.3579, over 16919.00 frames. ], tot_loss[loss=0.1925, ctc_loss=0.1242, cr_loss=0.3415, over 3320432.85 frames. ], batch size: 58, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:46:49,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=726903.3333333334, ans=0.0
2024-09-25 10:46:52,442 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 10:47:02,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.01 vs. limit=22.5
2024-09-25 10:47:09,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=726950.0, ans=0.125
2024-09-25 10:47:14,513 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.52 vs. limit=15.0
2024-09-25 10:47:22,561 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=7.15 vs. limit=15.0
2024-09-25 10:47:34,474 INFO [train.py:1198] (1/4) Epoch 40, batch 3850, loss[loss=0.2138, ctc_loss=0.1423, cr_loss=0.3575, over 11459.00 frames. ], tot_loss[loss=0.1936, ctc_loss=0.1251, cr_loss=0.3424, over 3278094.76 frames. ], batch size: 123, lr: 2.95e-03, grad_scale: 16.0
2024-09-25 10:47:37,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=727043.3333333334, ans=0.125
2024-09-25 10:47:45,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=727043.3333333334, ans=0.125
2024-09-25 10:47:48,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=727090.0, ans=0.125
2024-09-25 10:48:12,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=727136.6666666666, ans=0.125
2024-09-25 10:48:18,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=727136.6666666666, ans=10.0
2024-09-25 10:48:30,296 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.319e+02 1.427e+02 1.593e+02 2.504e+02, threshold=2.853e+02, percent-clipped=0.0
2024-09-25 10:48:35,859 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=6.08 vs. limit=10.0
2024-09-25 10:48:36,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=727230.0, ans=0.1
2024-09-25 10:48:38,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=727230.0, ans=0.0
2024-09-25 10:48:40,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=727230.0, ans=0.04949747468305833
2024-09-25 10:49:36,303 INFO [train.py:1198] (1/4) Epoch 41, batch 0, loss[loss=0.1669, ctc_loss=0.1061, cr_loss=0.3039, over 17078.00 frames. ], tot_loss[loss=0.1669, ctc_loss=0.1061, cr_loss=0.3039, over 17078.00 frames. ], batch size: 46, lr: 2.91e-03, grad_scale: 32.0
2024-09-25 10:49:36,303 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 10:49:51,734 INFO [train.py:1230] (1/4) Epoch 41, validation: loss=0.03537, ctc_loss=0.03537, cr_loss=1.035e-14, over 944034.00 frames. 
2024-09-25 10:49:51,734 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 10:49:52,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=727258.0, ans=0.125
2024-09-25 10:50:10,680 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=11.67 vs. limit=15.0
2024-09-25 10:50:10,894 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=7.67 vs. limit=15.0
2024-09-25 10:51:15,284 INFO [train.py:1198] (1/4) Epoch 41, batch 50, loss[loss=0.1873, ctc_loss=0.1216, cr_loss=0.3285, over 17137.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1232, cr_loss=0.3403, over 764128.42 frames. ], batch size: 45, lr: 2.91e-03, grad_scale: 32.0
2024-09-25 10:51:52,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=727584.6666666666, ans=0.125
2024-09-25 10:52:02,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=727631.3333333334, ans=0.2
2024-09-25 10:52:19,248 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.297e+02 1.379e+02 1.480e+02 1.921e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-25 10:52:27,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer_ff3.min_abs, batch_count=727678.0, ans=0.2
2024-09-25 10:52:30,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=727678.0, ans=0.125
2024-09-25 10:52:35,283 INFO [train.py:1198] (1/4) Epoch 41, batch 100, loss[loss=0.1839, ctc_loss=0.1212, cr_loss=0.3137, over 16999.00 frames. ], tot_loss[loss=0.1908, ctc_loss=0.1229, cr_loss=0.3395, over 1322011.17 frames. ], batch size: 56, lr: 2.91e-03, grad_scale: 32.0
2024-09-25 10:52:37,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=727724.6666666666, ans=0.125
2024-09-25 10:52:37,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=727724.6666666666, ans=0.0
2024-09-25 10:52:43,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=727724.6666666666, ans=0.125
2024-09-25 10:53:12,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=727818.0, ans=0.125
2024-09-25 10:53:34,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=727864.6666666666, ans=0.125
2024-09-25 10:53:43,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=727911.3333333334, ans=0.04949747468305833
2024-09-25 10:53:56,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=727911.3333333334, ans=0.125
2024-09-25 10:54:00,638 INFO [train.py:1198] (1/4) Epoch 41, batch 150, loss[loss=0.2298, ctc_loss=0.1482, cr_loss=0.4083, over 17053.00 frames. ], tot_loss[loss=0.1926, ctc_loss=0.1242, cr_loss=0.3419, over 1769666.67 frames. ], batch size: 56, lr: 2.91e-03, grad_scale: 32.0
2024-09-25 10:54:37,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=728051.3333333334, ans=0.125
2024-09-25 10:54:48,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.min_positive, batch_count=728051.3333333334, ans=0.025
2024-09-25 10:54:56,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=728098.0, ans=0.04949747468305833
2024-09-25 10:55:10,424 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.287e+02 1.355e+02 1.476e+02 1.968e+02, threshold=2.710e+02, percent-clipped=0.0
2024-09-25 10:55:27,696 INFO [train.py:1198] (1/4) Epoch 41, batch 200, loss[loss=0.1774, ctc_loss=0.1134, cr_loss=0.32, over 17227.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.124, cr_loss=0.3416, over 2118209.15 frames. ], batch size: 50, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 10:55:30,044 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.50 vs. limit=15.0
2024-09-25 10:56:44,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=728378.0, ans=0.025
2024-09-25 10:56:47,729 INFO [train.py:1198] (1/4) Epoch 41, batch 250, loss[loss=0.1706, ctc_loss=0.1075, cr_loss=0.3155, over 17134.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.123, cr_loss=0.3395, over 2391812.58 frames. ], batch size: 45, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 10:57:35,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=728564.6666666666, ans=0.125
2024-09-25 10:57:53,023 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.260e+02 1.343e+02 1.433e+02 1.845e+02, threshold=2.685e+02, percent-clipped=0.0
2024-09-25 10:58:07,361 INFO [train.py:1198] (1/4) Epoch 41, batch 300, loss[loss=0.1953, ctc_loss=0.1289, cr_loss=0.3321, over 17015.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1221, cr_loss=0.3379, over 2605992.71 frames. ], batch size: 51, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 10:58:15,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=728658.0, ans=0.0
2024-09-25 10:58:28,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=728704.6666666666, ans=0.0
2024-09-25 10:58:52,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=728751.3333333334, ans=0.0
2024-09-25 10:58:56,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=728751.3333333334, ans=0.125
2024-09-25 10:58:56,960 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=728751.3333333334, ans=0.125
2024-09-25 10:59:06,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=728798.0, ans=0.025
2024-09-25 10:59:09,838 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 10:59:25,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=728844.6666666666, ans=0.05
2024-09-25 10:59:32,119 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.16 vs. limit=15.0
2024-09-25 10:59:36,173 INFO [train.py:1198] (1/4) Epoch 41, batch 350, loss[loss=0.2239, ctc_loss=0.143, cr_loss=0.4047, over 17101.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1223, cr_loss=0.3388, over 2776959.18 frames. ], batch size: 49, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 10:59:54,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=728938.0, ans=0.125
2024-09-25 11:00:31,013 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.50 vs. limit=15.0
2024-09-25 11:00:44,742 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.270e+02 1.343e+02 1.483e+02 2.666e+02, threshold=2.685e+02, percent-clipped=0.0
2024-09-25 11:00:50,466 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.98 vs. limit=15.0
2024-09-25 11:00:59,198 INFO [train.py:1198] (1/4) Epoch 41, batch 400, loss[loss=0.191, ctc_loss=0.1187, cr_loss=0.3612, over 17246.00 frames. ], tot_loss[loss=0.1898, ctc_loss=0.1221, cr_loss=0.3387, over 2907546.94 frames. ], batch size: 44, lr: 2.91e-03, grad_scale: 32.0
2024-09-25 11:01:20,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=729171.3333333334, ans=0.0
2024-09-25 11:01:21,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=729171.3333333334, ans=0.0
2024-09-25 11:01:39,648 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=729218.0, ans=0.1
2024-09-25 11:01:53,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=729264.6666666666, ans=0.1
2024-09-25 11:02:17,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=729358.0, ans=0.125
2024-09-25 11:02:18,860 INFO [train.py:1198] (1/4) Epoch 41, batch 450, loss[loss=0.1633, ctc_loss=0.1038, cr_loss=0.2976, over 16282.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1219, cr_loss=0.3384, over 3011471.49 frames. ], batch size: 36, lr: 2.91e-03, grad_scale: 32.0
2024-09-25 11:02:22,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=729358.0, ans=0.125
2024-09-25 11:02:52,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=729451.3333333334, ans=0.125
2024-09-25 11:02:54,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=729451.3333333334, ans=0.025
2024-09-25 11:03:20,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=729498.0, ans=0.125
2024-09-25 11:03:26,326 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.254e+02 1.327e+02 1.463e+02 1.935e+02, threshold=2.654e+02, percent-clipped=0.0
2024-09-25 11:03:41,630 INFO [train.py:1198] (1/4) Epoch 41, batch 500, loss[loss=0.2496, ctc_loss=0.1683, cr_loss=0.4065, over 12148.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1219, cr_loss=0.3386, over 3081867.15 frames. ], batch size: 123, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 11:03:47,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=729591.3333333334, ans=0.125
2024-09-25 11:04:14,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=729638.0, ans=0.015
2024-09-25 11:04:45,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=729731.3333333334, ans=0.125
2024-09-25 11:05:09,614 INFO [train.py:1198] (1/4) Epoch 41, batch 550, loss[loss=0.2184, ctc_loss=0.1432, cr_loss=0.3758, over 17229.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1217, cr_loss=0.3381, over 3143091.64 frames. ], batch size: 55, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 11:05:22,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=729824.6666666666, ans=0.125
2024-09-25 11:05:24,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=729871.3333333334, ans=0.2
2024-09-25 11:05:50,737 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.29 vs. limit=22.5
2024-09-25 11:05:59,849 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 11:06:18,377 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.304e+02 1.377e+02 1.509e+02 2.527e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-25 11:06:29,789 INFO [train.py:1198] (1/4) Epoch 41, batch 600, loss[loss=0.1984, ctc_loss=0.1294, cr_loss=0.3449, over 17107.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1216, cr_loss=0.3378, over 3192810.54 frames. ], batch size: 49, lr: 2.91e-03, grad_scale: 8.0
2024-09-25 11:07:16,701 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=730198.0, ans=0.07
2024-09-25 11:07:21,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=730198.0, ans=0.0
2024-09-25 11:07:30,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=730198.0, ans=0.125
2024-09-25 11:07:47,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=730244.6666666666, ans=0.125
2024-09-25 11:07:49,972 INFO [train.py:1198] (1/4) Epoch 41, batch 650, loss[loss=0.1849, ctc_loss=0.1186, cr_loss=0.3316, over 17010.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1214, cr_loss=0.3373, over 3240852.32 frames. ], batch size: 51, lr: 2.91e-03, grad_scale: 8.0
2024-09-25 11:08:03,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=730291.3333333334, ans=0.0
2024-09-25 11:08:16,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=730338.0, ans=0.0
2024-09-25 11:08:16,369 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.20 vs. limit=22.5
2024-09-25 11:08:19,402 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=730338.0, ans=0.0
2024-09-25 11:08:23,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=730384.6666666666, ans=0.125
2024-09-25 11:09:06,926 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.289e+02 1.341e+02 1.440e+02 1.849e+02, threshold=2.683e+02, percent-clipped=0.0
2024-09-25 11:09:12,396 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.41 vs. limit=22.5
2024-09-25 11:09:18,185 INFO [train.py:1198] (1/4) Epoch 41, batch 700, loss[loss=0.1748, ctc_loss=0.1114, cr_loss=0.317, over 17166.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1209, cr_loss=0.3368, over 3274382.47 frames. ], batch size: 45, lr: 2.91e-03, grad_scale: 8.0
2024-09-25 11:09:52,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=730618.0, ans=0.125
2024-09-25 11:10:00,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=730618.0, ans=0.125
2024-09-25 11:10:11,298 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.75 vs. limit=22.5
2024-09-25 11:10:40,884 INFO [train.py:1198] (1/4) Epoch 41, batch 750, loss[loss=0.201, ctc_loss=0.132, cr_loss=0.3445, over 17018.00 frames. ], tot_loss[loss=0.1886, ctc_loss=0.1211, cr_loss=0.3378, over 3295230.96 frames. ], batch size: 51, lr: 2.91e-03, grad_scale: 8.0
2024-09-25 11:10:42,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=730758.0, ans=0.0
2024-09-25 11:11:49,459 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.291e+02 1.357e+02 1.455e+02 2.788e+02, threshold=2.714e+02, percent-clipped=1.0
2024-09-25 11:11:51,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=730944.6666666666, ans=0.125
2024-09-25 11:11:59,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=730991.3333333334, ans=0.0
2024-09-25 11:12:00,720 INFO [train.py:1198] (1/4) Epoch 41, batch 800, loss[loss=0.2169, ctc_loss=0.141, cr_loss=0.3794, over 17062.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1217, cr_loss=0.3395, over 3311468.34 frames. ], batch size: 52, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 11:12:07,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=730991.3333333334, ans=0.1
2024-09-25 11:12:15,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=731038.0, ans=0.0
2024-09-25 11:12:29,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=731038.0, ans=0.2
2024-09-25 11:13:20,135 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.25 vs. limit=15.0
2024-09-25 11:13:21,003 INFO [train.py:1198] (1/4) Epoch 41, batch 850, loss[loss=0.2069, ctc_loss=0.1352, cr_loss=0.3586, over 17296.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1217, cr_loss=0.3392, over 3332574.26 frames. ], batch size: 51, lr: 2.91e-03, grad_scale: 16.0
2024-09-25 11:13:24,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=731224.6666666666, ans=0.125
2024-09-25 11:13:50,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=731271.3333333334, ans=0.025
2024-09-25 11:14:29,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=731364.6666666666, ans=0.2
2024-09-25 11:14:34,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=731411.3333333334, ans=0.1
2024-09-25 11:14:34,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=731411.3333333334, ans=0.125
2024-09-25 11:14:35,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=731411.3333333334, ans=0.0
2024-09-25 11:14:37,169 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.274e+02 1.366e+02 1.473e+02 2.977e+02, threshold=2.732e+02, percent-clipped=1.0
2024-09-25 11:14:48,455 INFO [train.py:1198] (1/4) Epoch 41, batch 900, loss[loss=0.2013, ctc_loss=0.1315, cr_loss=0.3494, over 16923.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1217, cr_loss=0.3396, over 3340018.36 frames. ], batch size: 58, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:15:29,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=731551.3333333334, ans=0.0
2024-09-25 11:15:31,954 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.65 vs. limit=15.0
2024-09-25 11:15:46,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=731598.0, ans=0.0
2024-09-25 11:15:49,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=731598.0, ans=0.1
2024-09-25 11:16:10,880 INFO [train.py:1198] (1/4) Epoch 41, batch 950, loss[loss=0.2231, ctc_loss=0.1469, cr_loss=0.3806, over 17059.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1217, cr_loss=0.34, over 3354457.26 frames. ], batch size: 46, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:16:22,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=731691.3333333334, ans=0.04949747468305833
2024-09-25 11:16:28,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=731738.0, ans=0.125
2024-09-25 11:16:36,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=731738.0, ans=0.125
2024-09-25 11:17:10,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=731831.3333333334, ans=0.0
2024-09-25 11:17:19,942 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.310e+02 1.391e+02 1.460e+02 2.998e+02, threshold=2.782e+02, percent-clipped=1.0
2024-09-25 11:17:21,981 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 11:17:28,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=731878.0, ans=0.0
2024-09-25 11:17:31,153 INFO [train.py:1198] (1/4) Epoch 41, batch 1000, loss[loss=0.2282, ctc_loss=0.157, cr_loss=0.356, over 11948.00 frames. ], tot_loss[loss=0.1904, ctc_loss=0.1222, cr_loss=0.341, over 3347440.89 frames. ], batch size: 124, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:17:52,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=731971.3333333334, ans=0.1
2024-09-25 11:18:15,200 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=7.30 vs. limit=15.0
2024-09-25 11:18:22,276 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=732064.6666666666, ans=0.125
2024-09-25 11:18:22,751 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.75 vs. limit=15.0
2024-09-25 11:18:29,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=732064.6666666666, ans=0.125
2024-09-25 11:18:30,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=732064.6666666666, ans=0.0
2024-09-25 11:18:37,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=732111.3333333334, ans=0.025
2024-09-25 11:18:41,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=732111.3333333334, ans=0.0
2024-09-25 11:18:42,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=732111.3333333334, ans=0.125
2024-09-25 11:18:56,467 INFO [train.py:1198] (1/4) Epoch 41, batch 1050, loss[loss=0.169, ctc_loss=0.1113, cr_loss=0.2881, over 17055.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1216, cr_loss=0.3396, over 3354063.21 frames. ], batch size: 46, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:19:18,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=732204.6666666666, ans=0.0
2024-09-25 11:19:22,063 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=14.21 vs. limit=15.0
2024-09-25 11:19:47,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=732298.0, ans=0.125
2024-09-25 11:20:10,107 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.269e+02 1.332e+02 1.413e+02 1.822e+02, threshold=2.664e+02, percent-clipped=0.0
2024-09-25 11:20:21,421 INFO [train.py:1198] (1/4) Epoch 41, batch 1100, loss[loss=0.1714, ctc_loss=0.1082, cr_loss=0.3162, over 17023.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1216, cr_loss=0.3393, over 3354592.52 frames. ], batch size: 44, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:20:34,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=732391.3333333334, ans=0.125
2024-09-25 11:20:46,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=732438.0, ans=22.5
2024-09-25 11:20:49,128 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=732438.0, ans=0.2
2024-09-25 11:21:02,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=732484.6666666666, ans=0.125
2024-09-25 11:21:02,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=732484.6666666666, ans=0.125
2024-09-25 11:21:07,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=732484.6666666666, ans=0.125
2024-09-25 11:21:15,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=732531.3333333334, ans=0.125
2024-09-25 11:21:20,470 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.32 vs. limit=15.0
2024-09-25 11:21:21,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=732531.3333333334, ans=0.125
2024-09-25 11:21:21,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=732531.3333333334, ans=0.0
2024-09-25 11:21:21,722 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.28 vs. limit=22.5
2024-09-25 11:21:41,756 INFO [train.py:1198] (1/4) Epoch 41, batch 1150, loss[loss=0.1836, ctc_loss=0.117, cr_loss=0.333, over 17223.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1218, cr_loss=0.3395, over 3352870.73 frames. ], batch size: 50, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:21:42,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=732624.6666666666, ans=0.125
2024-09-25 11:21:49,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=732624.6666666666, ans=0.2
2024-09-25 11:21:51,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=732624.6666666666, ans=0.07
2024-09-25 11:22:07,953 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=732671.3333333334, ans=0.025
2024-09-25 11:22:42,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=732764.6666666666, ans=0.125
2024-09-25 11:22:44,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=732811.3333333334, ans=0.125
2024-09-25 11:22:50,423 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.312e+02 1.375e+02 1.453e+02 2.138e+02, threshold=2.750e+02, percent-clipped=0.0
2024-09-25 11:23:01,755 INFO [train.py:1198] (1/4) Epoch 41, batch 1200, loss[loss=0.1522, ctc_loss=0.09758, cr_loss=0.2734, over 16963.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1222, cr_loss=0.34, over 3341012.52 frames. ], batch size: 42, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:23:06,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=732858.0, ans=0.125
2024-09-25 11:23:13,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.44 vs. limit=10.0
2024-09-25 11:23:44,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=732951.3333333334, ans=0.0
2024-09-25 11:23:45,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=732951.3333333334, ans=0.2
2024-09-25 11:23:47,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=732951.3333333334, ans=0.0
2024-09-25 11:23:55,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=732998.0, ans=0.125
2024-09-25 11:24:17,718 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.39 vs. limit=15.0
2024-09-25 11:24:29,677 INFO [train.py:1198] (1/4) Epoch 41, batch 1250, loss[loss=0.1347, ctc_loss=0.08359, cr_loss=0.2556, over 17082.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.1222, cr_loss=0.3389, over 3340902.10 frames. ], batch size: 40, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:25:12,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=733184.6666666666, ans=0.125
2024-09-25 11:25:37,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=733278.0, ans=0.1
2024-09-25 11:25:40,678 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.293e+02 1.355e+02 1.454e+02 2.054e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-25 11:25:52,136 INFO [train.py:1198] (1/4) Epoch 41, batch 1300, loss[loss=0.1711, ctc_loss=0.1077, cr_loss=0.317, over 17166.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1222, cr_loss=0.3389, over 3347215.48 frames. ], batch size: 45, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:26:13,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=733371.3333333334, ans=0.2
2024-09-25 11:26:13,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=733371.3333333334, ans=0.125
2024-09-25 11:26:18,169 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 11:26:27,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=733418.0, ans=0.125
2024-09-25 11:26:34,779 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.65 vs. limit=15.0
2024-09-25 11:26:35,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=733418.0, ans=0.125
2024-09-25 11:26:42,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=733464.6666666666, ans=0.1
2024-09-25 11:26:46,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=733464.6666666666, ans=0.1
2024-09-25 11:26:57,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.45 vs. limit=15.0
2024-09-25 11:26:58,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.34 vs. limit=15.0
2024-09-25 11:27:00,256 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.73 vs. limit=10.0
2024-09-25 11:27:11,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=733558.0, ans=0.0
2024-09-25 11:27:12,355 INFO [train.py:1198] (1/4) Epoch 41, batch 1350, loss[loss=0.1783, ctc_loss=0.1112, cr_loss=0.3353, over 17079.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1227, cr_loss=0.34, over 3348272.94 frames. ], batch size: 43, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:27:12,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=733558.0, ans=0.125
2024-09-25 11:27:16,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=733558.0, ans=0.0
2024-09-25 11:27:40,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=733604.6666666666, ans=0.125
2024-09-25 11:28:20,787 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.331e+02 1.402e+02 1.512e+02 1.864e+02, threshold=2.804e+02, percent-clipped=0.0
2024-09-25 11:28:37,186 INFO [train.py:1198] (1/4) Epoch 41, batch 1400, loss[loss=0.1848, ctc_loss=0.1169, cr_loss=0.3394, over 17136.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1224, cr_loss=0.3394, over 3364081.60 frames. ], batch size: 48, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:28:45,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=733791.3333333334, ans=0.0
2024-09-25 11:29:03,172 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.74 vs. limit=12.0
2024-09-25 11:29:08,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=733838.0, ans=0.125
2024-09-25 11:29:15,520 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.54 vs. limit=15.0
2024-09-25 11:29:49,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=733978.0, ans=0.1
2024-09-25 11:30:00,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten.whitening_limit, batch_count=734024.6666666666, ans=22.5
2024-09-25 11:30:01,594 INFO [train.py:1198] (1/4) Epoch 41, batch 1450, loss[loss=0.2011, ctc_loss=0.1306, cr_loss=0.3527, over 16759.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1232, cr_loss=0.3414, over 3366751.32 frames. ], batch size: 61, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:30:36,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=734118.0, ans=0.0
2024-09-25 11:30:38,324 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 11:30:43,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=734118.0, ans=0.125
2024-09-25 11:30:55,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=734164.6666666666, ans=0.125
2024-09-25 11:30:55,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=734164.6666666666, ans=0.125
2024-09-25 11:31:07,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=734211.3333333334, ans=0.0
2024-09-25 11:31:10,081 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.281e+02 1.374e+02 1.455e+02 2.816e+02, threshold=2.747e+02, percent-clipped=1.0
2024-09-25 11:31:21,396 INFO [train.py:1198] (1/4) Epoch 41, batch 1500, loss[loss=0.1957, ctc_loss=0.1264, cr_loss=0.3469, over 17294.00 frames. ], tot_loss[loss=0.1901, ctc_loss=0.1222, cr_loss=0.3393, over 3373163.91 frames. ], batch size: 51, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:31:51,206 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.83 vs. limit=15.0
2024-09-25 11:32:28,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=734444.6666666666, ans=0.125
2024-09-25 11:32:38,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=734444.6666666666, ans=0.125
2024-09-25 11:32:41,150 INFO [train.py:1198] (1/4) Epoch 41, batch 1550, loss[loss=0.1742, ctc_loss=0.1129, cr_loss=0.3064, over 17310.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1214, cr_loss=0.3377, over 3373532.39 frames. ], batch size: 51, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:32:41,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=734491.3333333334, ans=0.125
2024-09-25 11:33:01,342 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.23 vs. limit=15.0
2024-09-25 11:33:07,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=734538.0, ans=0.2
2024-09-25 11:33:08,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=734538.0, ans=0.125
2024-09-25 11:33:12,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=734584.6666666666, ans=0.1
2024-09-25 11:33:15,252 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=734584.6666666666, ans=0.125
2024-09-25 11:33:59,384 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.278e+02 1.349e+02 1.424e+02 1.781e+02, threshold=2.698e+02, percent-clipped=0.0
2024-09-25 11:34:02,146 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.95 vs. limit=10.0
2024-09-25 11:34:06,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=734678.0, ans=0.1
2024-09-25 11:34:09,064 INFO [train.py:1198] (1/4) Epoch 41, batch 1600, loss[loss=0.2008, ctc_loss=0.1271, cr_loss=0.3686, over 17244.00 frames. ], tot_loss[loss=0.1898, ctc_loss=0.1221, cr_loss=0.3387, over 3374272.75 frames. ], batch size: 44, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:34:17,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=734724.6666666666, ans=0.125
2024-09-25 11:34:22,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=734724.6666666666, ans=0.125
2024-09-25 11:35:03,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=734864.6666666666, ans=0.125
2024-09-25 11:35:08,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=734864.6666666666, ans=0.1
2024-09-25 11:35:24,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=734911.3333333334, ans=0.1
2024-09-25 11:35:31,973 INFO [train.py:1198] (1/4) Epoch 41, batch 1650, loss[loss=0.2206, ctc_loss=0.1434, cr_loss=0.3858, over 16554.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1228, cr_loss=0.3393, over 3367527.37 frames. ], batch size: 66, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:35:41,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=734958.0, ans=0.125
2024-09-25 11:35:51,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=735004.6666666666, ans=0.0
2024-09-25 11:36:04,892 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.91 vs. limit=15.0
2024-09-25 11:36:09,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=735051.3333333334, ans=0.125
2024-09-25 11:36:29,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=735098.0, ans=0.125
2024-09-25 11:36:42,043 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.308e+02 1.358e+02 1.444e+02 2.176e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 11:36:51,846 INFO [train.py:1198] (1/4) Epoch 41, batch 1700, loss[loss=0.1708, ctc_loss=0.109, cr_loss=0.3092, over 17088.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1233, cr_loss=0.3406, over 3372195.33 frames. ], batch size: 40, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:36:52,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=735191.3333333334, ans=0.125
2024-09-25 11:36:53,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=735191.3333333334, ans=0.125
2024-09-25 11:36:55,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=735191.3333333334, ans=0.0
2024-09-25 11:37:16,771 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.03 vs. limit=15.0
2024-09-25 11:37:27,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=735284.6666666666, ans=0.125
2024-09-25 11:37:30,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=735284.6666666666, ans=0.2
2024-09-25 11:38:06,872 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.76 vs. limit=15.0
2024-09-25 11:38:12,454 INFO [train.py:1198] (1/4) Epoch 41, batch 1750, loss[loss=0.178, ctc_loss=0.1155, cr_loss=0.3125, over 17043.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1225, cr_loss=0.3391, over 3369279.62 frames. ], batch size: 39, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:38:41,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=735471.3333333334, ans=0.0
2024-09-25 11:38:43,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=735471.3333333334, ans=0.125
2024-09-25 11:38:58,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=735518.0, ans=0.1
2024-09-25 11:39:16,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=735564.6666666666, ans=0.2
2024-09-25 11:39:30,211 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.294e+02 1.365e+02 1.443e+02 2.325e+02, threshold=2.730e+02, percent-clipped=0.0
2024-09-25 11:39:32,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=19.94 vs. limit=22.5
2024-09-25 11:39:39,718 INFO [train.py:1198] (1/4) Epoch 41, batch 1800, loss[loss=0.1968, ctc_loss=0.1275, cr_loss=0.3468, over 16883.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1233, cr_loss=0.3408, over 3362786.54 frames. ], batch size: 58, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:39:44,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=735658.0, ans=0.0
2024-09-25 11:40:03,581 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.94 vs. limit=22.5
2024-09-25 11:40:17,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=735751.3333333334, ans=0.0
2024-09-25 11:40:32,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=735798.0, ans=0.0
2024-09-25 11:40:35,651 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=735798.0, ans=0.125
2024-09-25 11:40:36,257 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.31 vs. limit=15.0
2024-09-25 11:40:42,257 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.68 vs. limit=15.0
2024-09-25 11:40:48,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=735844.6666666666, ans=0.125
2024-09-25 11:40:59,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=735844.6666666666, ans=0.09899494936611666
2024-09-25 11:41:02,239 INFO [train.py:1198] (1/4) Epoch 41, batch 1850, loss[loss=0.1841, ctc_loss=0.1175, cr_loss=0.3332, over 17097.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.1228, cr_loss=0.3402, over 3364564.90 frames. ], batch size: 40, lr: 2.90e-03, grad_scale: 32.0
2024-09-25 11:41:13,861 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 11:41:16,279 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.63 vs. limit=5.0
2024-09-25 11:41:27,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=735938.0, ans=0.125
2024-09-25 11:41:32,658 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=735984.6666666666, ans=0.125
2024-09-25 11:41:53,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer_na.min_abs, batch_count=736031.3333333334, ans=0.02
2024-09-25 11:41:59,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=736031.3333333334, ans=0.125
2024-09-25 11:42:10,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=736078.0, ans=0.125
2024-09-25 11:42:13,780 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.264e+02 1.364e+02 1.473e+02 1.819e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-25 11:42:15,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=736078.0, ans=0.0
2024-09-25 11:42:21,750 INFO [train.py:1198] (1/4) Epoch 41, batch 1900, loss[loss=0.1942, ctc_loss=0.1214, cr_loss=0.3644, over 17157.00 frames. ], tot_loss[loss=0.1908, ctc_loss=0.1228, cr_loss=0.3401, over 3366129.48 frames. ], batch size: 45, lr: 2.90e-03, grad_scale: 16.0
2024-09-25 11:42:41,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=736171.3333333334, ans=0.0
2024-09-25 11:43:20,703 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 11:43:33,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=736311.3333333334, ans=0.0
2024-09-25 11:43:50,147 INFO [train.py:1198] (1/4) Epoch 41, batch 1950, loss[loss=0.2218, ctc_loss=0.1411, cr_loss=0.4035, over 17317.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.123, cr_loss=0.3414, over 3368373.05 frames. ], batch size: 49, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:43:54,363 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.26 vs. limit=15.0
2024-09-25 11:43:55,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=736358.0, ans=0.1
2024-09-25 11:44:03,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.37 vs. limit=15.0
2024-09-25 11:44:32,691 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.05 vs. limit=6.0
2024-09-25 11:44:54,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=736498.0, ans=0.0
2024-09-25 11:45:02,765 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.39 vs. limit=6.0
2024-09-25 11:45:05,174 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.301e+02 1.403e+02 1.536e+02 5.268e+02, threshold=2.807e+02, percent-clipped=2.0
2024-09-25 11:45:13,265 INFO [train.py:1198] (1/4) Epoch 41, batch 2000, loss[loss=0.1747, ctc_loss=0.1111, cr_loss=0.3184, over 17243.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1228, cr_loss=0.3414, over 3368556.63 frames. ], batch size: 44, lr: 2.89e-03, grad_scale: 32.0
2024-09-25 11:45:55,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=736684.6666666666, ans=0.0
2024-09-25 11:46:34,056 INFO [train.py:1198] (1/4) Epoch 41, batch 2050, loss[loss=0.2111, ctc_loss=0.1366, cr_loss=0.3729, over 17284.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1233, cr_loss=0.3415, over 3364920.19 frames. ], batch size: 51, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:46:47,808 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.97 vs. limit=15.0
2024-09-25 11:46:50,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=736871.3333333334, ans=0.2
2024-09-25 11:46:53,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=736871.3333333334, ans=0.125
2024-09-25 11:46:57,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer_ff2.min_abs, batch_count=736871.3333333334, ans=0.1
2024-09-25 11:46:58,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=736871.3333333334, ans=0.2
2024-09-25 11:47:47,792 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.289e+02 1.380e+02 1.485e+02 2.059e+02, threshold=2.761e+02, percent-clipped=0.0
2024-09-25 11:47:54,264 INFO [train.py:1198] (1/4) Epoch 41, batch 2100, loss[loss=0.1963, ctc_loss=0.1266, cr_loss=0.3485, over 17232.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1237, cr_loss=0.342, over 3366003.88 frames. ], batch size: 55, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:47:58,072 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.62 vs. limit=15.0
2024-09-25 11:48:24,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.37 vs. limit=22.5
2024-09-25 11:48:25,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=737104.6666666666, ans=0.025
2024-09-25 11:48:25,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=737104.6666666666, ans=0.125
2024-09-25 11:49:08,139 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=15.01 vs. limit=22.5
2024-09-25 11:49:21,659 INFO [train.py:1198] (1/4) Epoch 41, batch 2150, loss[loss=0.1879, ctc_loss=0.1217, cr_loss=0.3307, over 16935.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1237, cr_loss=0.3418, over 3368880.93 frames. ], batch size: 58, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:49:31,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=737291.3333333334, ans=0.2
2024-09-25 11:49:50,354 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=10.66 vs. limit=12.0
2024-09-25 11:50:12,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=737431.3333333334, ans=0.0
2024-09-25 11:50:23,428 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=737431.3333333334, ans=0.015
2024-09-25 11:50:37,767 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.274e+02 1.356e+02 1.503e+02 3.384e+02, threshold=2.711e+02, percent-clipped=1.0
2024-09-25 11:50:42,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=737524.6666666666, ans=0.1
2024-09-25 11:50:42,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=737524.6666666666, ans=0.125
2024-09-25 11:50:44,085 INFO [train.py:1198] (1/4) Epoch 41, batch 2200, loss[loss=0.2228, ctc_loss=0.1457, cr_loss=0.3858, over 17029.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1231, cr_loss=0.3405, over 3372323.65 frames. ], batch size: 53, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:51:13,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=737571.3333333334, ans=0.125
2024-09-25 11:51:13,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=737571.3333333334, ans=0.1
2024-09-25 11:51:55,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.55 vs. limit=6.0
2024-09-25 11:51:56,629 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=737711.3333333334, ans=0.04949747468305833
2024-09-25 11:51:56,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=737711.3333333334, ans=0.025
2024-09-25 11:52:04,267 INFO [train.py:1198] (1/4) Epoch 41, batch 2250, loss[loss=0.1951, ctc_loss=0.125, cr_loss=0.3501, over 17220.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1231, cr_loss=0.3408, over 3368371.08 frames. ], batch size: 50, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:52:12,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=737758.0, ans=0.0
2024-09-25 11:52:32,139 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.09 vs. limit=15.0
2024-09-25 11:52:52,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=737898.0, ans=0.125
2024-09-25 11:53:03,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=737898.0, ans=0.95
2024-09-25 11:53:21,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=737944.6666666666, ans=0.125
2024-09-25 11:53:22,627 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.265e+02 1.339e+02 1.413e+02 1.733e+02, threshold=2.679e+02, percent-clipped=0.0
2024-09-25 11:53:23,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=737944.6666666666, ans=0.125
2024-09-25 11:53:29,051 INFO [train.py:1198] (1/4) Epoch 41, batch 2300, loss[loss=0.1862, ctc_loss=0.1209, cr_loss=0.3263, over 15727.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1231, cr_loss=0.3402, over 3368801.57 frames. ], batch size: 74, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:53:39,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=737991.3333333334, ans=0.025
2024-09-25 11:54:53,982 INFO [train.py:1198] (1/4) Epoch 41, batch 2350, loss[loss=0.1698, ctc_loss=0.1065, cr_loss=0.3161, over 17117.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1228, cr_loss=0.3398, over 3360391.15 frames. ], batch size: 40, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:55:50,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=738364.6666666666, ans=0.125
2024-09-25 11:56:07,456 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.297e+02 1.370e+02 1.455e+02 1.687e+02, threshold=2.740e+02, percent-clipped=0.0
2024-09-25 11:56:07,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=738411.3333333334, ans=0.125
2024-09-25 11:56:12,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=738458.0, ans=0.5
2024-09-25 11:56:13,956 INFO [train.py:1198] (1/4) Epoch 41, batch 2400, loss[loss=0.2097, ctc_loss=0.1362, cr_loss=0.3672, over 16028.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1234, cr_loss=0.3402, over 3356250.48 frames. ], batch size: 74, lr: 2.89e-03, grad_scale: 32.0
2024-09-25 11:56:15,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=738458.0, ans=0.0
2024-09-25 11:56:17,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=738458.0, ans=0.1
2024-09-25 11:56:22,970 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.37 vs. limit=15.0
2024-09-25 11:56:44,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=738551.3333333334, ans=0.125
2024-09-25 11:56:54,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=738551.3333333334, ans=0.125
2024-09-25 11:56:55,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=738551.3333333334, ans=0.04949747468305833
2024-09-25 11:57:15,023 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=9.04 vs. limit=12.0
2024-09-25 11:57:17,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=738644.6666666666, ans=0.125
2024-09-25 11:57:33,333 INFO [train.py:1198] (1/4) Epoch 41, batch 2450, loss[loss=0.1825, ctc_loss=0.1186, cr_loss=0.3195, over 17213.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1234, cr_loss=0.3395, over 3350111.32 frames. ], batch size: 50, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:57:59,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=738738.0, ans=0.0
2024-09-25 11:58:55,647 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.283e+02 1.406e+02 1.500e+02 1.911e+02, threshold=2.813e+02, percent-clipped=0.0
2024-09-25 11:59:00,466 INFO [train.py:1198] (1/4) Epoch 41, batch 2500, loss[loss=0.2243, ctc_loss=0.1475, cr_loss=0.3842, over 17105.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1215, cr_loss=0.3366, over 3362345.60 frames. ], batch size: 49, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 11:59:08,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=738924.6666666666, ans=0.2
2024-09-25 11:59:15,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=738971.3333333334, ans=0.125
2024-09-25 11:59:24,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=738971.3333333334, ans=0.1
2024-09-25 11:59:30,097 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.45 vs. limit=10.0
2024-09-25 11:59:41,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=739018.0, ans=0.125
2024-09-25 11:59:45,054 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.79 vs. limit=15.0
2024-09-25 12:00:13,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=739111.3333333334, ans=0.0
2024-09-25 12:00:20,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=739111.3333333334, ans=0.125
2024-09-25 12:00:23,049 INFO [train.py:1198] (1/4) Epoch 41, batch 2550, loss[loss=0.151, ctc_loss=0.09609, cr_loss=0.2746, over 16272.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1219, cr_loss=0.3381, over 3365996.89 frames. ], batch size: 36, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 12:00:25,046 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:00:31,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.99 vs. limit=15.0
2024-09-25 12:01:05,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=739251.3333333334, ans=0.125
2024-09-25 12:01:38,422 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.313e+02 1.392e+02 1.468e+02 1.882e+02, threshold=2.785e+02, percent-clipped=0.0
2024-09-25 12:01:43,156 INFO [train.py:1198] (1/4) Epoch 41, batch 2600, loss[loss=0.1981, ctc_loss=0.1262, cr_loss=0.3596, over 17205.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1224, cr_loss=0.3395, over 3372095.33 frames. ], batch size: 50, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 12:02:10,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=739438.0, ans=0.125
2024-09-25 12:02:13,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=739484.6666666666, ans=0.125
2024-09-25 12:02:18,395 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=739484.6666666666, ans=0.0
2024-09-25 12:02:26,374 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=739484.6666666666, ans=0.1
2024-09-25 12:02:45,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=739578.0, ans=0.5
2024-09-25 12:02:47,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer_ff2.min_abs, batch_count=739578.0, ans=0.1
2024-09-25 12:03:07,751 INFO [train.py:1198] (1/4) Epoch 41, batch 2650, loss[loss=0.2279, ctc_loss=0.1533, cr_loss=0.3728, over 11945.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1237, cr_loss=0.3414, over 3351914.39 frames. ], batch size: 123, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 12:03:33,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=739671.3333333334, ans=0.2
2024-09-25 12:04:15,106 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=739811.3333333334, ans=0.025
2024-09-25 12:04:26,077 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.313e+02 1.405e+02 1.499e+02 1.845e+02, threshold=2.809e+02, percent-clipped=0.0
2024-09-25 12:04:30,838 INFO [train.py:1198] (1/4) Epoch 41, batch 2700, loss[loss=0.184, ctc_loss=0.1159, cr_loss=0.3404, over 17317.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1234, cr_loss=0.3413, over 3357158.95 frames. ], batch size: 51, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 12:04:35,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=739858.0, ans=0.0
2024-09-25 12:04:46,675 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:04:46,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=739858.0, ans=0.1
2024-09-25 12:04:53,873 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.29 vs. limit=12.0
2024-09-25 12:05:12,388 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=23.41 vs. limit=22.5
2024-09-25 12:05:23,221 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.50 vs. limit=22.5
2024-09-25 12:05:40,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=740044.6666666666, ans=0.2
2024-09-25 12:05:42,801 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.59 vs. limit=6.0
2024-09-25 12:05:53,460 INFO [train.py:1198] (1/4) Epoch 41, batch 2750, loss[loss=0.1849, ctc_loss=0.1206, cr_loss=0.3217, over 17165.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.123, cr_loss=0.3404, over 3361906.15 frames. ], batch size: 48, lr: 2.89e-03, grad_scale: 16.0
2024-09-25 12:05:58,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=740091.3333333334, ans=0.025
2024-09-25 12:06:05,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=740091.3333333334, ans=0.025
2024-09-25 12:06:22,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=740138.0, ans=0.125
2024-09-25 12:06:27,920 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:06:27,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=740184.6666666666, ans=0.125
2024-09-25 12:06:56,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=740278.0, ans=0.1
2024-09-25 12:07:09,100 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.274e+02 1.386e+02 1.486e+02 2.179e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 12:07:11,406 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.54 vs. limit=12.0
2024-09-25 12:07:14,038 INFO [train.py:1198] (1/4) Epoch 41, batch 2800, loss[loss=0.2162, ctc_loss=0.1436, cr_loss=0.363, over 16992.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1227, cr_loss=0.3396, over 3361602.38 frames. ], batch size: 53, lr: 2.89e-03, grad_scale: 32.0
2024-09-25 12:07:14,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=740324.6666666666, ans=0.025
2024-09-25 12:07:40,968 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.81 vs. limit=15.0
2024-09-25 12:07:42,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=740371.3333333334, ans=0.125
2024-09-25 12:07:55,475 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.90 vs. limit=10.0
2024-09-25 12:08:28,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=740511.3333333334, ans=0.0
2024-09-25 12:08:42,543 INFO [train.py:1198] (1/4) Epoch 41, batch 2850, loss[loss=0.1888, ctc_loss=0.1225, cr_loss=0.3318, over 17058.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.123, cr_loss=0.3397, over 3346282.01 frames. ], batch size: 46, lr: 2.89e-03, grad_scale: 32.0
2024-09-25 12:09:07,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=740604.6666666666, ans=0.125
2024-09-25 12:10:00,303 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.293e+02 1.358e+02 1.450e+02 1.925e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 12:10:05,285 INFO [train.py:1198] (1/4) Epoch 41, batch 2900, loss[loss=0.2338, ctc_loss=0.1543, cr_loss=0.3974, over 15242.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1233, cr_loss=0.3407, over 3348135.98 frames. ], batch size: 89, lr: 2.89e-03, grad_scale: 32.0
2024-09-25 12:11:11,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=740978.0, ans=0.1
2024-09-25 12:11:23,132 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.14 vs. limit=15.0
2024-09-25 12:11:25,601 INFO [train.py:1198] (1/4) Epoch 41, batch 2950, loss[loss=0.1737, ctc_loss=0.1083, cr_loss=0.3271, over 16931.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1235, cr_loss=0.3411, over 3338642.40 frames. ], batch size: 42, lr: 2.89e-03, grad_scale: 32.0
2024-09-25 12:11:26,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=741024.6666666666, ans=0.0
2024-09-25 12:11:30,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=741024.6666666666, ans=0.0
2024-09-25 12:12:20,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=741164.6666666666, ans=0.2
2024-09-25 12:12:25,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=741164.6666666666, ans=0.0
2024-09-25 12:12:31,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=741211.3333333334, ans=0.125
2024-09-25 12:12:33,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=741211.3333333334, ans=0.0
2024-09-25 12:12:40,613 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.313e+02 1.388e+02 1.485e+02 2.724e+02, threshold=2.776e+02, percent-clipped=1.0
2024-09-25 12:12:40,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=741211.3333333334, ans=0.0
2024-09-25 12:12:41,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=741211.3333333334, ans=0.125
2024-09-25 12:12:45,291 INFO [train.py:1198] (1/4) Epoch 41, batch 3000, loss[loss=0.1884, ctc_loss=0.1195, cr_loss=0.3448, over 17001.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1232, cr_loss=0.3416, over 3347962.34 frames. ], batch size: 53, lr: 2.89e-03, grad_scale: 32.0
2024-09-25 12:12:45,292 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 12:13:00,797 INFO [train.py:1230] (1/4) Epoch 41, validation: loss=0.03575, ctc_loss=0.03575, cr_loss=9.81e-15, over 944034.00 frames. 
2024-09-25 12:13:00,797 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 12:13:31,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=741304.6666666666, ans=0.125
2024-09-25 12:14:00,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=741398.0, ans=0.125
2024-09-25 12:14:07,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=741444.6666666666, ans=0.125
2024-09-25 12:14:11,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=741444.6666666666, ans=0.125
2024-09-25 12:14:18,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=741444.6666666666, ans=0.2
2024-09-25 12:14:26,512 INFO [train.py:1198] (1/4) Epoch 41, batch 3050, loss[loss=0.172, ctc_loss=0.1128, cr_loss=0.2963, over 17223.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.123, cr_loss=0.3411, over 3347205.64 frames. ], batch size: 50, lr: 2.88e-03, grad_scale: 32.0
2024-09-25 12:14:48,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=741538.0, ans=0.0
2024-09-25 12:15:04,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=741584.6666666666, ans=0.0
2024-09-25 12:15:29,076 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=741678.0, ans=0.125
2024-09-25 12:15:32,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=741678.0, ans=0.0
2024-09-25 12:15:36,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=741678.0, ans=0.025
2024-09-25 12:15:39,715 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.278e+02 1.352e+02 1.472e+02 2.246e+02, threshold=2.705e+02, percent-clipped=0.0
2024-09-25 12:15:44,427 INFO [train.py:1198] (1/4) Epoch 41, batch 3100, loss[loss=0.1636, ctc_loss=0.1044, cr_loss=0.2962, over 17052.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1221, cr_loss=0.3393, over 3351347.35 frames. ], batch size: 39, lr: 2.88e-03, grad_scale: 32.0
2024-09-25 12:16:11,857 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.37 vs. limit=15.0
2024-09-25 12:16:28,976 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.65 vs. limit=15.0
2024-09-25 12:16:42,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=741864.6666666666, ans=0.0
2024-09-25 12:16:42,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.max_positive, batch_count=741864.6666666666, ans=0.95
2024-09-25 12:16:58,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=741911.3333333334, ans=0.0
2024-09-25 12:17:04,754 INFO [train.py:1198] (1/4) Epoch 41, batch 3150, loss[loss=0.2037, ctc_loss=0.1308, cr_loss=0.3645, over 17006.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1223, cr_loss=0.3395, over 3355732.02 frames. ], batch size: 52, lr: 2.88e-03, grad_scale: 32.0
2024-09-25 12:17:49,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=742051.3333333334, ans=0.0
2024-09-25 12:18:10,610 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.09 vs. limit=15.0
2024-09-25 12:18:17,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=742144.6666666666, ans=0.125
2024-09-25 12:18:19,185 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.276e+02 1.346e+02 1.473e+02 3.100e+02, threshold=2.691e+02, percent-clipped=1.0
2024-09-25 12:18:23,855 INFO [train.py:1198] (1/4) Epoch 41, batch 3200, loss[loss=0.148, ctc_loss=0.09397, cr_loss=0.2702, over 17024.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1219, cr_loss=0.3383, over 3351000.70 frames. ], batch size: 44, lr: 2.88e-03, grad_scale: 32.0
2024-09-25 12:18:24,315 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:18:25,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=742191.3333333334, ans=0.2
2024-09-25 12:19:33,235 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:19:42,197 INFO [train.py:1198] (1/4) Epoch 41, batch 3250, loss[loss=0.1682, ctc_loss=0.106, cr_loss=0.3111, over 17089.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.1229, cr_loss=0.3406, over 3358346.52 frames. ], batch size: 43, lr: 2.88e-03, grad_scale: 8.0
2024-09-25 12:20:15,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=742518.0, ans=0.0
2024-09-25 12:20:26,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=742518.0, ans=10.0
2024-09-25 12:20:28,556 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.57 vs. limit=15.0
2024-09-25 12:20:35,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=742564.6666666666, ans=0.125
2024-09-25 12:20:58,869 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.297e+02 1.390e+02 1.461e+02 1.762e+02, threshold=2.780e+02, percent-clipped=0.0
2024-09-25 12:21:00,473 INFO [train.py:1198] (1/4) Epoch 41, batch 3300, loss[loss=0.2161, ctc_loss=0.1476, cr_loss=0.3423, over 11555.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.1228, cr_loss=0.3404, over 3355450.06 frames. ], batch size: 123, lr: 2.88e-03, grad_scale: 8.0
2024-09-25 12:21:21,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=742704.6666666666, ans=0.0
2024-09-25 12:21:24,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=742704.6666666666, ans=0.0
2024-09-25 12:21:29,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=742704.6666666666, ans=0.125
2024-09-25 12:21:42,363 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.28 vs. limit=15.0
2024-09-25 12:22:07,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=742844.6666666666, ans=0.1
2024-09-25 12:22:18,676 INFO [train.py:1198] (1/4) Epoch 41, batch 3350, loss[loss=0.2182, ctc_loss=0.1437, cr_loss=0.3725, over 14882.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1232, cr_loss=0.3414, over 3351009.53 frames. ], batch size: 89, lr: 2.88e-03, grad_scale: 8.0
2024-09-25 12:22:19,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=742891.3333333334, ans=0.125
2024-09-25 12:22:34,598 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=742938.0, ans=0.0
2024-09-25 12:22:48,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=742984.6666666666, ans=0.1
2024-09-25 12:23:18,783 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.40 vs. limit=12.0
2024-09-25 12:23:35,162 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.288e+02 1.390e+02 1.522e+02 3.340e+02, threshold=2.781e+02, percent-clipped=2.0
2024-09-25 12:23:36,768 INFO [train.py:1198] (1/4) Epoch 41, batch 3400, loss[loss=0.1931, ctc_loss=0.1224, cr_loss=0.3535, over 17024.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1235, cr_loss=0.3416, over 3352763.36 frames. ], batch size: 44, lr: 2.88e-03, grad_scale: 8.0
2024-09-25 12:24:39,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=743264.6666666666, ans=0.125
2024-09-25 12:25:00,981 INFO [train.py:1198] (1/4) Epoch 41, batch 3450, loss[loss=0.1753, ctc_loss=0.1129, cr_loss=0.3121, over 17103.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1239, cr_loss=0.3425, over 3354133.45 frames. ], batch size: 40, lr: 2.88e-03, grad_scale: 8.0
2024-09-25 12:25:10,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=743358.0, ans=0.0
2024-09-25 12:25:14,172 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.88 vs. limit=6.0
2024-09-25 12:26:03,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=743544.6666666666, ans=0.125
2024-09-25 12:26:06,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=743544.6666666666, ans=0.125
2024-09-25 12:26:10,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=6.00 vs. limit=15.0
2024-09-25 12:26:17,203 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.301e+02 1.363e+02 1.481e+02 3.003e+02, threshold=2.726e+02, percent-clipped=1.0
2024-09-25 12:26:18,734 INFO [train.py:1198] (1/4) Epoch 41, batch 3500, loss[loss=0.1988, ctc_loss=0.1244, cr_loss=0.3719, over 17054.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1236, cr_loss=0.342, over 3351319.08 frames. ], batch size: 46, lr: 2.88e-03, grad_scale: 8.0
2024-09-25 12:26:30,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=743591.3333333334, ans=0.0
2024-09-25 12:26:54,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten.whitening_limit, batch_count=743684.6666666666, ans=22.5
2024-09-25 12:26:55,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=743684.6666666666, ans=0.125
2024-09-25 12:27:03,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=743684.6666666666, ans=0.125
2024-09-25 12:27:06,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=743731.3333333334, ans=0.0
2024-09-25 12:27:10,210 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.35 vs. limit=6.0
2024-09-25 12:27:21,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.41 vs. limit=15.0
2024-09-25 12:27:33,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=743778.0, ans=0.125
2024-09-25 12:27:39,045 INFO [train.py:1198] (1/4) Epoch 41, batch 3550, loss[loss=0.1945, ctc_loss=0.1228, cr_loss=0.3583, over 16976.00 frames. ], tot_loss[loss=0.1922, ctc_loss=0.1237, cr_loss=0.3421, over 3351442.01 frames. ], batch size: 56, lr: 2.88e-03, grad_scale: 8.0
2024-09-25 12:27:51,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=743824.6666666666, ans=0.07
2024-09-25 12:27:56,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=743871.3333333334, ans=0.125
2024-09-25 12:27:59,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=743871.3333333334, ans=0.2
2024-09-25 12:28:05,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=743871.3333333334, ans=0.0
2024-09-25 12:28:08,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=743918.0, ans=0.125
2024-09-25 12:28:10,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=743918.0, ans=0.125
2024-09-25 12:28:27,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=743964.6666666666, ans=0.025
2024-09-25 12:28:35,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=743964.6666666666, ans=0.0
2024-09-25 12:28:37,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=743964.6666666666, ans=0.0
2024-09-25 12:28:55,697 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.288e+02 1.370e+02 1.460e+02 2.719e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-25 12:28:57,317 INFO [train.py:1198] (1/4) Epoch 41, batch 3600, loss[loss=0.1844, ctc_loss=0.1185, cr_loss=0.3299, over 16915.00 frames. ], tot_loss[loss=0.1924, ctc_loss=0.1239, cr_loss=0.3426, over 3351746.40 frames. ], batch size: 58, lr: 2.88e-03, grad_scale: 16.0
2024-09-25 12:29:30,596 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.33 vs. limit=15.0
2024-09-25 12:29:36,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=744151.3333333334, ans=0.09899494936611666
2024-09-25 12:29:48,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.20 vs. limit=15.0
2024-09-25 12:29:49,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=744198.0, ans=0.0
2024-09-25 12:30:15,424 INFO [train.py:1198] (1/4) Epoch 41, batch 3650, loss[loss=0.1739, ctc_loss=0.1123, cr_loss=0.308, over 16932.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1231, cr_loss=0.3409, over 3350352.43 frames. ], batch size: 42, lr: 2.88e-03, grad_scale: 16.0
2024-09-25 12:30:15,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=744291.3333333334, ans=0.1
2024-09-25 12:30:17,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=744291.3333333334, ans=0.1
2024-09-25 12:30:32,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=744338.0, ans=0.0
2024-09-25 12:30:34,507 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.41 vs. limit=15.0
2024-09-25 12:30:43,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=744338.0, ans=0.125
2024-09-25 12:30:53,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=744384.6666666666, ans=0.125
2024-09-25 12:31:03,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=744431.3333333334, ans=0.125
2024-09-25 12:31:32,488 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.315e+02 1.407e+02 1.494e+02 1.743e+02, threshold=2.814e+02, percent-clipped=0.0
2024-09-25 12:31:34,120 INFO [train.py:1198] (1/4) Epoch 41, batch 3700, loss[loss=0.2279, ctc_loss=0.1539, cr_loss=0.3699, over 11932.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1234, cr_loss=0.3412, over 3340584.28 frames. ], batch size: 123, lr: 2.88e-03, grad_scale: 16.0
2024-09-25 12:31:56,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=744571.3333333334, ans=0.125
2024-09-25 12:32:02,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=744571.3333333334, ans=0.125
2024-09-25 12:32:03,276 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.52 vs. limit=15.0
2024-09-25 12:32:07,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=744618.0, ans=0.125
2024-09-25 12:32:12,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=744618.0, ans=0.0
2024-09-25 12:32:18,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=744618.0, ans=0.025
2024-09-25 12:32:35,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=744711.3333333334, ans=0.125
2024-09-25 12:32:53,077 INFO [train.py:1198] (1/4) Epoch 41, batch 3750, loss[loss=0.2309, ctc_loss=0.1508, cr_loss=0.4001, over 14838.00 frames. ], tot_loss[loss=0.1913, ctc_loss=0.1232, cr_loss=0.3408, over 3334139.95 frames. ], batch size: 89, lr: 2.88e-03, grad_scale: 16.0
2024-09-25 12:32:54,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=744758.0, ans=0.0
2024-09-25 12:33:18,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=744804.6666666666, ans=0.025
2024-09-25 12:33:46,763 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.60 vs. limit=22.5
2024-09-25 12:33:49,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=744898.0, ans=0.2
2024-09-25 12:34:11,724 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.288e+02 1.369e+02 1.453e+02 1.957e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-25 12:34:14,054 INFO [train.py:1198] (1/4) Epoch 41, batch 3800, loss[loss=0.1994, ctc_loss=0.1299, cr_loss=0.3476, over 16916.00 frames. ], tot_loss[loss=0.1908, ctc_loss=0.1229, cr_loss=0.3395, over 3302594.57 frames. ], batch size: 58, lr: 2.88e-03, grad_scale: 16.0
2024-09-25 12:34:19,649 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=7.59 vs. limit=15.0
2024-09-25 12:34:23,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=744991.3333333334, ans=0.025
2024-09-25 12:34:28,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=745038.0, ans=0.0
2024-09-25 12:34:40,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=745038.0, ans=0.2
2024-09-25 12:34:45,412 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=745084.6666666666, ans=0.125
2024-09-25 12:35:04,380 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:35:07,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=745131.3333333334, ans=0.125
2024-09-25 12:35:07,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.63 vs. limit=15.0
2024-09-25 12:35:18,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=745178.0, ans=0.125
2024-09-25 12:35:24,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=745178.0, ans=0.125
2024-09-25 12:35:29,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=745178.0, ans=0.0
2024-09-25 12:35:29,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=745178.0, ans=0.0
2024-09-25 12:35:32,617 INFO [train.py:1198] (1/4) Epoch 41, batch 3850, loss[loss=0.1664, ctc_loss=0.1056, cr_loss=0.3042, over 17259.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1221, cr_loss=0.3374, over 3289146.09 frames. ], batch size: 44, lr: 2.88e-03, grad_scale: 16.0
2024-09-25 12:35:47,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=745271.3333333334, ans=0.125
2024-09-25 12:35:50,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=745271.3333333334, ans=0.025
2024-09-25 12:35:52,017 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:35:54,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=745271.3333333334, ans=0.025
2024-09-25 12:36:12,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=745318.0, ans=0.05
2024-09-25 12:36:18,207 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=10.61 vs. limit=22.5
2024-09-25 12:36:24,953 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:37:33,472 INFO [train.py:1198] (1/4) Epoch 42, batch 0, loss[loss=0.1866, ctc_loss=0.1191, cr_loss=0.3376, over 17012.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1191, cr_loss=0.3376, over 17012.00 frames. ], batch size: 51, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:37:33,473 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 12:37:48,886 INFO [train.py:1230] (1/4) Epoch 42, validation: loss=0.03453, ctc_loss=0.03453, cr_loss=1.019e-14, over 944034.00 frames. 
2024-09-25 12:37:48,887 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 12:37:51,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten.whitening_limit, batch_count=745439.3333333334, ans=15.0
2024-09-25 12:37:53,632 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.354e+02 1.491e+02 1.700e+02 3.066e+02, threshold=2.981e+02, percent-clipped=1.0
2024-09-25 12:37:57,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=745439.3333333334, ans=0.0
2024-09-25 12:38:00,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten.whitening_limit, batch_count=745439.3333333334, ans=15.0
2024-09-25 12:38:06,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer_ff2.min_abs, batch_count=745486.0, ans=0.1
2024-09-25 12:38:08,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=745486.0, ans=0.125
2024-09-25 12:38:16,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=745486.0, ans=0.125
2024-09-25 12:39:03,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=745626.0, ans=0.0
2024-09-25 12:39:11,032 INFO [train.py:1198] (1/4) Epoch 42, batch 50, loss[loss=0.1874, ctc_loss=0.1208, cr_loss=0.333, over 17263.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1216, cr_loss=0.3368, over 759361.80 frames. ], batch size: 44, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:39:11,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=745672.6666666666, ans=0.1
2024-09-25 12:39:11,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_abs, batch_count=745672.6666666666, ans=0.5
2024-09-25 12:39:22,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=745672.6666666666, ans=0.125
2024-09-25 12:39:48,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=745766.0, ans=0.125
2024-09-25 12:40:02,877 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.82 vs. limit=6.0
2024-09-25 12:40:15,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=745812.6666666666, ans=0.0
2024-09-25 12:40:37,093 INFO [train.py:1198] (1/4) Epoch 42, batch 100, loss[loss=0.19, ctc_loss=0.123, cr_loss=0.3349, over 17023.00 frames. ], tot_loss[loss=0.192, ctc_loss=0.1238, cr_loss=0.3409, over 1338085.23 frames. ], batch size: 56, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:40:41,769 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.296e+02 1.378e+02 1.504e+02 1.895e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 12:40:46,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=745906.0, ans=0.125
2024-09-25 12:41:01,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=745952.6666666666, ans=0.05
2024-09-25 12:41:03,377 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.65 vs. limit=15.0
2024-09-25 12:41:16,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=745999.3333333334, ans=0.2
2024-09-25 12:41:18,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=745999.3333333334, ans=0.125
2024-09-25 12:41:31,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.max_abs, batch_count=746046.0, ans=10.0
2024-09-25 12:41:37,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=746046.0, ans=0.04949747468305833
2024-09-25 12:41:39,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=746046.0, ans=0.125
2024-09-25 12:41:40,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=746046.0, ans=0.2
2024-09-25 12:41:50,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=746092.6666666666, ans=0.015
2024-09-25 12:41:59,754 INFO [train.py:1198] (1/4) Epoch 42, batch 150, loss[loss=0.1689, ctc_loss=0.1071, cr_loss=0.3088, over 17246.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1211, cr_loss=0.3371, over 1795148.62 frames. ], batch size: 42, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:42:37,366 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.12 vs. limit=6.0
2024-09-25 12:42:39,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=746232.6666666666, ans=0.125
2024-09-25 12:42:59,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=746279.3333333334, ans=0.0
2024-09-25 12:43:15,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=746326.0, ans=0.1
2024-09-25 12:43:19,865 INFO [train.py:1198] (1/4) Epoch 42, batch 200, loss[loss=0.1741, ctc_loss=0.1087, cr_loss=0.3269, over 17041.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1222, cr_loss=0.3392, over 2141282.87 frames. ], batch size: 44, lr: 2.84e-03, grad_scale: 16.0
2024-09-25 12:43:22,169 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=10.10 vs. limit=12.0
2024-09-25 12:43:26,427 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.308e+02 1.400e+02 1.491e+02 1.853e+02, threshold=2.800e+02, percent-clipped=0.0
2024-09-25 12:43:30,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=746372.6666666666, ans=0.1
2024-09-25 12:43:52,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=746419.3333333334, ans=0.2
2024-09-25 12:43:57,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=746466.0, ans=0.0
2024-09-25 12:44:43,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=746559.3333333334, ans=0.025
2024-09-25 12:44:45,844 INFO [train.py:1198] (1/4) Epoch 42, batch 250, loss[loss=0.1757, ctc_loss=0.1115, cr_loss=0.3211, over 16942.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.122, cr_loss=0.3396, over 2416333.29 frames. ], batch size: 42, lr: 2.84e-03, grad_scale: 16.0
2024-09-25 12:44:58,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=746606.0, ans=0.125
2024-09-25 12:46:13,664 INFO [train.py:1198] (1/4) Epoch 42, batch 300, loss[loss=0.2068, ctc_loss=0.1334, cr_loss=0.3674, over 17231.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1217, cr_loss=0.3391, over 2627183.62 frames. ], batch size: 55, lr: 2.84e-03, grad_scale: 16.0
2024-09-25 12:46:15,716 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=746839.3333333334, ans=0.125
2024-09-25 12:46:20,041 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.307e+02 1.390e+02 1.481e+02 2.059e+02, threshold=2.780e+02, percent-clipped=0.0
2024-09-25 12:46:24,317 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.71 vs. limit=12.0
2024-09-25 12:46:32,244 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.74 vs. limit=12.0
2024-09-25 12:46:39,057 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.66 vs. limit=6.0
2024-09-25 12:46:48,276 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.10 vs. limit=15.0
2024-09-25 12:46:57,570 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.61 vs. limit=15.0
2024-09-25 12:47:14,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=746979.3333333334, ans=0.025
2024-09-25 12:47:33,913 INFO [train.py:1198] (1/4) Epoch 42, batch 350, loss[loss=0.1816, ctc_loss=0.1134, cr_loss=0.3411, over 17060.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1213, cr_loss=0.338, over 2785883.48 frames. ], batch size: 46, lr: 2.84e-03, grad_scale: 16.0
2024-09-25 12:47:58,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=747119.3333333334, ans=0.125
2024-09-25 12:48:15,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=747166.0, ans=0.025
2024-09-25 12:48:20,676 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=747212.6666666666, ans=0.0
2024-09-25 12:48:22,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=747212.6666666666, ans=0.125
2024-09-25 12:48:24,319 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.43 vs. limit=22.5
2024-09-25 12:48:28,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=747212.6666666666, ans=0.1
2024-09-25 12:48:41,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=747259.3333333334, ans=0.0
2024-09-25 12:48:52,285 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.46 vs. limit=15.0
2024-09-25 12:48:56,551 INFO [train.py:1198] (1/4) Epoch 42, batch 400, loss[loss=0.1962, ctc_loss=0.1247, cr_loss=0.3573, over 15902.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1225, cr_loss=0.3409, over 2913610.02 frames. ], batch size: 74, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:49:02,843 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.303e+02 1.377e+02 1.468e+02 2.064e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-25 12:49:14,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=747352.6666666666, ans=0.1
2024-09-25 12:49:17,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=747352.6666666666, ans=0.125
2024-09-25 12:49:21,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=747352.6666666666, ans=0.09899494936611666
2024-09-25 12:49:37,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=747399.3333333334, ans=0.1
2024-09-25 12:49:42,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=747399.3333333334, ans=0.1
2024-09-25 12:49:59,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=747446.0, ans=0.09899494936611666
2024-09-25 12:50:01,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=747492.6666666666, ans=0.125
2024-09-25 12:50:03,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=747492.6666666666, ans=0.1
2024-09-25 12:50:04,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=747492.6666666666, ans=0.0
2024-09-25 12:50:21,630 INFO [train.py:1198] (1/4) Epoch 42, batch 450, loss[loss=0.2056, ctc_loss=0.1326, cr_loss=0.365, over 17231.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1228, cr_loss=0.3413, over 3006239.93 frames. ], batch size: 55, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:50:26,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=747539.3333333334, ans=0.0
2024-09-25 12:51:44,414 INFO [train.py:1198] (1/4) Epoch 42, batch 500, loss[loss=0.2046, ctc_loss=0.1312, cr_loss=0.367, over 17018.00 frames. ], tot_loss[loss=0.1908, ctc_loss=0.1227, cr_loss=0.3407, over 3086064.26 frames. ], batch size: 56, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:51:50,062 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.96 vs. limit=15.0
2024-09-25 12:51:50,867 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.301e+02 1.386e+02 1.488e+02 2.359e+02, threshold=2.773e+02, percent-clipped=0.0
2024-09-25 12:52:15,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=747866.0, ans=0.125
2024-09-25 12:52:15,820 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.74 vs. limit=15.0
2024-09-25 12:52:16,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=747866.0, ans=0.2
2024-09-25 12:52:34,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=747912.6666666666, ans=0.025
2024-09-25 12:53:04,112 INFO [train.py:1198] (1/4) Epoch 42, batch 550, loss[loss=0.1761, ctc_loss=0.1107, cr_loss=0.3273, over 17157.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1218, cr_loss=0.3389, over 3145756.57 frames. ], batch size: 40, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:53:20,767 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.99 vs. limit=15.0
2024-09-25 12:53:28,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=748052.6666666666, ans=0.1
2024-09-25 12:53:51,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=748099.3333333334, ans=0.1
2024-09-25 12:53:53,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=748146.0, ans=0.1
2024-09-25 12:53:59,891 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=748146.0, ans=0.0
2024-09-25 12:54:26,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=748192.6666666666, ans=0.0
2024-09-25 12:54:29,012 INFO [train.py:1198] (1/4) Epoch 42, batch 600, loss[loss=0.1848, ctc_loss=0.1165, cr_loss=0.3415, over 17106.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1216, cr_loss=0.3383, over 3186407.27 frames. ], batch size: 40, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:54:35,339 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.284e+02 1.358e+02 1.453e+02 2.155e+02, threshold=2.716e+02, percent-clipped=0.0
2024-09-25 12:54:51,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=748286.0, ans=0.1
2024-09-25 12:55:01,410 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 12:55:07,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=748332.6666666666, ans=0.125
2024-09-25 12:55:10,850 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=748332.6666666666, ans=0.5
2024-09-25 12:55:24,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=748379.3333333334, ans=0.1
2024-09-25 12:55:36,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=748426.0, ans=0.025
2024-09-25 12:55:51,814 INFO [train.py:1198] (1/4) Epoch 42, batch 650, loss[loss=0.2061, ctc_loss=0.1333, cr_loss=0.3639, over 16915.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.122, cr_loss=0.3391, over 3228227.87 frames. ], batch size: 58, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:56:12,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=748519.3333333334, ans=0.0
2024-09-25 12:56:30,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=748566.0, ans=0.125
2024-09-25 12:56:53,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.25 vs. limit=15.0
2024-09-25 12:57:14,805 INFO [train.py:1198] (1/4) Epoch 42, batch 700, loss[loss=0.2389, ctc_loss=0.1604, cr_loss=0.3922, over 11462.00 frames. ], tot_loss[loss=0.1898, ctc_loss=0.122, cr_loss=0.3389, over 3253734.43 frames. ], batch size: 123, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:57:21,242 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.310e+02 1.390e+02 1.481e+02 1.937e+02, threshold=2.780e+02, percent-clipped=0.0
2024-09-25 12:57:39,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=748752.6666666666, ans=0.0
2024-09-25 12:57:42,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=748752.6666666666, ans=0.1
2024-09-25 12:58:10,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=748846.0, ans=0.125
2024-09-25 12:58:17,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=748892.6666666666, ans=0.125
2024-09-25 12:58:27,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=748892.6666666666, ans=0.0
2024-09-25 12:58:33,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=748939.3333333334, ans=0.125
2024-09-25 12:58:34,915 INFO [train.py:1198] (1/4) Epoch 42, batch 750, loss[loss=0.1742, ctc_loss=0.109, cr_loss=0.3259, over 17244.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1215, cr_loss=0.3375, over 3279766.51 frames. ], batch size: 44, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 12:59:07,370 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.58 vs. limit=15.0
2024-09-25 12:59:30,175 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=749079.3333333334, ans=0.1
2024-09-25 12:59:57,713 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 13:00:00,473 INFO [train.py:1198] (1/4) Epoch 42, batch 800, loss[loss=0.1757, ctc_loss=0.1122, cr_loss=0.3178, over 17116.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1209, cr_loss=0.337, over 3300163.04 frames. ], batch size: 40, lr: 2.84e-03, grad_scale: 32.0
2024-09-25 13:00:06,821 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.303e+02 1.405e+02 1.532e+02 1.999e+02, threshold=2.810e+02, percent-clipped=0.0
2024-09-25 13:00:47,376 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.54 vs. limit=22.5
2024-09-25 13:00:48,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=749266.0, ans=0.0
2024-09-25 13:00:48,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=749266.0, ans=0.1
2024-09-25 13:00:53,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=749312.6666666666, ans=0.125
2024-09-25 13:01:22,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=749359.3333333334, ans=0.125
2024-09-25 13:01:25,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=749406.0, ans=0.125
2024-09-25 13:01:26,597 INFO [train.py:1198] (1/4) Epoch 42, batch 850, loss[loss=0.2024, ctc_loss=0.1294, cr_loss=0.365, over 17150.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1209, cr_loss=0.3375, over 3315513.36 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 32.0
2024-09-25 13:01:28,689 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=749406.0, ans=0.125
2024-09-25 13:01:55,193 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.58 vs. limit=15.0
2024-09-25 13:01:56,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=749452.6666666666, ans=0.0
2024-09-25 13:02:07,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=749499.3333333334, ans=0.125
2024-09-25 13:02:18,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=749546.0, ans=0.0
2024-09-25 13:02:39,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=749592.6666666666, ans=0.0
2024-09-25 13:02:46,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=749639.3333333334, ans=0.125
2024-09-25 13:02:47,617 INFO [train.py:1198] (1/4) Epoch 42, batch 900, loss[loss=0.2076, ctc_loss=0.1355, cr_loss=0.3608, over 17309.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1221, cr_loss=0.3397, over 3310856.77 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 32.0
2024-09-25 13:02:51,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=749639.3333333334, ans=0.0
2024-09-25 13:02:53,984 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.278e+02 1.357e+02 1.447e+02 3.889e+02, threshold=2.715e+02, percent-clipped=1.0
2024-09-25 13:03:02,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=749686.0, ans=0.125
2024-09-25 13:03:23,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=749732.6666666666, ans=0.125
2024-09-25 13:03:27,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=749732.6666666666, ans=0.125
2024-09-25 13:03:32,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=749732.6666666666, ans=0.125
2024-09-25 13:03:55,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=749826.0, ans=0.125
2024-09-25 13:03:59,357 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.32 vs. limit=15.0
2024-09-25 13:04:11,573 INFO [train.py:1198] (1/4) Epoch 42, batch 950, loss[loss=0.2201, ctc_loss=0.1436, cr_loss=0.3825, over 17006.00 frames. ], tot_loss[loss=0.1904, ctc_loss=0.1223, cr_loss=0.3408, over 3321560.65 frames. ], batch size: 53, lr: 2.83e-03, grad_scale: 32.0
2024-09-25 13:04:27,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=749872.6666666666, ans=0.05
2024-09-25 13:05:06,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=750012.6666666666, ans=0.125
2024-09-25 13:05:28,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=750059.3333333334, ans=0.0
2024-09-25 13:05:36,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=750106.0, ans=0.5
2024-09-25 13:05:37,406 INFO [train.py:1198] (1/4) Epoch 42, batch 1000, loss[loss=0.1837, ctc_loss=0.1183, cr_loss=0.3268, over 17247.00 frames. ], tot_loss[loss=0.1904, ctc_loss=0.1223, cr_loss=0.3405, over 3327491.13 frames. ], batch size: 44, lr: 2.83e-03, grad_scale: 32.0
2024-09-25 13:05:43,608 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.283e+02 1.358e+02 1.462e+02 1.840e+02, threshold=2.715e+02, percent-clipped=0.0
2024-09-25 13:06:10,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=750199.3333333334, ans=0.125
2024-09-25 13:06:14,142 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.01 vs. limit=6.0
2024-09-25 13:06:42,753 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.29 vs. limit=15.0
2024-09-25 13:06:59,677 INFO [train.py:1198] (1/4) Epoch 42, batch 1050, loss[loss=0.1708, ctc_loss=0.1102, cr_loss=0.3031, over 17055.00 frames. ], tot_loss[loss=0.1904, ctc_loss=0.1224, cr_loss=0.34, over 3331816.42 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 32.0
2024-09-25 13:07:04,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=750339.3333333334, ans=0.125
2024-09-25 13:07:21,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=750386.0, ans=0.125
2024-09-25 13:07:29,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=750386.0, ans=0.125
2024-09-25 13:07:29,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=750386.0, ans=0.1
2024-09-25 13:07:40,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=750432.6666666666, ans=0.125
2024-09-25 13:07:46,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=750479.3333333334, ans=0.1
2024-09-25 13:07:49,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=750479.3333333334, ans=0.1
2024-09-25 13:08:20,229 INFO [train.py:1198] (1/4) Epoch 42, batch 1100, loss[loss=0.1707, ctc_loss=0.1087, cr_loss=0.3102, over 17236.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1227, cr_loss=0.3398, over 3321795.11 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:08:28,293 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.338e+02 1.422e+02 1.527e+02 1.799e+02, threshold=2.844e+02, percent-clipped=0.0
2024-09-25 13:08:33,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=750572.6666666666, ans=0.125
2024-09-25 13:08:56,606 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 13:09:28,973 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=750759.3333333334, ans=0.2
2024-09-25 13:09:33,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=750759.3333333334, ans=0.0
2024-09-25 13:09:35,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=750759.3333333334, ans=0.0
2024-09-25 13:09:44,703 INFO [train.py:1198] (1/4) Epoch 42, batch 1150, loss[loss=0.2248, ctc_loss=0.1496, cr_loss=0.3763, over 17222.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1228, cr_loss=0.3396, over 3333381.88 frames. ], batch size: 55, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:09:44,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=750806.0, ans=0.125
2024-09-25 13:10:10,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=750852.6666666666, ans=0.1
2024-09-25 13:10:10,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=750852.6666666666, ans=0.0
2024-09-25 13:10:14,846 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=750852.6666666666, ans=0.2
2024-09-25 13:10:14,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=750852.6666666666, ans=0.0
2024-09-25 13:10:35,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=750946.0, ans=0.2
2024-09-25 13:10:38,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=750946.0, ans=0.125
2024-09-25 13:11:09,660 INFO [train.py:1198] (1/4) Epoch 42, batch 1200, loss[loss=0.1859, ctc_loss=0.1179, cr_loss=0.3402, over 16990.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1232, cr_loss=0.3404, over 3330030.90 frames. ], batch size: 56, lr: 2.83e-03, grad_scale: 32.0
2024-09-25 13:11:17,515 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.296e+02 1.388e+02 1.485e+02 1.813e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-25 13:11:57,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=751179.3333333334, ans=0.125
2024-09-25 13:12:26,733 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.99 vs. limit=15.0
2024-09-25 13:12:29,347 INFO [train.py:1198] (1/4) Epoch 42, batch 1250, loss[loss=0.1763, ctc_loss=0.1108, cr_loss=0.3278, over 16948.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1225, cr_loss=0.3391, over 3334984.54 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 32.0
2024-09-25 13:12:37,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=751272.6666666666, ans=0.0
2024-09-25 13:13:01,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.05 vs. limit=15.0
2024-09-25 13:13:03,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=751366.0, ans=0.125
2024-09-25 13:13:23,807 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=751412.6666666666, ans=0.125
2024-09-25 13:13:23,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=751412.6666666666, ans=0.025
2024-09-25 13:13:51,190 INFO [train.py:1198] (1/4) Epoch 42, batch 1300, loss[loss=0.201, ctc_loss=0.1342, cr_loss=0.3338, over 17020.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.1221, cr_loss=0.3387, over 3345380.04 frames. ], batch size: 53, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:13:57,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=751506.0, ans=0.1
2024-09-25 13:13:59,713 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=751506.0, ans=0.125
2024-09-25 13:14:00,451 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.49 vs. limit=15.0
2024-09-25 13:14:00,856 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.307e+02 1.370e+02 1.468e+02 2.127e+02, threshold=2.740e+02, percent-clipped=0.0
2024-09-25 13:14:15,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=751552.6666666666, ans=0.05
2024-09-25 13:14:34,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=751599.3333333334, ans=0.2
2024-09-25 13:14:42,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=751646.0, ans=0.0
2024-09-25 13:14:45,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=751646.0, ans=0.125
2024-09-25 13:15:10,162 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.60 vs. limit=6.0
2024-09-25 13:15:15,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=751739.3333333334, ans=0.0
2024-09-25 13:15:16,744 INFO [train.py:1198] (1/4) Epoch 42, batch 1350, loss[loss=0.1662, ctc_loss=0.105, cr_loss=0.3059, over 17127.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1216, cr_loss=0.338, over 3349751.60 frames. ], batch size: 40, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:15:27,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=751739.3333333334, ans=0.035
2024-09-25 13:16:15,880 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.54 vs. limit=15.0
2024-09-25 13:16:29,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=751926.0, ans=0.0
2024-09-25 13:16:38,719 INFO [train.py:1198] (1/4) Epoch 42, batch 1400, loss[loss=0.1759, ctc_loss=0.1116, cr_loss=0.3217, over 17110.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1219, cr_loss=0.3389, over 3343447.27 frames. ], batch size: 40, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:16:48,177 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.272e+02 1.355e+02 1.432e+02 2.085e+02, threshold=2.710e+02, percent-clipped=0.0
2024-09-25 13:17:02,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=752019.3333333334, ans=0.0
2024-09-25 13:17:26,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=752112.6666666666, ans=0.125
2024-09-25 13:17:58,206 INFO [train.py:1198] (1/4) Epoch 42, batch 1450, loss[loss=0.1567, ctc_loss=0.09682, cr_loss=0.2992, over 16683.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1219, cr_loss=0.3394, over 3347276.49 frames. ], batch size: 37, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:17:58,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=752206.0, ans=0.125
2024-09-25 13:18:21,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=752252.6666666666, ans=0.025
2024-09-25 13:18:25,345 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.29 vs. limit=12.0
2024-09-25 13:18:45,253 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.10 vs. limit=15.0
2024-09-25 13:18:52,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=752346.0, ans=0.0
2024-09-25 13:19:11,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=752392.6666666666, ans=0.0
2024-09-25 13:19:23,690 INFO [train.py:1198] (1/4) Epoch 42, batch 1500, loss[loss=0.2064, ctc_loss=0.1376, cr_loss=0.3436, over 17030.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3384, over 3342670.38 frames. ], batch size: 56, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:19:31,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=752439.3333333334, ans=0.0
2024-09-25 13:19:33,185 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.208e+02 1.278e+02 1.340e+02 1.432e+02 2.576e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-25 13:19:39,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=752486.0, ans=0.125
2024-09-25 13:19:57,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=752532.6666666666, ans=0.0
2024-09-25 13:19:57,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_abs, batch_count=752532.6666666666, ans=0.5
2024-09-25 13:20:02,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=752532.6666666666, ans=0.125
2024-09-25 13:20:05,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=752532.6666666666, ans=0.125
2024-09-25 13:20:27,910 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.35 vs. limit=6.0
2024-09-25 13:20:33,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=752626.0, ans=0.025
2024-09-25 13:20:36,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=752626.0, ans=0.125
2024-09-25 13:20:48,641 INFO [train.py:1198] (1/4) Epoch 42, batch 1550, loss[loss=0.1671, ctc_loss=0.1025, cr_loss=0.323, over 17257.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.121, cr_loss=0.3381, over 3344196.05 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:21:01,830 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 13:21:14,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=752719.3333333334, ans=0.125
2024-09-25 13:21:21,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=752766.0, ans=0.125
2024-09-25 13:21:22,593 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=752766.0, ans=0.1
2024-09-25 13:21:50,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=752812.6666666666, ans=0.0
2024-09-25 13:22:06,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=752859.3333333334, ans=0.1
2024-09-25 13:22:09,036 INFO [train.py:1198] (1/4) Epoch 42, batch 1600, loss[loss=0.1998, ctc_loss=0.1297, cr_loss=0.3503, over 16460.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1226, cr_loss=0.3402, over 3332705.58 frames. ], batch size: 66, lr: 2.83e-03, grad_scale: 16.0
2024-09-25 13:22:20,350 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.308e+02 1.394e+02 1.520e+02 2.214e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-25 13:22:22,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=752906.0, ans=0.125
2024-09-25 13:22:31,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=752952.6666666666, ans=0.125
2024-09-25 13:22:36,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=752952.6666666666, ans=0.125
2024-09-25 13:22:36,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=752952.6666666666, ans=0.125
2024-09-25 13:22:59,335 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=753046.0, ans=0.0
2024-09-25 13:23:07,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=753046.0, ans=0.0
2024-09-25 13:23:07,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=753046.0, ans=0.125
2024-09-25 13:23:31,967 INFO [train.py:1198] (1/4) Epoch 42, batch 1650, loss[loss=0.19, ctc_loss=0.1231, cr_loss=0.3347, over 17027.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.1228, cr_loss=0.3404, over 3330095.84 frames. ], batch size: 56, lr: 2.83e-03, grad_scale: 8.0
2024-09-25 13:23:49,183 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.66 vs. limit=6.0
2024-09-25 13:24:07,649 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.37 vs. limit=15.0
2024-09-25 13:24:19,557 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.98 vs. limit=10.0
2024-09-25 13:24:38,248 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.48 vs. limit=15.0
2024-09-25 13:24:48,068 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.64 vs. limit=22.5
2024-09-25 13:24:55,037 INFO [train.py:1198] (1/4) Epoch 42, batch 1700, loss[loss=0.1865, ctc_loss=0.1153, cr_loss=0.356, over 16998.00 frames. ], tot_loss[loss=0.1911, ctc_loss=0.1229, cr_loss=0.3409, over 3333649.85 frames. ], batch size: 44, lr: 2.83e-03, grad_scale: 8.0
2024-09-25 13:25:00,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=753372.6666666666, ans=0.0
2024-09-25 13:25:10,263 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.274e+02 1.365e+02 1.477e+02 2.615e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-25 13:25:40,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=753466.0, ans=0.1
2024-09-25 13:25:43,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=753466.0, ans=0.125
2024-09-25 13:25:59,403 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 13:26:19,505 INFO [train.py:1198] (1/4) Epoch 42, batch 1750, loss[loss=0.1506, ctc_loss=0.09596, cr_loss=0.273, over 17101.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.1229, cr_loss=0.3414, over 3349326.50 frames. ], batch size: 43, lr: 2.83e-03, grad_scale: 8.0
2024-09-25 13:26:37,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=753652.6666666666, ans=0.0
2024-09-25 13:26:37,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=753652.6666666666, ans=0.04949747468305833
2024-09-25 13:27:40,002 INFO [train.py:1198] (1/4) Epoch 42, batch 1800, loss[loss=0.1852, ctc_loss=0.1178, cr_loss=0.3368, over 17218.00 frames. ], tot_loss[loss=0.1923, ctc_loss=0.1238, cr_loss=0.3422, over 3340972.82 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 8.0
2024-09-25 13:27:52,868 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.293e+02 1.332e+02 1.454e+02 1.867e+02, threshold=2.665e+02, percent-clipped=0.0
2024-09-25 13:28:14,422 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.91 vs. limit=15.0
2024-09-25 13:29:04,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=754072.6666666666, ans=0.125
2024-09-25 13:29:05,784 INFO [train.py:1198] (1/4) Epoch 42, batch 1850, loss[loss=0.1467, ctc_loss=0.0934, cr_loss=0.2666, over 17004.00 frames. ], tot_loss[loss=0.1921, ctc_loss=0.1238, cr_loss=0.3419, over 3338222.93 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 8.0
2024-09-25 13:29:15,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=754072.6666666666, ans=0.125
2024-09-25 13:29:23,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=754119.3333333334, ans=0.0
2024-09-25 13:29:30,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=754119.3333333334, ans=0.04949747468305833
2024-09-25 13:29:43,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=754166.0, ans=10.0
2024-09-25 13:29:50,350 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.71 vs. limit=6.0
2024-09-25 13:30:14,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=754259.3333333334, ans=0.125
2024-09-25 13:30:25,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.max_positive, batch_count=754259.3333333334, ans=0.95
2024-09-25 13:30:31,298 INFO [train.py:1198] (1/4) Epoch 42, batch 1900, loss[loss=0.2079, ctc_loss=0.1346, cr_loss=0.3662, over 16779.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.1228, cr_loss=0.3402, over 3338243.11 frames. ], batch size: 61, lr: 2.83e-03, grad_scale: 8.0
2024-09-25 13:30:44,084 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.292e+02 1.376e+02 1.481e+02 2.312e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-25 13:30:56,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=754352.6666666666, ans=0.125
2024-09-25 13:31:01,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=754399.3333333334, ans=0.0
2024-09-25 13:31:22,787 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.61 vs. limit=15.0
2024-09-25 13:31:39,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=754492.6666666666, ans=0.125
2024-09-25 13:31:50,786 INFO [train.py:1198] (1/4) Epoch 42, batch 1950, loss[loss=0.1948, ctc_loss=0.1244, cr_loss=0.352, over 16742.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1234, cr_loss=0.341, over 3324928.88 frames. ], batch size: 61, lr: 2.83e-03, grad_scale: 8.0
2024-09-25 13:32:02,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.24 vs. limit=15.0
2024-09-25 13:32:03,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=754539.3333333334, ans=0.07
2024-09-25 13:32:26,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=754632.6666666666, ans=0.0
2024-09-25 13:32:33,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=754632.6666666666, ans=15.0
2024-09-25 13:32:34,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=754632.6666666666, ans=0.125
2024-09-25 13:32:39,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=754679.3333333334, ans=0.04949747468305833
2024-09-25 13:32:50,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=754679.3333333334, ans=0.025
2024-09-25 13:32:54,237 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.30 vs. limit=15.0
2024-09-25 13:33:01,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=754726.0, ans=0.1
2024-09-25 13:33:10,943 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.11 vs. limit=15.0
2024-09-25 13:33:12,670 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=7.88 vs. limit=15.0
2024-09-25 13:33:13,475 INFO [train.py:1198] (1/4) Epoch 42, batch 2000, loss[loss=0.1812, ctc_loss=0.1152, cr_loss=0.3301, over 17296.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1232, cr_loss=0.3408, over 3331779.58 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:33:23,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=754772.6666666666, ans=0.025
2024-09-25 13:33:25,999 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.295e+02 1.343e+02 1.422e+02 2.059e+02, threshold=2.687e+02, percent-clipped=0.0
2024-09-25 13:33:26,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=754772.6666666666, ans=0.0
2024-09-25 13:33:44,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=754866.0, ans=0.2
2024-09-25 13:34:06,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.79 vs. limit=22.5
2024-09-25 13:34:21,160 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.76 vs. limit=10.0
2024-09-25 13:34:33,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=754959.3333333334, ans=0.0
2024-09-25 13:34:36,178 INFO [train.py:1198] (1/4) Epoch 42, batch 2050, loss[loss=0.188, ctc_loss=0.1224, cr_loss=0.3279, over 16830.00 frames. ], tot_loss[loss=0.1919, ctc_loss=0.1237, cr_loss=0.3414, over 3321908.87 frames. ], batch size: 61, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:35:00,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=755052.6666666666, ans=0.025
2024-09-25 13:35:04,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=755052.6666666666, ans=0.1
2024-09-25 13:35:16,493 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=4.98 vs. limit=15.0
2024-09-25 13:35:22,261 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.83 vs. limit=15.0
2024-09-25 13:36:01,516 INFO [train.py:1198] (1/4) Epoch 42, batch 2100, loss[loss=0.197, ctc_loss=0.129, cr_loss=0.34, over 17299.00 frames. ], tot_loss[loss=0.1912, ctc_loss=0.123, cr_loss=0.3406, over 3333961.66 frames. ], batch size: 51, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:36:14,532 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.293e+02 1.359e+02 1.448e+02 2.316e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-25 13:36:20,326 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.19 vs. limit=15.0
2024-09-25 13:37:21,952 INFO [train.py:1198] (1/4) Epoch 42, batch 2150, loss[loss=0.1733, ctc_loss=0.1093, cr_loss=0.3202, over 17306.00 frames. ], tot_loss[loss=0.1918, ctc_loss=0.1235, cr_loss=0.3415, over 3342967.99 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:37:22,326 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=755472.6666666666, ans=0.0
2024-09-25 13:37:48,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=755519.3333333334, ans=0.0
2024-09-25 13:38:13,315 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=11.51 vs. limit=12.0
2024-09-25 13:38:44,292 INFO [train.py:1198] (1/4) Epoch 42, batch 2200, loss[loss=0.1947, ctc_loss=0.1259, cr_loss=0.3442, over 17144.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.1233, cr_loss=0.3415, over 3338769.80 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:38:44,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=755706.0, ans=0.025
2024-09-25 13:38:52,110 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 13:38:59,636 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.309e+02 1.387e+02 1.496e+02 2.285e+02, threshold=2.773e+02, percent-clipped=0.0
2024-09-25 13:39:13,281 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=6.53 vs. limit=15.0
2024-09-25 13:39:18,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=755799.3333333334, ans=0.025
2024-09-25 13:39:39,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=755846.0, ans=0.0
2024-09-25 13:39:44,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=755846.0, ans=0.125
2024-09-25 13:39:51,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=755892.6666666666, ans=0.0
2024-09-25 13:40:09,778 INFO [train.py:1198] (1/4) Epoch 42, batch 2250, loss[loss=0.1628, ctc_loss=0.1021, cr_loss=0.3031, over 17027.00 frames. ], tot_loss[loss=0.1914, ctc_loss=0.1231, cr_loss=0.3414, over 3346150.55 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:40:11,766 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 13:40:13,813 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.59 vs. limit=15.0
2024-09-25 13:40:23,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=755939.3333333334, ans=0.0
2024-09-25 13:40:25,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=755939.3333333334, ans=0.2
2024-09-25 13:40:35,810 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.18 vs. limit=22.5
2024-09-25 13:40:54,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=756032.6666666666, ans=0.2
2024-09-25 13:41:00,368 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.14 vs. limit=22.5
2024-09-25 13:41:00,936 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.max_abs, batch_count=756079.3333333334, ans=10.0
2024-09-25 13:41:10,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=756079.3333333334, ans=0.025
2024-09-25 13:41:23,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=756126.0, ans=0.0
2024-09-25 13:41:33,094 INFO [train.py:1198] (1/4) Epoch 42, batch 2300, loss[loss=0.2029, ctc_loss=0.131, cr_loss=0.3595, over 17136.00 frames. ], tot_loss[loss=0.1905, ctc_loss=0.1225, cr_loss=0.3401, over 3352611.46 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:41:44,574 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=756172.6666666666, ans=0.05
2024-09-25 13:41:45,867 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.265e+02 1.352e+02 1.460e+02 1.967e+02, threshold=2.704e+02, percent-clipped=0.0
2024-09-25 13:41:46,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=756172.6666666666, ans=0.125
2024-09-25 13:41:46,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=756172.6666666666, ans=0.125
2024-09-25 13:42:23,667 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.24 vs. limit=12.0
2024-09-25 13:42:28,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=756312.6666666666, ans=0.125
2024-09-25 13:42:29,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.48 vs. limit=15.0
2024-09-25 13:42:47,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=756359.3333333334, ans=0.1
2024-09-25 13:42:49,284 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.20 vs. limit=6.0
2024-09-25 13:42:53,053 INFO [train.py:1198] (1/4) Epoch 42, batch 2350, loss[loss=0.1562, ctc_loss=0.09696, cr_loss=0.296, over 17094.00 frames. ], tot_loss[loss=0.1904, ctc_loss=0.1223, cr_loss=0.3403, over 3347894.01 frames. ], batch size: 40, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:43:18,938 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.90 vs. limit=12.0
2024-09-25 13:43:20,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=756452.6666666666, ans=0.0
2024-09-25 13:43:27,167 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=16.11 vs. limit=15.0
2024-09-25 13:43:27,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=756499.3333333334, ans=0.0
2024-09-25 13:43:29,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=756499.3333333334, ans=0.125
2024-09-25 13:43:42,662 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.97 vs. limit=6.0
2024-09-25 13:43:54,455 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=756546.0, ans=0.125
2024-09-25 13:43:56,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=756546.0, ans=0.125
2024-09-25 13:44:04,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=756592.6666666666, ans=0.125
2024-09-25 13:44:18,217 INFO [train.py:1198] (1/4) Epoch 42, batch 2400, loss[loss=0.1644, ctc_loss=0.1022, cr_loss=0.3111, over 17058.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.122, cr_loss=0.3397, over 3355305.54 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 32.0
2024-09-25 13:44:26,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=756639.3333333334, ans=0.025
2024-09-25 13:44:30,873 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.304e+02 1.412e+02 1.522e+02 4.391e+02, threshold=2.825e+02, percent-clipped=1.0
2024-09-25 13:44:32,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=756686.0, ans=0.125
2024-09-25 13:44:50,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=756732.6666666666, ans=0.125
2024-09-25 13:45:23,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=756779.3333333334, ans=0.1
2024-09-25 13:45:33,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=756826.0, ans=0.125
2024-09-25 13:45:35,882 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.97 vs. limit=15.0
2024-09-25 13:45:43,398 INFO [train.py:1198] (1/4) Epoch 42, batch 2450, loss[loss=0.2111, ctc_loss=0.1379, cr_loss=0.3663, over 17352.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1225, cr_loss=0.3404, over 3360640.82 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 32.0
2024-09-25 13:45:50,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=756872.6666666666, ans=0.025
2024-09-25 13:46:06,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=756919.3333333334, ans=0.125
2024-09-25 13:46:08,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=756919.3333333334, ans=0.2
2024-09-25 13:46:32,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=757012.6666666666, ans=0.125
2024-09-25 13:46:41,047 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=8.36 vs. limit=15.0
2024-09-25 13:46:50,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.57 vs. limit=15.0
2024-09-25 13:46:56,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=757059.3333333334, ans=0.125
2024-09-25 13:46:56,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=757059.3333333334, ans=0.125
2024-09-25 13:47:03,851 INFO [train.py:1198] (1/4) Epoch 42, batch 2500, loss[loss=0.1644, ctc_loss=0.1021, cr_loss=0.3118, over 16935.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1219, cr_loss=0.3393, over 3362487.16 frames. ], batch size: 42, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:47:15,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=757106.0, ans=0.025
2024-09-25 13:47:18,220 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.315e+02 1.378e+02 1.471e+02 2.327e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 13:48:09,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=757292.6666666666, ans=0.2
2024-09-25 13:48:25,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=757339.3333333334, ans=0.2
2024-09-25 13:48:26,433 INFO [train.py:1198] (1/4) Epoch 42, batch 2550, loss[loss=0.2212, ctc_loss=0.145, cr_loss=0.381, over 16542.00 frames. ], tot_loss[loss=0.1894, ctc_loss=0.1216, cr_loss=0.339, over 3373823.06 frames. ], batch size: 66, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:48:34,617 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=757339.3333333334, ans=0.125
2024-09-25 13:49:11,128 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.12 vs. limit=22.5
2024-09-25 13:49:33,637 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.62 vs. limit=22.5
2024-09-25 13:49:48,975 INFO [train.py:1198] (1/4) Epoch 42, batch 2600, loss[loss=0.1462, ctc_loss=0.09219, cr_loss=0.27, over 16278.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1211, cr_loss=0.3374, over 3372372.65 frames. ], batch size: 36, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:50:05,736 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.054e+02 1.272e+02 1.357e+02 1.427e+02 2.038e+02, threshold=2.713e+02, percent-clipped=0.0
2024-09-25 13:50:09,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.15 vs. limit=10.0
2024-09-25 13:50:32,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=757666.0, ans=0.1
2024-09-25 13:50:32,838 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.38 vs. limit=15.0
2024-09-25 13:51:01,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.79 vs. limit=22.5
2024-09-25 13:51:02,562 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=757759.3333333334, ans=0.125
2024-09-25 13:51:13,648 INFO [train.py:1198] (1/4) Epoch 42, batch 2650, loss[loss=0.1847, ctc_loss=0.1161, cr_loss=0.3429, over 17232.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1218, cr_loss=0.3388, over 3361133.07 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:51:17,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=757806.0, ans=0.125
2024-09-25 13:51:52,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=757899.3333333334, ans=0.125
2024-09-25 13:52:18,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=757992.6666666666, ans=0.125
2024-09-25 13:52:34,235 INFO [train.py:1198] (1/4) Epoch 42, batch 2700, loss[loss=0.1713, ctc_loss=0.1103, cr_loss=0.3047, over 17039.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1219, cr_loss=0.3391, over 3356704.59 frames. ], batch size: 44, lr: 2.82e-03, grad_scale: 8.0
2024-09-25 13:52:45,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=758039.3333333334, ans=0.125
2024-09-25 13:52:50,044 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.301e+02 1.411e+02 1.546e+02 2.916e+02, threshold=2.822e+02, percent-clipped=1.0
2024-09-25 13:52:55,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=758086.0, ans=15.0
2024-09-25 13:52:56,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=758086.0, ans=0.0
2024-09-25 13:53:33,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=758179.3333333334, ans=0.035
2024-09-25 13:53:34,522 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=758179.3333333334, ans=0.2
2024-09-25 13:53:59,182 INFO [train.py:1198] (1/4) Epoch 42, batch 2750, loss[loss=0.1761, ctc_loss=0.1113, cr_loss=0.3241, over 17142.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.122, cr_loss=0.3396, over 3344798.22 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 8.0
2024-09-25 13:54:13,991 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.05 vs. limit=22.5
2024-09-25 13:55:19,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=758459.3333333334, ans=0.125
2024-09-25 13:55:22,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=758506.0, ans=0.025
2024-09-25 13:55:23,620 INFO [train.py:1198] (1/4) Epoch 42, batch 2800, loss[loss=0.1775, ctc_loss=0.1105, cr_loss=0.335, over 17064.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1217, cr_loss=0.3379, over 3332567.37 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:55:39,669 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.295e+02 1.378e+02 1.487e+02 2.331e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 13:56:08,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=758599.3333333334, ans=0.07
2024-09-25 13:56:10,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=758646.0, ans=0.1
2024-09-25 13:56:43,681 INFO [train.py:1198] (1/4) Epoch 42, batch 2850, loss[loss=0.1424, ctc_loss=0.09048, cr_loss=0.2594, over 17123.00 frames. ], tot_loss[loss=0.1886, ctc_loss=0.1212, cr_loss=0.3373, over 3350376.32 frames. ], batch size: 40, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:57:03,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=758786.0, ans=0.125
2024-09-25 13:57:21,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=758832.6666666666, ans=0.1
2024-09-25 13:57:28,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=758832.6666666666, ans=0.125
2024-09-25 13:57:35,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=758879.3333333334, ans=0.0
2024-09-25 13:58:05,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=758972.6666666666, ans=0.125
2024-09-25 13:58:06,778 INFO [train.py:1198] (1/4) Epoch 42, batch 2900, loss[loss=0.1983, ctc_loss=0.1262, cr_loss=0.3609, over 17177.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1218, cr_loss=0.3387, over 3344623.55 frames. ], batch size: 45, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:58:15,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=758972.6666666666, ans=0.04949747468305833
2024-09-25 13:58:22,533 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.263e+02 1.349e+02 1.466e+02 2.283e+02, threshold=2.697e+02, percent-clipped=0.0
2024-09-25 13:58:43,822 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.43 vs. limit=22.5
2024-09-25 13:58:46,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=759066.0, ans=0.125
2024-09-25 13:58:54,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=759066.0, ans=0.125
2024-09-25 13:58:59,225 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=759112.6666666666, ans=0.025
2024-09-25 13:59:12,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=759159.3333333334, ans=0.125
2024-09-25 13:59:29,385 INFO [train.py:1198] (1/4) Epoch 42, batch 2950, loss[loss=0.1872, ctc_loss=0.1224, cr_loss=0.3241, over 17140.00 frames. ], tot_loss[loss=0.1898, ctc_loss=0.1219, cr_loss=0.3393, over 3348411.54 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 13:59:32,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=759206.0, ans=0.0
2024-09-25 13:59:45,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.min_positive, batch_count=759252.6666666666, ans=0.025
2024-09-25 13:59:47,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=759252.6666666666, ans=0.125
2024-09-25 14:00:48,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=759392.6666666666, ans=0.2
2024-09-25 14:00:53,413 INFO [train.py:1198] (1/4) Epoch 42, batch 3000, loss[loss=0.155, ctc_loss=0.09806, cr_loss=0.2847, over 16767.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1223, cr_loss=0.34, over 3354607.94 frames. ], batch size: 37, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 14:00:53,413 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 14:01:08,871 INFO [train.py:1230] (1/4) Epoch 42, validation: loss=0.03543, ctc_loss=0.03543, cr_loss=1.019e-14, over 944034.00 frames. 
2024-09-25 14:01:08,872 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 14:01:12,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=759439.3333333334, ans=0.0
2024-09-25 14:01:24,612 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.302e+02 1.378e+02 1.459e+02 2.338e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 14:01:44,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=759532.6666666666, ans=0.125
2024-09-25 14:02:09,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=759626.0, ans=0.1
2024-09-25 14:02:26,580 INFO [train.py:1198] (1/4) Epoch 42, batch 3050, loss[loss=0.2279, ctc_loss=0.1503, cr_loss=0.388, over 16535.00 frames. ], tot_loss[loss=0.1909, ctc_loss=0.1227, cr_loss=0.341, over 3339165.30 frames. ], batch size: 66, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 14:02:47,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=759719.3333333334, ans=0.025
2024-09-25 14:02:48,209 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.75 vs. limit=22.5
2024-09-25 14:02:51,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=759719.3333333334, ans=10.0
2024-09-25 14:03:18,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=759812.6666666666, ans=0.125
2024-09-25 14:03:45,317 INFO [train.py:1198] (1/4) Epoch 42, batch 3100, loss[loss=0.1762, ctc_loss=0.1118, cr_loss=0.3223, over 17068.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.1228, cr_loss=0.3411, over 3348008.39 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 16.0
2024-09-25 14:04:00,900 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.293e+02 1.378e+02 1.463e+02 2.447e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 14:04:04,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=759952.6666666666, ans=0.0
2024-09-25 14:04:13,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=759952.6666666666, ans=0.125
2024-09-25 14:04:46,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=760092.6666666666, ans=0.125
2024-09-25 14:04:51,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=760092.6666666666, ans=0.2
2024-09-25 14:05:03,355 INFO [train.py:1198] (1/4) Epoch 42, batch 3150, loss[loss=0.1922, ctc_loss=0.1231, cr_loss=0.3451, over 17232.00 frames. ], tot_loss[loss=0.1905, ctc_loss=0.1224, cr_loss=0.3404, over 3351859.42 frames. ], batch size: 55, lr: 2.81e-03, grad_scale: 16.0
2024-09-25 14:05:54,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=760279.3333333334, ans=0.0
2024-09-25 14:06:00,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=760279.3333333334, ans=0.2
2024-09-25 14:06:19,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=760326.0, ans=0.125
2024-09-25 14:06:23,519 INFO [train.py:1198] (1/4) Epoch 42, batch 3200, loss[loss=0.1581, ctc_loss=0.09958, cr_loss=0.2927, over 17210.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.121, cr_loss=0.3363, over 3352978.64 frames. ], batch size: 41, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:06:33,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=760372.6666666666, ans=0.0
2024-09-25 14:06:36,832 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.87 vs. limit=15.0
2024-09-25 14:06:39,248 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.277e+02 1.364e+02 1.425e+02 3.566e+02, threshold=2.728e+02, percent-clipped=1.0
2024-09-25 14:06:40,154 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.87 vs. limit=6.0
2024-09-25 14:06:56,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=760466.0, ans=0.04949747468305833
2024-09-25 14:07:07,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=760466.0, ans=0.125
2024-09-25 14:07:12,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=760512.6666666666, ans=0.2
2024-09-25 14:07:43,979 INFO [train.py:1198] (1/4) Epoch 42, batch 3250, loss[loss=0.2248, ctc_loss=0.1481, cr_loss=0.3837, over 15092.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1206, cr_loss=0.335, over 3353662.65 frames. ], batch size: 89, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:07:52,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=760606.0, ans=0.1
2024-09-25 14:08:49,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=760792.6666666666, ans=0.1
2024-09-25 14:08:51,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=760792.6666666666, ans=0.04949747468305833
2024-09-25 14:08:53,627 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.55 vs. limit=22.5
2024-09-25 14:09:02,228 INFO [train.py:1198] (1/4) Epoch 42, batch 3300, loss[loss=0.2162, ctc_loss=0.1412, cr_loss=0.375, over 17219.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1216, cr_loss=0.3364, over 3339828.71 frames. ], batch size: 47, lr: 2.81e-03, grad_scale: 16.0
2024-09-25 14:09:19,406 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.292e+02 1.397e+02 1.503e+02 2.274e+02, threshold=2.794e+02, percent-clipped=0.0
2024-09-25 14:09:19,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=760886.0, ans=0.125
2024-09-25 14:09:21,152 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=760886.0, ans=0.125
2024-09-25 14:09:30,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=760886.0, ans=0.2
2024-09-25 14:09:51,367 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.43 vs. limit=10.0
2024-09-25 14:10:14,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=761026.0, ans=0.1
2024-09-25 14:10:22,239 INFO [train.py:1198] (1/4) Epoch 42, batch 3350, loss[loss=0.2174, ctc_loss=0.1414, cr_loss=0.3798, over 16718.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1217, cr_loss=0.3369, over 3347077.76 frames. ], batch size: 61, lr: 2.81e-03, grad_scale: 16.0
2024-09-25 14:10:25,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=761072.6666666666, ans=0.125
2024-09-25 14:10:30,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=761072.6666666666, ans=0.025
2024-09-25 14:10:32,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=761072.6666666666, ans=0.125
2024-09-25 14:10:36,013 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.01 vs. limit=15.0
2024-09-25 14:11:02,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=761166.0, ans=0.125
2024-09-25 14:11:23,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=761212.6666666666, ans=0.2
2024-09-25 14:11:29,897 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.78 vs. limit=12.0
2024-09-25 14:11:31,522 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.21 vs. limit=22.5
2024-09-25 14:11:38,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=761259.3333333334, ans=0.0
2024-09-25 14:11:42,862 INFO [train.py:1198] (1/4) Epoch 42, batch 3400, loss[loss=0.1914, ctc_loss=0.1212, cr_loss=0.3512, over 16994.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1224, cr_loss=0.3382, over 3346626.66 frames. ], batch size: 53, lr: 2.81e-03, grad_scale: 16.0
2024-09-25 14:12:00,049 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.312e+02 1.394e+02 1.492e+02 2.078e+02, threshold=2.788e+02, percent-clipped=0.0
2024-09-25 14:12:05,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=761352.6666666666, ans=0.2
2024-09-25 14:12:43,067 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=761446.0, ans=0.0
2024-09-25 14:12:50,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=761492.6666666666, ans=0.0
2024-09-25 14:13:01,388 INFO [train.py:1198] (1/4) Epoch 42, batch 3450, loss[loss=0.2211, ctc_loss=0.1489, cr_loss=0.3608, over 11555.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1223, cr_loss=0.3385, over 3350536.09 frames. ], batch size: 123, lr: 2.81e-03, grad_scale: 16.0
2024-09-25 14:13:25,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=761586.0, ans=0.125
2024-09-25 14:13:44,456 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.28 vs. limit=22.5
2024-09-25 14:14:03,787 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.07 vs. limit=15.0
2024-09-25 14:14:13,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=761726.0, ans=0.125
2024-09-25 14:14:19,996 INFO [train.py:1198] (1/4) Epoch 42, batch 3500, loss[loss=0.1857, ctc_loss=0.1188, cr_loss=0.3345, over 17302.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1215, cr_loss=0.3375, over 3353832.06 frames. ], batch size: 49, lr: 2.81e-03, grad_scale: 16.0
2024-09-25 14:14:20,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=761772.6666666666, ans=10.0
2024-09-25 14:14:20,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=761772.6666666666, ans=0.95
2024-09-25 14:14:26,422 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=761772.6666666666, ans=0.125
2024-09-25 14:14:37,159 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.281e+02 1.384e+02 1.487e+02 1.767e+02, threshold=2.768e+02, percent-clipped=0.0
2024-09-25 14:14:40,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=761819.3333333334, ans=0.0
2024-09-25 14:15:16,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=761912.6666666666, ans=0.1
2024-09-25 14:15:19,095 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.59 vs. limit=15.0
2024-09-25 14:15:19,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=761912.6666666666, ans=0.125
2024-09-25 14:15:40,263 INFO [train.py:1198] (1/4) Epoch 42, batch 3550, loss[loss=0.1773, ctc_loss=0.1143, cr_loss=0.3153, over 17217.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.1221, cr_loss=0.3392, over 3353487.86 frames. ], batch size: 47, lr: 2.81e-03, grad_scale: 16.0
2024-09-25 14:15:42,521 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=8.19 vs. limit=15.0
2024-09-25 14:16:15,234 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=6.76 vs. limit=15.0
2024-09-25 14:16:36,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=762146.0, ans=0.125
2024-09-25 14:16:37,340 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.97 vs. limit=15.0
2024-09-25 14:16:46,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=762192.6666666666, ans=0.125
2024-09-25 14:16:49,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=762192.6666666666, ans=0.125
2024-09-25 14:16:55,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=12.38 vs. limit=12.0
2024-09-25 14:16:58,304 INFO [train.py:1198] (1/4) Epoch 42, batch 3600, loss[loss=0.2054, ctc_loss=0.1319, cr_loss=0.3676, over 17096.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1223, cr_loss=0.3397, over 3361145.40 frames. ], batch size: 49, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:17:01,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=762239.3333333334, ans=0.07
2024-09-25 14:17:10,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=762239.3333333334, ans=0.1
2024-09-25 14:17:15,187 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.282e+02 1.377e+02 1.493e+02 1.761e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-25 14:17:15,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=762286.0, ans=0.125
2024-09-25 14:17:38,475 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.98 vs. limit=22.5
2024-09-25 14:18:08,859 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=762426.0, ans=0.0
2024-09-25 14:18:11,186 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.48 vs. limit=6.0
2024-09-25 14:18:17,993 INFO [train.py:1198] (1/4) Epoch 42, batch 3650, loss[loss=0.2248, ctc_loss=0.1437, cr_loss=0.4055, over 16755.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.122, cr_loss=0.3394, over 3359483.43 frames. ], batch size: 61, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:18:51,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=762566.0, ans=0.125
2024-09-25 14:18:54,642 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=762566.0, ans=0.125
2024-09-25 14:19:12,656 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.10 vs. limit=22.5
2024-09-25 14:19:30,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=762659.3333333334, ans=0.125
2024-09-25 14:19:38,301 INFO [train.py:1198] (1/4) Epoch 42, batch 3700, loss[loss=0.1698, ctc_loss=0.1101, cr_loss=0.2987, over 16947.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1216, cr_loss=0.338, over 3343619.36 frames. ], batch size: 42, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:19:55,539 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.181e+02 1.273e+02 1.387e+02 1.507e+02 1.911e+02, threshold=2.774e+02, percent-clipped=0.0
2024-09-25 14:20:07,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=762752.6666666666, ans=0.125
2024-09-25 14:20:12,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=762799.3333333334, ans=0.0
2024-09-25 14:20:15,492 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=762799.3333333334, ans=0.1
2024-09-25 14:20:35,787 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=762846.0, ans=0.125
2024-09-25 14:20:57,614 INFO [train.py:1198] (1/4) Epoch 42, batch 3750, loss[loss=0.2188, ctc_loss=0.1481, cr_loss=0.3537, over 14985.00 frames. ], tot_loss[loss=0.1901, ctc_loss=0.1223, cr_loss=0.339, over 3331140.47 frames. ], batch size: 89, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:21:19,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=762986.0, ans=0.2
2024-09-25 14:21:39,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=763032.6666666666, ans=0.125
2024-09-25 14:21:41,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=763032.6666666666, ans=0.1
2024-09-25 14:21:43,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=763079.3333333334, ans=0.2
2024-09-25 14:22:03,813 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.44 vs. limit=15.0
2024-09-25 14:22:15,973 INFO [train.py:1198] (1/4) Epoch 42, batch 3800, loss[loss=0.1985, ctc_loss=0.1295, cr_loss=0.3449, over 17007.00 frames. ], tot_loss[loss=0.1901, ctc_loss=0.1224, cr_loss=0.3386, over 3307649.64 frames. ], batch size: 51, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:22:33,278 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.315e+02 1.408e+02 1.487e+02 2.777e+02, threshold=2.816e+02, percent-clipped=1.0
2024-09-25 14:23:02,308 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.95 vs. limit=22.5
2024-09-25 14:23:04,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=763312.6666666666, ans=0.07
2024-09-25 14:23:22,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=763359.3333333334, ans=0.2
2024-09-25 14:23:26,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=763359.3333333334, ans=0.125
2024-09-25 14:23:28,507 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 14:23:30,639 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=6.26 vs. limit=12.0
2024-09-25 14:23:34,530 INFO [train.py:1198] (1/4) Epoch 42, batch 3850, loss[loss=0.1543, ctc_loss=0.09577, cr_loss=0.2929, over 17029.00 frames. ], tot_loss[loss=0.1904, ctc_loss=0.1227, cr_loss=0.3384, over 3267022.88 frames. ], batch size: 39, lr: 2.81e-03, grad_scale: 32.0
2024-09-25 14:23:52,120 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 14:25:36,253 INFO [train.py:1198] (1/4) Epoch 43, batch 0, loss[loss=0.1868, ctc_loss=0.1229, cr_loss=0.3192, over 16613.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1229, cr_loss=0.3192, over 16613.00 frames. ], batch size: 66, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:25:36,253 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 14:25:49,917 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.2.self_attn_weights, attn_weights_entropy = tensor([2.3218, 2.7715, 2.6785, 2.7191, 2.5473, 2.5216, 2.8495, 2.8931],
       device='cuda:1')
2024-09-25 14:25:51,491 INFO [train.py:1230] (1/4) Epoch 43, validation: loss=0.03486, ctc_loss=0.03486, cr_loss=1.051e-14, over 944034.00 frames. 
2024-09-25 14:25:51,491 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 14:25:58,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=763620.6666666666, ans=0.125
2024-09-25 14:26:04,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=763620.6666666666, ans=0.0
2024-09-25 14:26:15,238 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.334e+02 1.498e+02 1.673e+02 2.107e+02, threshold=2.995e+02, percent-clipped=0.0
2024-09-25 14:26:28,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=763714.0, ans=0.125
2024-09-25 14:26:48,756 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 14:27:10,606 INFO [train.py:1198] (1/4) Epoch 43, batch 50, loss[loss=0.1771, ctc_loss=0.1135, cr_loss=0.3179, over 17043.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1206, cr_loss=0.3397, over 762596.47 frames. ], batch size: 51, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:27:12,801 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.85 vs. limit=15.0
2024-09-25 14:27:22,593 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.67 vs. limit=15.0
2024-09-25 14:27:30,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=763900.6666666666, ans=0.125
2024-09-25 14:28:28,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=764087.3333333334, ans=10.0
2024-09-25 14:28:30,181 INFO [train.py:1198] (1/4) Epoch 43, batch 100, loss[loss=0.2015, ctc_loss=0.1326, cr_loss=0.3444, over 17300.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1207, cr_loss=0.3381, over 1345331.84 frames. ], batch size: 49, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:28:54,052 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.307e+02 1.401e+02 1.473e+02 2.012e+02, threshold=2.802e+02, percent-clipped=0.0
2024-09-25 14:29:15,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=764180.6666666666, ans=0.1
2024-09-25 14:29:25,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=764227.3333333334, ans=0.1
2024-09-25 14:29:44,084 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=9.03 vs. limit=22.5
2024-09-25 14:29:50,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=764274.0, ans=0.2
2024-09-25 14:29:54,510 INFO [train.py:1198] (1/4) Epoch 43, batch 150, loss[loss=0.1564, ctc_loss=0.0989, cr_loss=0.2875, over 16691.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.1218, cr_loss=0.3405, over 1795841.36 frames. ], batch size: 37, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:30:00,590 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.42 vs. limit=15.0
2024-09-25 14:30:17,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=764367.3333333334, ans=0.125
2024-09-25 14:30:18,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=764367.3333333334, ans=0.1
2024-09-25 14:30:27,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=764414.0, ans=0.1
2024-09-25 14:30:28,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=764414.0, ans=0.125
2024-09-25 14:30:28,919 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=764414.0, ans=0.1
2024-09-25 14:30:57,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=764460.6666666666, ans=0.125
2024-09-25 14:31:09,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=764507.3333333334, ans=0.05
2024-09-25 14:31:17,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=764507.3333333334, ans=0.1
2024-09-25 14:31:19,818 INFO [train.py:1198] (1/4) Epoch 43, batch 200, loss[loss=0.1651, ctc_loss=0.1065, cr_loss=0.2928, over 17175.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.1199, cr_loss=0.3366, over 2151950.17 frames. ], batch size: 45, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:31:29,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=764554.0, ans=0.125
2024-09-25 14:31:43,639 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.287e+02 1.384e+02 1.479e+02 1.740e+02, threshold=2.769e+02, percent-clipped=0.0
2024-09-25 14:31:51,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=764647.3333333334, ans=0.0
2024-09-25 14:32:18,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=764694.0, ans=0.0
2024-09-25 14:32:22,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=764740.6666666666, ans=0.0
2024-09-25 14:32:39,303 INFO [train.py:1198] (1/4) Epoch 43, batch 250, loss[loss=0.2009, ctc_loss=0.1278, cr_loss=0.3654, over 17073.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1206, cr_loss=0.3379, over 2410460.97 frames. ], batch size: 46, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:33:32,994 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=764927.3333333334, ans=0.125
2024-09-25 14:33:39,730 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.16 vs. limit=6.0
2024-09-25 14:33:59,384 INFO [train.py:1198] (1/4) Epoch 43, batch 300, loss[loss=0.2019, ctc_loss=0.1316, cr_loss=0.3512, over 16704.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1215, cr_loss=0.3393, over 2627960.59 frames. ], batch size: 61, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:34:16,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=765067.3333333334, ans=0.125
2024-09-25 14:34:19,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=765067.3333333334, ans=0.2
2024-09-25 14:34:25,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.77 vs. limit=12.0
2024-09-25 14:34:25,901 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.299e+02 1.370e+02 1.451e+02 2.001e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-25 14:34:37,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=765114.0, ans=0.0
2024-09-25 14:35:10,577 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 14:35:24,626 INFO [train.py:1198] (1/4) Epoch 43, batch 350, loss[loss=0.1711, ctc_loss=0.1093, cr_loss=0.3092, over 16928.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1212, cr_loss=0.3393, over 2785940.37 frames. ], batch size: 42, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:36:20,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=765394.0, ans=0.125
2024-09-25 14:36:27,381 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.96 vs. limit=15.0
2024-09-25 14:36:41,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=765440.6666666666, ans=0.0
2024-09-25 14:36:51,771 INFO [train.py:1198] (1/4) Epoch 43, batch 400, loss[loss=0.1507, ctc_loss=0.09594, cr_loss=0.274, over 16356.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1211, cr_loss=0.3392, over 2914221.67 frames. ], batch size: 36, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:36:52,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=765487.3333333334, ans=0.125
2024-09-25 14:37:15,438 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.295e+02 1.341e+02 1.449e+02 2.336e+02, threshold=2.681e+02, percent-clipped=0.0
2024-09-25 14:37:31,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=765580.6666666666, ans=0.025
2024-09-25 14:37:35,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=765580.6666666666, ans=0.2
2024-09-25 14:37:38,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=765627.3333333334, ans=0.125
2024-09-25 14:38:02,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=765674.0, ans=0.025
2024-09-25 14:38:11,327 INFO [train.py:1198] (1/4) Epoch 43, batch 450, loss[loss=0.1401, ctc_loss=0.08949, cr_loss=0.2529, over 17197.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1203, cr_loss=0.3366, over 3008516.72 frames. ], batch size: 41, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:38:11,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=765720.6666666666, ans=0.0
2024-09-25 14:38:11,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=765720.6666666666, ans=0.0
2024-09-25 14:38:17,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=765720.6666666666, ans=0.035
2024-09-25 14:39:02,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=765860.6666666666, ans=0.125
2024-09-25 14:39:02,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=765860.6666666666, ans=0.1
2024-09-25 14:39:04,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=765860.6666666666, ans=0.2
2024-09-25 14:39:33,566 INFO [train.py:1198] (1/4) Epoch 43, batch 500, loss[loss=0.2164, ctc_loss=0.1407, cr_loss=0.3787, over 17107.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1207, cr_loss=0.3372, over 3085620.80 frames. ], batch size: 49, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:39:52,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=766000.6666666666, ans=0.125
2024-09-25 14:39:54,172 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=766000.6666666666, ans=0.0
2024-09-25 14:39:55,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=766000.6666666666, ans=0.025
2024-09-25 14:40:00,058 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.354e+02 1.449e+02 1.544e+02 3.228e+02, threshold=2.898e+02, percent-clipped=1.0
2024-09-25 14:40:06,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=766047.3333333334, ans=0.2
2024-09-25 14:40:40,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=766140.6666666666, ans=0.125
2024-09-25 14:40:49,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=766140.6666666666, ans=0.125
2024-09-25 14:40:57,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=766140.6666666666, ans=0.09899494936611666
2024-09-25 14:41:00,788 INFO [train.py:1198] (1/4) Epoch 43, batch 550, loss[loss=0.175, ctc_loss=0.1107, cr_loss=0.3212, over 17226.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1207, cr_loss=0.3371, over 3144684.32 frames. ], batch size: 50, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:41:04,627 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.24 vs. limit=12.0
2024-09-25 14:41:09,272 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.04 vs. limit=6.0
2024-09-25 14:41:10,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=766187.3333333334, ans=0.07
2024-09-25 14:41:12,728 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=4.16 vs. limit=6.0
2024-09-25 14:41:15,215 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=766234.0, ans=0.125
2024-09-25 14:41:23,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=766234.0, ans=0.0
2024-09-25 14:41:31,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=766280.6666666666, ans=0.025
2024-09-25 14:41:39,704 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.44 vs. limit=15.0
2024-09-25 14:41:56,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=766327.3333333334, ans=0.125
2024-09-25 14:42:20,463 INFO [train.py:1198] (1/4) Epoch 43, batch 600, loss[loss=0.2155, ctc_loss=0.1398, cr_loss=0.3784, over 17233.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1203, cr_loss=0.3363, over 3190387.07 frames. ], batch size: 50, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:42:30,338 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer_na.min_abs, batch_count=766420.6666666666, ans=0.02
2024-09-25 14:42:35,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=766467.3333333334, ans=0.1
2024-09-25 14:42:35,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=766467.3333333334, ans=0.2
2024-09-25 14:42:38,819 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.81 vs. limit=15.0
2024-09-25 14:42:44,284 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.302e+02 1.423e+02 1.505e+02 3.409e+02, threshold=2.845e+02, percent-clipped=1.0
2024-09-25 14:43:24,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=766607.3333333334, ans=0.0
2024-09-25 14:43:40,674 INFO [train.py:1198] (1/4) Epoch 43, batch 650, loss[loss=0.1918, ctc_loss=0.1204, cr_loss=0.357, over 17318.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1198, cr_loss=0.3356, over 3238002.74 frames. ], batch size: 49, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:44:00,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=766700.6666666666, ans=0.2
2024-09-25 14:44:08,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=766700.6666666666, ans=0.0
2024-09-25 14:44:10,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=766700.6666666666, ans=0.0
2024-09-25 14:44:50,792 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=766840.6666666666, ans=0.125
2024-09-25 14:45:06,411 INFO [train.py:1198] (1/4) Epoch 43, batch 700, loss[loss=0.1743, ctc_loss=0.11, cr_loss=0.3214, over 17162.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1196, cr_loss=0.335, over 3267558.55 frames. ], batch size: 45, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:45:24,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=766934.0, ans=0.04949747468305833
2024-09-25 14:45:30,412 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.308e+02 1.372e+02 1.490e+02 1.932e+02, threshold=2.743e+02, percent-clipped=0.0
2024-09-25 14:45:45,561 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=19.66 vs. limit=22.5
2024-09-25 14:46:16,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=767074.0, ans=0.1
2024-09-25 14:46:26,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=767074.0, ans=0.0
2024-09-25 14:46:29,245 INFO [train.py:1198] (1/4) Epoch 43, batch 750, loss[loss=0.1434, ctc_loss=0.08888, cr_loss=0.2728, over 17116.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1205, cr_loss=0.3363, over 3292846.67 frames. ], batch size: 40, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:46:31,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=767120.6666666666, ans=0.0
2024-09-25 14:46:51,742 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 14:47:04,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=767214.0, ans=0.125
2024-09-25 14:47:06,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_abs, batch_count=767214.0, ans=0.5
2024-09-25 14:47:09,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer_ff2.min_abs, batch_count=767214.0, ans=0.1
2024-09-25 14:47:15,219 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.50 vs. limit=22.5
2024-09-25 14:47:49,078 INFO [train.py:1198] (1/4) Epoch 43, batch 800, loss[loss=0.212, ctc_loss=0.1379, cr_loss=0.3704, over 17018.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1205, cr_loss=0.3367, over 3309025.14 frames. ], batch size: 44, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:47:53,303 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=9.83 vs. limit=15.0
2024-09-25 14:48:11,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=767400.6666666666, ans=0.0
2024-09-25 14:48:12,577 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.303e+02 1.387e+02 1.475e+02 2.331e+02, threshold=2.774e+02, percent-clipped=0.0
2024-09-25 14:48:52,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=767540.6666666666, ans=0.125
2024-09-25 14:49:01,590 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.88 vs. limit=10.0
2024-09-25 14:49:08,214 INFO [train.py:1198] (1/4) Epoch 43, batch 850, loss[loss=0.1441, ctc_loss=0.08962, cr_loss=0.2722, over 16249.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1205, cr_loss=0.3363, over 3327806.74 frames. ], batch size: 36, lr: 2.77e-03, grad_scale: 32.0
2024-09-25 14:49:20,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=767587.3333333334, ans=0.125
2024-09-25 14:49:30,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=767634.0, ans=0.2
2024-09-25 14:49:58,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=767680.6666666666, ans=0.015
2024-09-25 14:50:01,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=767727.3333333334, ans=0.125
2024-09-25 14:50:06,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=767727.3333333334, ans=0.1
2024-09-25 14:50:08,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=767727.3333333334, ans=0.0
2024-09-25 14:50:13,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=767727.3333333334, ans=0.0
2024-09-25 14:50:16,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=767774.0, ans=0.0
2024-09-25 14:50:33,649 INFO [train.py:1198] (1/4) Epoch 43, batch 900, loss[loss=0.1791, ctc_loss=0.1133, cr_loss=0.3292, over 17205.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1208, cr_loss=0.337, over 3337002.45 frames. ], batch size: 50, lr: 2.77e-03, grad_scale: 16.0
2024-09-25 14:50:58,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=767867.3333333334, ans=0.0
2024-09-25 14:51:03,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=767867.3333333334, ans=0.1
2024-09-25 14:51:04,267 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.318e+02 1.401e+02 1.527e+02 2.167e+02, threshold=2.803e+02, percent-clipped=0.0
2024-09-25 14:51:17,847 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=3.94 vs. limit=15.0
2024-09-25 14:51:47,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer_ff2.min_abs, batch_count=768007.3333333334, ans=0.1
2024-09-25 14:51:57,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=768054.0, ans=0.125
2024-09-25 14:51:58,511 INFO [train.py:1198] (1/4) Epoch 43, batch 950, loss[loss=0.193, ctc_loss=0.1238, cr_loss=0.3458, over 17224.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1203, cr_loss=0.3356, over 3343779.64 frames. ], batch size: 47, lr: 2.77e-03, grad_scale: 16.0
2024-09-25 14:52:05,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=768054.0, ans=0.0
2024-09-25 14:52:08,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=768054.0, ans=0.125
2024-09-25 14:52:29,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=768147.3333333334, ans=0.125
2024-09-25 14:52:58,957 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.18 vs. limit=6.0
2024-09-25 14:53:06,819 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.70 vs. limit=10.0
2024-09-25 14:53:18,817 INFO [train.py:1198] (1/4) Epoch 43, batch 1000, loss[loss=0.2092, ctc_loss=0.134, cr_loss=0.3759, over 17211.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1208, cr_loss=0.3363, over 3333306.60 frames. ], batch size: 55, lr: 2.77e-03, grad_scale: 16.0
2024-09-25 14:53:23,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=768287.3333333334, ans=0.0
2024-09-25 14:53:44,360 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.294e+02 1.416e+02 1.489e+02 2.363e+02, threshold=2.832e+02, percent-clipped=0.0
2024-09-25 14:54:03,281 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.59 vs. limit=15.0
2024-09-25 14:54:07,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=768427.3333333334, ans=0.0
2024-09-25 14:54:10,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=768427.3333333334, ans=0.125
2024-09-25 14:54:13,478 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=768427.3333333334, ans=0.2
2024-09-25 14:54:44,013 INFO [train.py:1198] (1/4) Epoch 43, batch 1050, loss[loss=0.2264, ctc_loss=0.1483, cr_loss=0.3904, over 16909.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.121, cr_loss=0.3364, over 3333730.39 frames. ], batch size: 58, lr: 2.77e-03, grad_scale: 16.0
2024-09-25 14:55:12,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=768567.3333333334, ans=0.125
2024-09-25 14:55:24,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=768614.0, ans=0.0
2024-09-25 14:55:27,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=768614.0, ans=0.0
2024-09-25 14:55:29,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=768614.0, ans=0.125
2024-09-25 14:55:37,480 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.90 vs. limit=15.0
2024-09-25 14:55:56,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=768707.3333333334, ans=0.1
2024-09-25 14:56:08,918 INFO [train.py:1198] (1/4) Epoch 43, batch 1100, loss[loss=0.2294, ctc_loss=0.1547, cr_loss=0.3737, over 14816.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1208, cr_loss=0.3367, over 3343022.53 frames. ], batch size: 88, lr: 2.77e-03, grad_scale: 16.0
2024-09-25 14:56:12,310 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=768754.0, ans=0.125
2024-09-25 14:56:34,424 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.288e+02 1.354e+02 1.479e+02 2.223e+02, threshold=2.708e+02, percent-clipped=0.0
2024-09-25 14:56:49,928 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.59 vs. limit=6.0
2024-09-25 14:57:01,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=768894.0, ans=0.125
2024-09-25 14:57:09,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=768894.0, ans=0.2
2024-09-25 14:57:11,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=768940.6666666666, ans=0.2
2024-09-25 14:57:12,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=768940.6666666666, ans=0.125
2024-09-25 14:57:15,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=768940.6666666666, ans=0.07
2024-09-25 14:57:23,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=768940.6666666666, ans=0.1
2024-09-25 14:57:24,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=768940.6666666666, ans=0.0
2024-09-25 14:57:28,424 INFO [train.py:1198] (1/4) Epoch 43, batch 1150, loss[loss=0.2039, ctc_loss=0.1304, cr_loss=0.3672, over 17022.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1214, cr_loss=0.3382, over 3347451.26 frames. ], batch size: 53, lr: 2.77e-03, grad_scale: 16.0
2024-09-25 14:57:33,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=768987.3333333334, ans=0.0
2024-09-25 14:57:41,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=768987.3333333334, ans=0.125
2024-09-25 14:57:46,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=769034.0, ans=0.09899494936611666
2024-09-25 14:57:48,470 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.48 vs. limit=15.0
2024-09-25 14:58:09,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=769080.6666666666, ans=0.025
2024-09-25 14:58:26,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=769127.3333333334, ans=0.125
2024-09-25 14:58:26,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=769127.3333333334, ans=0.0
2024-09-25 14:58:48,784 INFO [train.py:1198] (1/4) Epoch 43, batch 1200, loss[loss=0.1968, ctc_loss=0.1276, cr_loss=0.3462, over 17030.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1208, cr_loss=0.3373, over 3352349.90 frames. ], batch size: 51, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 14:59:11,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=769267.3333333334, ans=0.125
2024-09-25 14:59:14,325 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.304e+02 1.385e+02 1.497e+02 1.972e+02, threshold=2.770e+02, percent-clipped=0.0
2024-09-25 14:59:28,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=769314.0, ans=0.07
2024-09-25 14:59:45,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=769360.6666666666, ans=0.025
2024-09-25 15:00:13,438 INFO [train.py:1198] (1/4) Epoch 43, batch 1250, loss[loss=0.2349, ctc_loss=0.1569, cr_loss=0.39, over 11388.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1214, cr_loss=0.3387, over 3340246.14 frames. ], batch size: 123, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:00:27,042 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=15.57 vs. limit=22.5
2024-09-25 15:00:45,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=769500.6666666666, ans=0.0
2024-09-25 15:01:04,484 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.71 vs. limit=15.0
2024-09-25 15:01:08,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=769594.0, ans=0.125
2024-09-25 15:01:16,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=769594.0, ans=0.125
2024-09-25 15:01:31,678 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.82 vs. limit=22.5
2024-09-25 15:01:34,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=769640.6666666666, ans=0.1
2024-09-25 15:01:38,710 INFO [train.py:1198] (1/4) Epoch 43, batch 1300, loss[loss=0.1711, ctc_loss=0.1082, cr_loss=0.3145, over 17272.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.121, cr_loss=0.3383, over 3336904.24 frames. ], batch size: 42, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:02:01,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.min_positive, batch_count=769734.0, ans=0.05
2024-09-25 15:02:04,021 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.298e+02 1.391e+02 1.472e+02 1.717e+02, threshold=2.781e+02, percent-clipped=0.0
2024-09-25 15:02:11,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten.whitening_limit, batch_count=769780.6666666666, ans=15.0
2024-09-25 15:02:31,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=769827.3333333334, ans=0.125
2024-09-25 15:02:41,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=769874.0, ans=0.125
2024-09-25 15:02:58,578 INFO [train.py:1198] (1/4) Epoch 43, batch 1350, loss[loss=0.154, ctc_loss=0.0984, cr_loss=0.2778, over 17234.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1214, cr_loss=0.3388, over 3341448.58 frames. ], batch size: 41, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:03:19,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=769967.3333333334, ans=0.125
2024-09-25 15:04:04,172 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=13.15 vs. limit=15.0
2024-09-25 15:04:06,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=770107.3333333334, ans=0.2
2024-09-25 15:04:21,711 INFO [train.py:1198] (1/4) Epoch 43, batch 1400, loss[loss=0.1962, ctc_loss=0.1245, cr_loss=0.3587, over 16006.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1214, cr_loss=0.3381, over 3339528.96 frames. ], batch size: 74, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:04:44,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.52 vs. limit=10.0
2024-09-25 15:04:50,140 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.179e+02 1.309e+02 1.380e+02 1.484e+02 2.530e+02, threshold=2.759e+02, percent-clipped=0.0
2024-09-25 15:05:09,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=770247.3333333334, ans=0.0
2024-09-25 15:05:20,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=770294.0, ans=0.09899494936611666
2024-09-25 15:05:46,977 INFO [train.py:1198] (1/4) Epoch 43, batch 1450, loss[loss=0.1714, ctc_loss=0.1076, cr_loss=0.3193, over 17268.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1219, cr_loss=0.3392, over 3353265.87 frames. ], batch size: 42, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:05:50,914 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=9.25 vs. limit=12.0
2024-09-25 15:05:52,775 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.19 vs. limit=15.0
2024-09-25 15:06:36,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=770527.3333333334, ans=0.125
2024-09-25 15:06:58,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=770574.0, ans=0.0
2024-09-25 15:07:09,744 INFO [train.py:1198] (1/4) Epoch 43, batch 1500, loss[loss=0.1654, ctc_loss=0.1061, cr_loss=0.2963, over 17165.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1217, cr_loss=0.3391, over 3363660.86 frames. ], batch size: 41, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:07:13,620 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.92 vs. limit=22.5
2024-09-25 15:07:27,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=770667.3333333334, ans=0.0
2024-09-25 15:07:29,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=770667.3333333334, ans=0.125
2024-09-25 15:07:35,090 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.284e+02 1.350e+02 1.419e+02 1.777e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-25 15:07:59,795 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.04 vs. limit=15.0
2024-09-25 15:08:03,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.66 vs. limit=10.0
2024-09-25 15:08:21,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=770807.3333333334, ans=0.1
2024-09-25 15:08:26,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=770807.3333333334, ans=0.07
2024-09-25 15:08:28,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=770854.0, ans=0.025
2024-09-25 15:08:29,384 INFO [train.py:1198] (1/4) Epoch 43, batch 1550, loss[loss=0.1729, ctc_loss=0.1089, cr_loss=0.3198, over 17190.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1215, cr_loss=0.3391, over 3365916.90 frames. ], batch size: 41, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:08:37,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=5.89 vs. limit=12.0
2024-09-25 15:08:42,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=770854.0, ans=0.0
2024-09-25 15:09:29,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=770994.0, ans=0.125
2024-09-25 15:09:36,161 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.95 vs. limit=12.0
2024-09-25 15:09:45,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=771040.6666666666, ans=0.125
2024-09-25 15:09:54,492 INFO [train.py:1198] (1/4) Epoch 43, batch 1600, loss[loss=0.1861, ctc_loss=0.1201, cr_loss=0.3296, over 17295.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1214, cr_loss=0.339, over 3375388.16 frames. ], batch size: 51, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:10:00,006 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.93 vs. limit=15.0
2024-09-25 15:10:09,919 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=7.67 vs. limit=15.0
2024-09-25 15:10:20,125 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.278e+02 1.396e+02 1.510e+02 2.224e+02, threshold=2.791e+02, percent-clipped=0.0
2024-09-25 15:10:20,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=771134.0, ans=0.09899494936611666
2024-09-25 15:10:47,920 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=9.45 vs. limit=15.0
2024-09-25 15:10:54,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=771227.3333333334, ans=0.125
2024-09-25 15:10:59,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=771227.3333333334, ans=0.2
2024-09-25 15:11:08,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=771274.0, ans=0.0
2024-09-25 15:11:19,901 INFO [train.py:1198] (1/4) Epoch 43, batch 1650, loss[loss=0.1729, ctc_loss=0.1123, cr_loss=0.3028, over 17188.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1205, cr_loss=0.3369, over 3370885.35 frames. ], batch size: 41, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:11:20,250 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=771320.6666666666, ans=0.125
2024-09-25 15:11:39,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=771367.3333333334, ans=0.125
2024-09-25 15:11:42,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=771367.3333333334, ans=0.125
2024-09-25 15:11:57,568 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.60 vs. limit=10.0
2024-09-25 15:12:23,592 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=771507.3333333334, ans=0.125
2024-09-25 15:12:39,451 INFO [train.py:1198] (1/4) Epoch 43, batch 1700, loss[loss=0.2029, ctc_loss=0.1296, cr_loss=0.3667, over 17041.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.1199, cr_loss=0.3363, over 3366710.59 frames. ], batch size: 56, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:12:55,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=771600.6666666666, ans=0.0
2024-09-25 15:13:05,042 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.300e+02 1.382e+02 1.486e+02 1.912e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-25 15:13:05,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=771600.6666666666, ans=0.125
2024-09-25 15:13:10,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=771647.3333333334, ans=0.05
2024-09-25 15:13:50,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=771740.6666666666, ans=0.0
2024-09-25 15:14:00,170 INFO [train.py:1198] (1/4) Epoch 43, batch 1750, loss[loss=0.1855, ctc_loss=0.1195, cr_loss=0.3298, over 17149.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1205, cr_loss=0.3369, over 3357416.77 frames. ], batch size: 45, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:14:00,977 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.37 vs. limit=15.0
2024-09-25 15:14:22,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=771834.0, ans=0.0
2024-09-25 15:14:27,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=771834.0, ans=0.0
2024-09-25 15:15:20,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=771974.0, ans=0.2
2024-09-25 15:15:24,845 INFO [train.py:1198] (1/4) Epoch 43, batch 1800, loss[loss=0.2016, ctc_loss=0.1396, cr_loss=0.31, over 11792.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1212, cr_loss=0.3387, over 3349677.65 frames. ], batch size: 123, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:15:38,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=772020.6666666666, ans=0.1
2024-09-25 15:15:50,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=772067.3333333334, ans=0.125
2024-09-25 15:15:54,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=772067.3333333334, ans=0.125
2024-09-25 15:15:55,551 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.295e+02 1.383e+02 1.459e+02 2.589e+02, threshold=2.767e+02, percent-clipped=0.0
2024-09-25 15:16:02,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=772114.0, ans=0.0
2024-09-25 15:16:16,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=772160.6666666666, ans=0.0
2024-09-25 15:16:18,576 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=11.62 vs. limit=12.0
2024-09-25 15:16:29,788 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.36 vs. limit=15.0
2024-09-25 15:16:42,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=772207.3333333334, ans=0.025
2024-09-25 15:16:49,941 INFO [train.py:1198] (1/4) Epoch 43, batch 1850, loss[loss=0.2153, ctc_loss=0.1398, cr_loss=0.3772, over 16692.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1207, cr_loss=0.338, over 3353464.91 frames. ], batch size: 61, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:17:30,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=772347.3333333334, ans=0.125
2024-09-25 15:17:45,142 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=772394.0, ans=0.125
2024-09-25 15:17:54,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=772440.6666666666, ans=0.125
2024-09-25 15:18:01,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=772440.6666666666, ans=0.2
2024-09-25 15:18:04,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=772440.6666666666, ans=0.1
2024-09-25 15:18:10,615 INFO [train.py:1198] (1/4) Epoch 43, batch 1900, loss[loss=0.1421, ctc_loss=0.08899, cr_loss=0.2656, over 16713.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1203, cr_loss=0.337, over 3353395.65 frames. ], batch size: 37, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:18:15,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=772487.3333333334, ans=0.0
2024-09-25 15:18:16,262 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.11 vs. limit=15.0
2024-09-25 15:18:25,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=772534.0, ans=0.0
2024-09-25 15:18:36,192 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.296e+02 1.386e+02 1.467e+02 1.936e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 15:18:58,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=772627.3333333334, ans=0.0
2024-09-25 15:19:25,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=772674.0, ans=0.1
2024-09-25 15:19:35,529 INFO [train.py:1198] (1/4) Epoch 43, batch 1950, loss[loss=0.1729, ctc_loss=0.1086, cr_loss=0.3213, over 17036.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1209, cr_loss=0.3381, over 3354971.37 frames. ], batch size: 53, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:19:46,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=772720.6666666666, ans=0.125
2024-09-25 15:20:11,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=772814.0, ans=0.0
2024-09-25 15:20:19,113 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=772814.0, ans=0.2
2024-09-25 15:20:50,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=772907.3333333334, ans=0.0
2024-09-25 15:21:00,968 INFO [train.py:1198] (1/4) Epoch 43, batch 2000, loss[loss=0.1998, ctc_loss=0.1268, cr_loss=0.3652, over 16997.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1205, cr_loss=0.3372, over 3359876.24 frames. ], batch size: 53, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:21:15,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=773000.6666666666, ans=0.125
2024-09-25 15:21:24,394 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=20.12 vs. limit=22.5
2024-09-25 15:21:26,607 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.269e+02 1.346e+02 1.449e+02 2.025e+02, threshold=2.692e+02, percent-clipped=0.0
2024-09-25 15:21:27,434 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.94 vs. limit=15.0
2024-09-25 15:21:33,618 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.53 vs. limit=15.0
2024-09-25 15:21:38,479 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.35 vs. limit=15.0
2024-09-25 15:21:54,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=773094.0, ans=0.125
2024-09-25 15:22:17,134 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.01 vs. limit=15.0
2024-09-25 15:22:20,791 INFO [train.py:1198] (1/4) Epoch 43, batch 2050, loss[loss=0.1761, ctc_loss=0.1116, cr_loss=0.3228, over 16960.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3383, over 3361745.20 frames. ], batch size: 42, lr: 2.76e-03, grad_scale: 32.0
2024-09-25 15:22:48,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=773234.0, ans=0.125
2024-09-25 15:23:01,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=773280.6666666666, ans=0.0
2024-09-25 15:23:36,495 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=773374.0, ans=0.1
2024-09-25 15:23:40,844 INFO [train.py:1198] (1/4) Epoch 43, batch 2100, loss[loss=0.2193, ctc_loss=0.148, cr_loss=0.3566, over 15111.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1203, cr_loss=0.3363, over 3361376.91 frames. ], batch size: 88, lr: 2.76e-03, grad_scale: 16.0
2024-09-25 15:23:55,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=773467.3333333334, ans=0.1
2024-09-25 15:23:58,886 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 15:24:10,504 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.257e+02 1.347e+02 1.435e+02 1.926e+02, threshold=2.695e+02, percent-clipped=0.0
2024-09-25 15:24:25,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=773514.0, ans=0.125
2024-09-25 15:24:45,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=773560.6666666666, ans=0.2
2024-09-25 15:25:05,722 INFO [train.py:1198] (1/4) Epoch 43, batch 2150, loss[loss=0.1634, ctc_loss=0.1029, cr_loss=0.3025, over 17215.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1203, cr_loss=0.3372, over 3367019.36 frames. ], batch size: 41, lr: 2.76e-03, grad_scale: 16.0
2024-09-25 15:25:18,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=773654.0, ans=0.2
2024-09-25 15:25:31,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=773700.6666666666, ans=0.2
2024-09-25 15:26:08,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=773794.0, ans=0.125
2024-09-25 15:26:20,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=773840.6666666666, ans=0.125
2024-09-25 15:26:31,341 INFO [train.py:1198] (1/4) Epoch 43, batch 2200, loss[loss=0.1538, ctc_loss=0.09652, cr_loss=0.2864, over 17058.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1196, cr_loss=0.3361, over 3361250.45 frames. ], batch size: 39, lr: 2.76e-03, grad_scale: 16.0
2024-09-25 15:26:37,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=773887.3333333334, ans=0.1
2024-09-25 15:26:50,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=773934.0, ans=0.0
2024-09-25 15:26:58,330 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.324e+02 1.413e+02 1.558e+02 2.550e+02, threshold=2.826e+02, percent-clipped=0.0
2024-09-25 15:27:24,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=774027.3333333334, ans=0.2
2024-09-25 15:27:50,922 INFO [train.py:1198] (1/4) Epoch 43, batch 2250, loss[loss=0.1872, ctc_loss=0.1217, cr_loss=0.3278, over 17307.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1202, cr_loss=0.3371, over 3365917.19 frames. ], batch size: 49, lr: 2.76e-03, grad_scale: 16.0
2024-09-25 15:28:00,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=774120.6666666666, ans=0.2
2024-09-25 15:28:23,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=774214.0, ans=0.0
2024-09-25 15:28:50,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=774260.6666666666, ans=0.2
2024-09-25 15:29:13,565 INFO [train.py:1198] (1/4) Epoch 43, batch 2300, loss[loss=0.194, ctc_loss=0.1263, cr_loss=0.3383, over 16909.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3388, over 3363248.66 frames. ], batch size: 58, lr: 2.76e-03, grad_scale: 16.0
2024-09-25 15:29:41,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=774400.6666666666, ans=0.125
2024-09-25 15:29:42,935 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.301e+02 1.372e+02 1.481e+02 2.158e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-25 15:29:54,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=774447.3333333334, ans=0.0
2024-09-25 15:30:02,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=774494.0, ans=0.2
2024-09-25 15:30:07,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.out_whiten.whitening_limit, batch_count=774494.0, ans=8.0
2024-09-25 15:30:11,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=774494.0, ans=0.0
2024-09-25 15:30:17,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=774540.6666666666, ans=0.025
2024-09-25 15:30:37,780 INFO [train.py:1198] (1/4) Epoch 43, batch 2350, loss[loss=0.1682, ctc_loss=0.1077, cr_loss=0.3024, over 17075.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1209, cr_loss=0.338, over 3365957.15 frames. ], batch size: 43, lr: 2.76e-03, grad_scale: 16.0
2024-09-25 15:30:54,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=774634.0, ans=0.0
2024-09-25 15:30:56,643 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=774634.0, ans=0.1
2024-09-25 15:31:22,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=774680.6666666666, ans=0.0
2024-09-25 15:31:44,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=774774.0, ans=0.025
2024-09-25 15:31:47,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=774774.0, ans=0.025
2024-09-25 15:31:53,168 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.43 vs. limit=22.5
2024-09-25 15:32:00,369 INFO [train.py:1198] (1/4) Epoch 43, batch 2400, loss[loss=0.1932, ctc_loss=0.1253, cr_loss=0.3397, over 17136.00 frames. ], tot_loss[loss=0.188, ctc_loss=0.1205, cr_loss=0.3376, over 3372177.45 frames. ], batch size: 48, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:32:10,344 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 15:32:10,766 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=7.66 vs. limit=15.0
2024-09-25 15:32:27,693 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.292e+02 1.363e+02 1.432e+02 2.291e+02, threshold=2.726e+02, percent-clipped=0.0
2024-09-25 15:32:42,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=774914.0, ans=0.09899494936611666
2024-09-25 15:32:57,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=774960.6666666666, ans=0.125
2024-09-25 15:32:58,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=774960.6666666666, ans=0.125
2024-09-25 15:33:06,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=775007.3333333334, ans=0.07
2024-09-25 15:33:08,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=775007.3333333334, ans=0.125
2024-09-25 15:33:20,860 INFO [train.py:1198] (1/4) Epoch 43, batch 2450, loss[loss=0.1871, ctc_loss=0.1183, cr_loss=0.3438, over 17292.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1215, cr_loss=0.3388, over 3363570.98 frames. ], batch size: 49, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:33:25,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=775054.0, ans=0.125
2024-09-25 15:33:37,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=775100.6666666666, ans=0.07
2024-09-25 15:33:46,339 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=775100.6666666666, ans=0.2
2024-09-25 15:33:46,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=775100.6666666666, ans=0.0
2024-09-25 15:34:12,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=775194.0, ans=0.025
2024-09-25 15:34:14,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=775194.0, ans=0.0
2024-09-25 15:34:14,580 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=775194.0, ans=0.1
2024-09-25 15:34:17,727 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=775194.0, ans=0.1
2024-09-25 15:34:31,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=775240.6666666666, ans=0.0
2024-09-25 15:34:45,575 INFO [train.py:1198] (1/4) Epoch 43, batch 2500, loss[loss=0.207, ctc_loss=0.137, cr_loss=0.35, over 17005.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1215, cr_loss=0.3388, over 3361316.52 frames. ], batch size: 53, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:34:49,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=775287.3333333334, ans=0.125
2024-09-25 15:35:02,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=775334.0, ans=0.125
2024-09-25 15:35:13,016 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.284e+02 1.374e+02 1.483e+02 2.994e+02, threshold=2.747e+02, percent-clipped=1.0
2024-09-25 15:35:14,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=775334.0, ans=0.0
2024-09-25 15:35:15,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=775334.0, ans=0.125
2024-09-25 15:35:19,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=775380.6666666666, ans=0.125
2024-09-25 15:35:32,876 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=7.74 vs. limit=15.0
2024-09-25 15:35:35,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=775427.3333333334, ans=0.0
2024-09-25 15:35:35,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=775427.3333333334, ans=0.125
2024-09-25 15:35:54,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=775474.0, ans=0.125
2024-09-25 15:35:59,417 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.15 vs. limit=10.0
2024-09-25 15:36:02,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=775474.0, ans=0.125
2024-09-25 15:36:02,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=775474.0, ans=0.035
2024-09-25 15:36:10,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=775520.6666666666, ans=0.125
2024-09-25 15:36:11,634 INFO [train.py:1198] (1/4) Epoch 43, batch 2550, loss[loss=0.2029, ctc_loss=0.1314, cr_loss=0.3571, over 17007.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1212, cr_loss=0.3382, over 3355985.59 frames. ], batch size: 56, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:36:23,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=775520.6666666666, ans=0.0
2024-09-25 15:36:26,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=775567.3333333334, ans=0.125
2024-09-25 15:36:47,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=775614.0, ans=0.2
2024-09-25 15:37:15,654 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.75 vs. limit=12.0
2024-09-25 15:37:18,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=775707.3333333334, ans=0.0
2024-09-25 15:37:32,113 INFO [train.py:1198] (1/4) Epoch 43, batch 2600, loss[loss=0.1793, ctc_loss=0.1141, cr_loss=0.3261, over 17352.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1205, cr_loss=0.3369, over 3359776.44 frames. ], batch size: 48, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:37:45,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=775754.0, ans=0.09899494936611666
2024-09-25 15:37:49,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=775800.6666666666, ans=0.1
2024-09-25 15:37:58,961 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.271e+02 1.387e+02 1.465e+02 2.023e+02, threshold=2.774e+02, percent-clipped=0.0
2024-09-25 15:38:39,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=775940.6666666666, ans=0.125
2024-09-25 15:38:51,970 INFO [train.py:1198] (1/4) Epoch 43, batch 2650, loss[loss=0.1563, ctc_loss=0.09584, cr_loss=0.3022, over 17278.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1211, cr_loss=0.3382, over 3367049.89 frames. ], batch size: 42, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:39:32,950 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=776080.6666666666, ans=0.0
2024-09-25 15:39:39,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=776080.6666666666, ans=0.125
2024-09-25 15:39:59,733 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 15:40:13,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=776174.0, ans=0.125
2024-09-25 15:40:16,626 INFO [train.py:1198] (1/4) Epoch 43, batch 2700, loss[loss=0.1629, ctc_loss=0.1029, cr_loss=0.3002, over 17105.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1217, cr_loss=0.3397, over 3369308.51 frames. ], batch size: 40, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:40:26,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=776220.6666666666, ans=0.0
2024-09-25 15:40:26,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=776220.6666666666, ans=0.0
2024-09-25 15:40:37,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=776267.3333333334, ans=0.04949747468305833
2024-09-25 15:40:49,149 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.311e+02 1.386e+02 1.489e+02 1.705e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 15:40:51,593 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.12 vs. limit=15.0
2024-09-25 15:41:09,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=776360.6666666666, ans=0.125
2024-09-25 15:41:21,894 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.48 vs. limit=15.0
2024-09-25 15:41:41,505 INFO [train.py:1198] (1/4) Epoch 43, batch 2750, loss[loss=0.1965, ctc_loss=0.1255, cr_loss=0.3551, over 17018.00 frames. ], tot_loss[loss=0.1905, ctc_loss=0.1223, cr_loss=0.3408, over 3356714.68 frames. ], batch size: 51, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:41:41,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=776454.0, ans=0.2
2024-09-25 15:41:49,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=776454.0, ans=0.0
2024-09-25 15:42:22,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=776547.3333333334, ans=0.125
2024-09-25 15:42:39,961 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=776594.0, ans=0.0
2024-09-25 15:42:40,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=776594.0, ans=0.1
2024-09-25 15:42:56,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=776640.6666666666, ans=0.125
2024-09-25 15:43:02,156 INFO [train.py:1198] (1/4) Epoch 43, batch 2800, loss[loss=0.206, ctc_loss=0.1363, cr_loss=0.3487, over 17292.00 frames. ], tot_loss[loss=0.1894, ctc_loss=0.1216, cr_loss=0.3389, over 3354238.15 frames. ], batch size: 46, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:43:13,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=776687.3333333334, ans=0.125
2024-09-25 15:43:24,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=776734.0, ans=0.125
2024-09-25 15:43:29,176 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.317e+02 1.406e+02 1.487e+02 1.823e+02, threshold=2.813e+02, percent-clipped=0.0
2024-09-25 15:43:31,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=776734.0, ans=0.0
2024-09-25 15:43:34,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=776780.6666666666, ans=0.0
2024-09-25 15:43:38,009 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.75 vs. limit=6.0
2024-09-25 15:44:02,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=776827.3333333334, ans=0.0
2024-09-25 15:44:12,364 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=776874.0, ans=0.125
2024-09-25 15:44:18,465 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.09 vs. limit=22.5
2024-09-25 15:44:27,271 INFO [train.py:1198] (1/4) Epoch 43, batch 2850, loss[loss=0.2028, ctc_loss=0.1379, cr_loss=0.3242, over 11763.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1208, cr_loss=0.3372, over 3350496.31 frames. ], batch size: 123, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:44:32,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=776920.6666666666, ans=0.125
2024-09-25 15:44:38,614 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=776920.6666666666, ans=0.125
2024-09-25 15:44:50,449 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.53 vs. limit=15.0
2024-09-25 15:45:01,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=3.82 vs. limit=12.0
2024-09-25 15:45:09,151 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 15:45:27,449 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=777060.6666666666, ans=0.0
2024-09-25 15:45:33,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=777107.3333333334, ans=0.125
2024-09-25 15:45:46,885 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.05 vs. limit=12.0
2024-09-25 15:45:52,191 INFO [train.py:1198] (1/4) Epoch 43, batch 2900, loss[loss=0.1992, ctc_loss=0.1291, cr_loss=0.351, over 17015.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1214, cr_loss=0.3379, over 3338270.00 frames. ], batch size: 56, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:46:19,411 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.318e+02 1.398e+02 1.451e+02 2.249e+02, threshold=2.795e+02, percent-clipped=0.0
2024-09-25 15:46:21,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=777200.6666666666, ans=0.125
2024-09-25 15:47:10,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.16 vs. limit=15.0
2024-09-25 15:47:12,742 INFO [train.py:1198] (1/4) Epoch 43, batch 2950, loss[loss=0.1936, ctc_loss=0.122, cr_loss=0.3578, over 16980.00 frames. ], tot_loss[loss=0.188, ctc_loss=0.1207, cr_loss=0.3368, over 3344717.75 frames. ], batch size: 53, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:47:31,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.85 vs. limit=10.0
2024-09-25 15:47:38,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=777434.0, ans=0.125
2024-09-25 15:47:48,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=777480.6666666666, ans=0.125
2024-09-25 15:47:51,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=777480.6666666666, ans=0.0
2024-09-25 15:48:04,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=777527.3333333334, ans=0.1
2024-09-25 15:48:09,319 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 15:48:18,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=777574.0, ans=0.2
2024-09-25 15:48:22,680 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.43 vs. limit=15.0
2024-09-25 15:48:33,057 INFO [train.py:1198] (1/4) Epoch 43, batch 3000, loss[loss=0.2294, ctc_loss=0.152, cr_loss=0.3871, over 15020.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.121, cr_loss=0.3376, over 3351109.35 frames. ], batch size: 90, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:48:33,057 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 15:48:50,096 INFO [train.py:1230] (1/4) Epoch 43, validation: loss=0.03539, ctc_loss=0.03539, cr_loss=1.015e-14, over 944034.00 frames. 
2024-09-25 15:48:50,097 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 15:48:59,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=777620.6666666666, ans=0.125
2024-09-25 15:49:16,945 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.300e+02 1.373e+02 1.452e+02 1.992e+02, threshold=2.746e+02, percent-clipped=0.0
2024-09-25 15:49:41,521 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=777760.6666666666, ans=0.125
2024-09-25 15:50:01,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=777807.3333333334, ans=0.125
2024-09-25 15:50:10,657 INFO [train.py:1198] (1/4) Epoch 43, batch 3050, loss[loss=0.142, ctc_loss=0.08921, cr_loss=0.2637, over 17264.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1213, cr_loss=0.3383, over 3351015.95 frames. ], batch size: 42, lr: 2.75e-03, grad_scale: 16.0
2024-09-25 15:50:15,653 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 15:50:17,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=777854.0, ans=0.0
2024-09-25 15:50:51,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=777947.3333333334, ans=0.2
2024-09-25 15:51:03,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=777994.0, ans=0.1
2024-09-25 15:51:20,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=778040.6666666666, ans=0.2
2024-09-25 15:51:28,334 INFO [train.py:1198] (1/4) Epoch 43, batch 3100, loss[loss=0.1584, ctc_loss=0.1011, cr_loss=0.2864, over 16294.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.121, cr_loss=0.3377, over 3343012.02 frames. ], batch size: 36, lr: 2.75e-03, grad_scale: 16.0
2024-09-25 15:51:52,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=778134.0, ans=0.125
2024-09-25 15:51:58,668 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.269e+02 1.353e+02 1.457e+02 2.895e+02, threshold=2.707e+02, percent-clipped=1.0
2024-09-25 15:52:18,006 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=778227.3333333334, ans=0.1
2024-09-25 15:52:30,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=778227.3333333334, ans=0.1
2024-09-25 15:52:51,552 INFO [train.py:1198] (1/4) Epoch 43, batch 3150, loss[loss=0.2069, ctc_loss=0.1319, cr_loss=0.3749, over 17040.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.121, cr_loss=0.3383, over 3350991.64 frames. ], batch size: 56, lr: 2.75e-03, grad_scale: 16.0
2024-09-25 15:52:51,904 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=778320.6666666666, ans=0.1
2024-09-25 15:54:10,165 INFO [train.py:1198] (1/4) Epoch 43, batch 3200, loss[loss=0.2024, ctc_loss=0.1264, cr_loss=0.3799, over 17096.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1207, cr_loss=0.3373, over 3360536.43 frames. ], batch size: 49, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:54:10,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=778554.0, ans=0.125
2024-09-25 15:54:20,272 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=5.22 vs. limit=12.0
2024-09-25 15:54:21,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=778554.0, ans=0.025
2024-09-25 15:54:23,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=778554.0, ans=0.04949747468305833
2024-09-25 15:54:34,505 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.04 vs. limit=15.0
2024-09-25 15:54:38,168 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.301e+02 1.369e+02 1.499e+02 2.028e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-25 15:54:41,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=778647.3333333334, ans=0.2
2024-09-25 15:54:43,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=778647.3333333334, ans=0.1
2024-09-25 15:54:46,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=778647.3333333334, ans=0.125
2024-09-25 15:54:59,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=778694.0, ans=0.125
2024-09-25 15:55:28,972 INFO [train.py:1198] (1/4) Epoch 43, batch 3250, loss[loss=0.1993, ctc_loss=0.1281, cr_loss=0.3561, over 16988.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1213, cr_loss=0.339, over 3357697.00 frames. ], batch size: 53, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:56:17,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=778927.3333333334, ans=0.125
2024-09-25 15:56:40,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=778974.0, ans=0.0
2024-09-25 15:56:46,769 INFO [train.py:1198] (1/4) Epoch 43, batch 3300, loss[loss=0.2037, ctc_loss=0.1321, cr_loss=0.3581, over 16928.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1218, cr_loss=0.3397, over 3349422.11 frames. ], batch size: 58, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:57:14,795 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.314e+02 1.383e+02 1.471e+02 2.059e+02, threshold=2.766e+02, percent-clipped=0.0
2024-09-25 15:58:04,605 INFO [train.py:1198] (1/4) Epoch 43, batch 3350, loss[loss=0.2377, ctc_loss=0.1558, cr_loss=0.4099, over 15127.00 frames. ], tot_loss[loss=0.1907, ctc_loss=0.1225, cr_loss=0.3411, over 3355350.64 frames. ], batch size: 89, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:58:26,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=779300.6666666666, ans=0.125
2024-09-25 15:58:29,150 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn2.whiten, num_groups=1, num_channels=192, metric=11.95 vs. limit=22.5
2024-09-25 15:58:30,428 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.66 vs. limit=10.0
2024-09-25 15:58:39,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=779347.3333333334, ans=0.0
2024-09-25 15:59:04,991 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=779394.0, ans=0.1
2024-09-25 15:59:23,220 INFO [train.py:1198] (1/4) Epoch 43, batch 3400, loss[loss=0.1858, ctc_loss=0.1166, cr_loss=0.3458, over 17297.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1223, cr_loss=0.3403, over 3334969.67 frames. ], batch size: 46, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 15:59:27,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=779487.3333333334, ans=0.1
2024-09-25 15:59:32,847 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=779487.3333333334, ans=0.1
2024-09-25 15:59:55,330 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.291e+02 1.360e+02 1.462e+02 4.308e+02, threshold=2.720e+02, percent-clipped=1.0
2024-09-25 16:00:25,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=779627.3333333334, ans=0.125
2024-09-25 16:00:37,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=779674.0, ans=0.0
2024-09-25 16:00:45,026 INFO [train.py:1198] (1/4) Epoch 43, batch 3450, loss[loss=0.1944, ctc_loss=0.1252, cr_loss=0.3459, over 16954.00 frames. ], tot_loss[loss=0.1898, ctc_loss=0.1219, cr_loss=0.3394, over 3346199.52 frames. ], batch size: 42, lr: 2.75e-03, grad_scale: 32.0
2024-09-25 16:00:50,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=779720.6666666666, ans=0.125
2024-09-25 16:00:53,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=779720.6666666666, ans=0.0
2024-09-25 16:00:56,297 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=779720.6666666666, ans=0.0
2024-09-25 16:00:56,684 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.70 vs. limit=15.0
2024-09-25 16:00:59,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=779767.3333333334, ans=0.025
2024-09-25 16:01:15,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=779814.0, ans=0.125
2024-09-25 16:02:03,507 INFO [train.py:1198] (1/4) Epoch 43, batch 3500, loss[loss=0.1898, ctc_loss=0.1229, cr_loss=0.3346, over 17212.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1217, cr_loss=0.339, over 3353630.93 frames. ], batch size: 47, lr: 2.75e-03, grad_scale: 16.0
2024-09-25 16:02:13,655 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.23 vs. limit=15.0
2024-09-25 16:02:35,356 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.311e+02 1.389e+02 1.472e+02 1.940e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-25 16:02:43,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=780047.3333333334, ans=0.0
2024-09-25 16:03:26,021 INFO [train.py:1198] (1/4) Epoch 43, batch 3550, loss[loss=0.1712, ctc_loss=0.11, cr_loss=0.3058, over 16966.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1212, cr_loss=0.3381, over 3355652.31 frames. ], batch size: 42, lr: 2.75e-03, grad_scale: 16.0
2024-09-25 16:03:47,214 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.40 vs. limit=15.0
2024-09-25 16:03:48,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=780234.0, ans=0.125
2024-09-25 16:03:48,378 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=780234.0, ans=0.125
2024-09-25 16:04:23,811 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.55 vs. limit=15.0
2024-09-25 16:04:44,773 INFO [train.py:1198] (1/4) Epoch 43, batch 3600, loss[loss=0.1657, ctc_loss=0.1051, cr_loss=0.303, over 17286.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.121, cr_loss=0.3386, over 3365570.01 frames. ], batch size: 42, lr: 2.74e-03, grad_scale: 32.0
2024-09-25 16:04:49,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=780420.6666666666, ans=0.1
2024-09-25 16:04:51,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=780420.6666666666, ans=0.1
2024-09-25 16:04:59,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=780467.3333333334, ans=0.125
2024-09-25 16:05:07,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=780467.3333333334, ans=0.1
2024-09-25 16:05:14,401 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.305e+02 1.407e+02 1.507e+02 1.948e+02, threshold=2.813e+02, percent-clipped=0.0
2024-09-25 16:05:53,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=780607.3333333334, ans=0.0
2024-09-25 16:06:01,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=780654.0, ans=0.025
2024-09-25 16:06:03,264 INFO [train.py:1198] (1/4) Epoch 43, batch 3650, loss[loss=0.187, ctc_loss=0.1218, cr_loss=0.326, over 17290.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1211, cr_loss=0.3382, over 3342794.07 frames. ], batch size: 49, lr: 2.74e-03, grad_scale: 32.0
2024-09-25 16:06:11,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=780654.0, ans=0.125
2024-09-25 16:06:26,166 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.48 vs. limit=15.0
2024-09-25 16:06:31,933 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=780700.6666666666, ans=0.1
2024-09-25 16:06:42,837 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 16:06:44,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=780747.3333333334, ans=0.125
2024-09-25 16:06:54,830 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.97 vs. limit=10.0
2024-09-25 16:07:11,568 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.97 vs. limit=15.0
2024-09-25 16:07:21,990 INFO [train.py:1198] (1/4) Epoch 43, batch 3700, loss[loss=0.1552, ctc_loss=0.102, cr_loss=0.2661, over 16230.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1218, cr_loss=0.3394, over 3335115.13 frames. ], batch size: 36, lr: 2.74e-03, grad_scale: 32.0
2024-09-25 16:07:28,525 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.81 vs. limit=10.0
2024-09-25 16:07:34,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=780887.3333333334, ans=0.125
2024-09-25 16:07:45,483 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.93 vs. limit=15.0
2024-09-25 16:07:52,633 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.295e+02 1.384e+02 1.500e+02 2.354e+02, threshold=2.769e+02, percent-clipped=0.0
2024-09-25 16:07:56,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=780980.6666666666, ans=0.1
2024-09-25 16:08:03,767 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 16:08:14,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=781027.3333333334, ans=0.125
2024-09-25 16:08:41,139 INFO [train.py:1198] (1/4) Epoch 43, batch 3750, loss[loss=0.2117, ctc_loss=0.1355, cr_loss=0.3812, over 17032.00 frames. ], tot_loss[loss=0.1915, ctc_loss=0.1232, cr_loss=0.3417, over 3327877.88 frames. ], batch size: 52, lr: 2.74e-03, grad_scale: 16.0
2024-09-25 16:08:41,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=781120.6666666666, ans=0.0
2024-09-25 16:08:44,855 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.09 vs. limit=12.0
2024-09-25 16:09:06,633 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=12.01 vs. limit=12.0
2024-09-25 16:09:51,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=781307.3333333334, ans=0.2
2024-09-25 16:10:01,216 INFO [train.py:1198] (1/4) Epoch 43, batch 3800, loss[loss=0.2, ctc_loss=0.1287, cr_loss=0.3564, over 16905.00 frames. ], tot_loss[loss=0.1917, ctc_loss=0.1234, cr_loss=0.3416, over 3310960.18 frames. ], batch size: 58, lr: 2.74e-03, grad_scale: 16.0
2024-09-25 16:10:10,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=781354.0, ans=0.125
2024-09-25 16:10:11,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=781354.0, ans=0.2
2024-09-25 16:10:22,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.80 vs. limit=22.5
2024-09-25 16:10:24,180 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.51 vs. limit=15.0
2024-09-25 16:10:28,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=781400.6666666666, ans=0.125
2024-09-25 16:10:32,811 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.355e+02 1.451e+02 1.569e+02 1.796e+02, threshold=2.903e+02, percent-clipped=0.0
2024-09-25 16:10:33,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=781447.3333333334, ans=0.125
2024-09-25 16:10:36,631 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.50 vs. limit=15.0
2024-09-25 16:11:04,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=781540.6666666666, ans=0.0
2024-09-25 16:11:08,177 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=781540.6666666666, ans=0.0
2024-09-25 16:11:17,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=781540.6666666666, ans=0.125
2024-09-25 16:11:20,436 INFO [train.py:1198] (1/4) Epoch 43, batch 3850, loss[loss=0.189, ctc_loss=0.1256, cr_loss=0.317, over 12253.00 frames. ], tot_loss[loss=0.1927, ctc_loss=0.1242, cr_loss=0.3421, over 3273554.02 frames. ], batch size: 124, lr: 2.74e-03, grad_scale: 16.0
2024-09-25 16:11:30,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=781587.3333333334, ans=0.1
2024-09-25 16:11:58,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=781680.6666666666, ans=0.125
2024-09-25 16:12:25,487 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.40 vs. limit=10.0
2024-09-25 16:12:26,902 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=781774.0, ans=0.0
2024-09-25 16:13:18,296 INFO [train.py:1198] (1/4) Epoch 44, batch 0, loss[loss=0.2158, ctc_loss=0.1403, cr_loss=0.3774, over 17004.00 frames. ], tot_loss[loss=0.2158, ctc_loss=0.1403, cr_loss=0.3774, over 17004.00 frames. ], batch size: 52, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:13:18,297 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 16:13:33,583 INFO [train.py:1230] (1/4) Epoch 44, validation: loss=0.03507, ctc_loss=0.03507, cr_loss=1.053e-14, over 944034.00 frames. 
2024-09-25 16:13:33,583 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 16:13:43,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=781802.0, ans=0.2
2024-09-25 16:13:56,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=781848.6666666666, ans=0.0
2024-09-25 16:14:14,723 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.434e+02 1.591e+02 1.721e+02 2.734e+02, threshold=3.183e+02, percent-clipped=0.0
2024-09-25 16:14:15,576 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.14 vs. limit=22.5
2024-09-25 16:14:43,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=781988.6666666666, ans=0.05
2024-09-25 16:14:44,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=781988.6666666666, ans=0.0
2024-09-25 16:14:49,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=781988.6666666666, ans=0.1
2024-09-25 16:14:56,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=781988.6666666666, ans=0.0
2024-09-25 16:14:59,046 INFO [train.py:1198] (1/4) Epoch 44, batch 50, loss[loss=0.218, ctc_loss=0.142, cr_loss=0.3803, over 17209.00 frames. ], tot_loss[loss=0.1902, ctc_loss=0.1221, cr_loss=0.3406, over 764489.83 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:15:02,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=782035.3333333334, ans=0.2
2024-09-25 16:15:15,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=782082.0, ans=0.0
2024-09-25 16:15:20,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=782082.0, ans=0.1
2024-09-25 16:15:33,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=782128.6666666666, ans=0.025
2024-09-25 16:15:36,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=782128.6666666666, ans=0.125
2024-09-25 16:15:37,758 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=782128.6666666666, ans=0.0
2024-09-25 16:15:53,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=782175.3333333334, ans=0.1
2024-09-25 16:15:55,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=782175.3333333334, ans=10.0
2024-09-25 16:15:57,612 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.83 vs. limit=15.0
2024-09-25 16:16:12,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=782222.0, ans=0.1
2024-09-25 16:16:18,758 INFO [train.py:1198] (1/4) Epoch 44, batch 100, loss[loss=0.1649, ctc_loss=0.1032, cr_loss=0.3086, over 17066.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1205, cr_loss=0.3369, over 1333257.49 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:16:20,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=782268.6666666666, ans=0.1
2024-09-25 16:16:35,619 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=7.50 vs. limit=15.0
2024-09-25 16:16:44,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=782315.3333333334, ans=0.125
2024-09-25 16:16:47,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=782315.3333333334, ans=0.125
2024-09-25 16:16:49,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=782362.0, ans=0.1
2024-09-25 16:17:00,126 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.309e+02 1.396e+02 1.536e+02 2.062e+02, threshold=2.792e+02, percent-clipped=0.0
2024-09-25 16:17:03,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=782362.0, ans=0.125
2024-09-25 16:17:26,932 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.19 vs. limit=10.0
2024-09-25 16:17:35,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=782455.3333333334, ans=0.125
2024-09-25 16:17:41,586 INFO [train.py:1198] (1/4) Epoch 44, batch 150, loss[loss=0.1975, ctc_loss=0.1259, cr_loss=0.3579, over 17155.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.1201, cr_loss=0.3358, over 1789727.65 frames. ], batch size: 45, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:17:53,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=782502.0, ans=0.2
2024-09-25 16:18:21,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=782595.3333333334, ans=0.125
2024-09-25 16:18:32,771 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 16:18:51,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=782688.6666666666, ans=0.1
2024-09-25 16:19:07,046 INFO [train.py:1198] (1/4) Epoch 44, batch 200, loss[loss=0.1934, ctc_loss=0.1225, cr_loss=0.3545, over 17287.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1218, cr_loss=0.3387, over 2123942.34 frames. ], batch size: 51, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:19:15,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=782735.3333333334, ans=0.125
2024-09-25 16:19:23,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=782782.0, ans=0.2
2024-09-25 16:19:30,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=782782.0, ans=0.025
2024-09-25 16:19:48,491 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.333e+02 1.408e+02 1.534e+02 2.430e+02, threshold=2.816e+02, percent-clipped=0.0
2024-09-25 16:20:09,201 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.81 vs. limit=6.0
2024-09-25 16:20:11,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=782875.3333333334, ans=0.2
2024-09-25 16:20:18,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=782922.0, ans=0.0
2024-09-25 16:20:30,830 INFO [train.py:1198] (1/4) Epoch 44, batch 250, loss[loss=0.184, ctc_loss=0.1151, cr_loss=0.3443, over 17044.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1208, cr_loss=0.3372, over 2392990.74 frames. ], batch size: 51, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:20:31,703 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=16.80 vs. limit=22.5
2024-09-25 16:20:45,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=783015.3333333334, ans=0.125
2024-09-25 16:20:47,306 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.03 vs. limit=12.0
2024-09-25 16:20:56,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=783015.3333333334, ans=0.125
2024-09-25 16:21:47,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=783155.3333333334, ans=0.125
2024-09-25 16:21:54,065 INFO [train.py:1198] (1/4) Epoch 44, batch 300, loss[loss=0.1693, ctc_loss=0.1067, cr_loss=0.3129, over 17062.00 frames. ], tot_loss[loss=0.1896, ctc_loss=0.1218, cr_loss=0.3392, over 2600575.33 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:22:02,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=783202.0, ans=0.2
2024-09-25 16:22:03,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=783202.0, ans=0.2
2024-09-25 16:22:12,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=783248.6666666666, ans=0.125
2024-09-25 16:22:19,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=783248.6666666666, ans=0.125
2024-09-25 16:22:32,113 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.326e+02 1.428e+02 1.549e+02 2.699e+02, threshold=2.856e+02, percent-clipped=0.0
2024-09-25 16:22:42,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=783342.0, ans=0.0
2024-09-25 16:22:54,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=783342.0, ans=0.1
2024-09-25 16:23:08,009 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.62 vs. limit=6.0
2024-09-25 16:23:16,539 INFO [train.py:1198] (1/4) Epoch 44, batch 350, loss[loss=0.1813, ctc_loss=0.111, cr_loss=0.3515, over 17256.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1211, cr_loss=0.3381, over 2765324.54 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:23:37,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=783482.0, ans=0.0
2024-09-25 16:24:07,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=783575.3333333334, ans=0.07
2024-09-25 16:24:22,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=783622.0, ans=0.125
2024-09-25 16:24:42,413 INFO [train.py:1198] (1/4) Epoch 44, batch 400, loss[loss=0.2099, ctc_loss=0.1379, cr_loss=0.36, over 17042.00 frames. ], tot_loss[loss=0.1901, ctc_loss=0.1221, cr_loss=0.3398, over 2891961.24 frames. ], batch size: 52, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:25:20,975 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.320e+02 1.420e+02 1.550e+02 2.069e+02, threshold=2.840e+02, percent-clipped=0.0
2024-09-25 16:25:50,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=783855.3333333334, ans=0.025
2024-09-25 16:25:51,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=783855.3333333334, ans=0.025
2024-09-25 16:26:02,567 INFO [train.py:1198] (1/4) Epoch 44, batch 450, loss[loss=0.1491, ctc_loss=0.09441, cr_loss=0.2733, over 16733.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1214, cr_loss=0.3393, over 2994109.92 frames. ], batch size: 37, lr: 2.71e-03, grad_scale: 32.0
2024-09-25 16:26:12,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=783902.0, ans=0.125
2024-09-25 16:26:14,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=783902.0, ans=0.1
2024-09-25 16:26:25,963 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.07 vs. limit=15.0
2024-09-25 16:26:56,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=784042.0, ans=0.2
2024-09-25 16:26:57,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=784042.0, ans=0.125
2024-09-25 16:27:28,172 INFO [train.py:1198] (1/4) Epoch 44, batch 500, loss[loss=0.2098, ctc_loss=0.1357, cr_loss=0.3706, over 17035.00 frames. ], tot_loss[loss=0.1886, ctc_loss=0.121, cr_loss=0.3378, over 3079366.46 frames. ], batch size: 52, lr: 2.71e-03, grad_scale: 16.0
2024-09-25 16:28:10,662 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.312e+02 1.366e+02 1.455e+02 1.775e+02, threshold=2.732e+02, percent-clipped=0.0
2024-09-25 16:28:12,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=784228.6666666666, ans=0.2
2024-09-25 16:28:32,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten.whitening_limit, batch_count=784275.3333333334, ans=15.0
2024-09-25 16:28:50,233 INFO [train.py:1198] (1/4) Epoch 44, batch 550, loss[loss=0.2023, ctc_loss=0.1313, cr_loss=0.3547, over 17100.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1204, cr_loss=0.3367, over 3144139.28 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 16.0
2024-09-25 16:28:50,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=784368.6666666666, ans=0.1
2024-09-25 16:28:56,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=784368.6666666666, ans=0.1
2024-09-25 16:29:13,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=784415.3333333334, ans=0.125
2024-09-25 16:29:25,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=784462.0, ans=0.125
2024-09-25 16:30:15,130 INFO [train.py:1198] (1/4) Epoch 44, batch 600, loss[loss=0.2412, ctc_loss=0.1607, cr_loss=0.4024, over 15081.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1213, cr_loss=0.3382, over 3193082.00 frames. ], batch size: 89, lr: 2.71e-03, grad_scale: 16.0
2024-09-25 16:30:23,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=784602.0, ans=0.0
2024-09-25 16:30:25,159 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=784602.0, ans=0.0
2024-09-25 16:30:28,509 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 16:30:49,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=784695.3333333334, ans=0.125
2024-09-25 16:30:55,328 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.308e+02 1.386e+02 1.485e+02 2.110e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 16:31:08,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=784742.0, ans=0.125
2024-09-25 16:31:24,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=784788.6666666666, ans=0.125
2024-09-25 16:31:35,712 INFO [train.py:1198] (1/4) Epoch 44, batch 650, loss[loss=0.2296, ctc_loss=0.1519, cr_loss=0.3885, over 14927.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1208, cr_loss=0.3372, over 3225892.64 frames. ], batch size: 89, lr: 2.71e-03, grad_scale: 16.0
2024-09-25 16:31:39,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.16 vs. limit=6.0
2024-09-25 16:32:28,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=784975.3333333334, ans=0.0
2024-09-25 16:32:30,601 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=784975.3333333334, ans=0.125
2024-09-25 16:32:43,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=785022.0, ans=0.125
2024-09-25 16:32:46,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=785022.0, ans=0.025
2024-09-25 16:32:46,723 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=785022.0, ans=0.0
2024-09-25 16:32:53,349 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.41 vs. limit=15.0
2024-09-25 16:32:56,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=785022.0, ans=0.125
2024-09-25 16:32:59,117 INFO [train.py:1198] (1/4) Epoch 44, batch 700, loss[loss=0.2247, ctc_loss=0.1432, cr_loss=0.4076, over 15963.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1216, cr_loss=0.3393, over 3258079.22 frames. ], batch size: 74, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:33:09,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=785068.6666666666, ans=0.035
2024-09-25 16:33:40,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=785162.0, ans=0.2
2024-09-25 16:33:41,733 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.292e+02 1.388e+02 1.458e+02 2.267e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-25 16:33:42,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=785162.0, ans=0.1
2024-09-25 16:33:56,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=785208.6666666666, ans=0.0
2024-09-25 16:33:58,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=785208.6666666666, ans=10.0
2024-09-25 16:34:19,544 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=785255.3333333334, ans=0.0
2024-09-25 16:34:24,043 INFO [train.py:1198] (1/4) Epoch 44, batch 750, loss[loss=0.2063, ctc_loss=0.1362, cr_loss=0.3508, over 17008.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1223, cr_loss=0.3402, over 3282648.13 frames. ], batch size: 51, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:34:30,609 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=785302.0, ans=0.0
2024-09-25 16:35:11,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=9.80 vs. limit=12.0
2024-09-25 16:35:43,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.max_abs, batch_count=785488.6666666666, ans=10.0
2024-09-25 16:35:46,667 INFO [train.py:1198] (1/4) Epoch 44, batch 800, loss[loss=0.1938, ctc_loss=0.1259, cr_loss=0.3393, over 17305.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1212, cr_loss=0.3381, over 3301983.96 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:35:50,891 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.11 vs. limit=22.5
2024-09-25 16:36:07,848 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=785582.0, ans=0.0
2024-09-25 16:36:10,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=785582.0, ans=10.0
2024-09-25 16:36:15,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=785582.0, ans=0.0
2024-09-25 16:36:18,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=785628.6666666666, ans=0.0
2024-09-25 16:36:18,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=785628.6666666666, ans=0.0
2024-09-25 16:36:26,439 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.304e+02 1.360e+02 1.506e+02 2.198e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-25 16:36:42,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=785675.3333333334, ans=10.0
2024-09-25 16:37:09,075 INFO [train.py:1198] (1/4) Epoch 44, batch 850, loss[loss=0.1485, ctc_loss=0.09347, cr_loss=0.275, over 17040.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3384, over 3314885.96 frames. ], batch size: 39, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:37:17,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=785768.6666666666, ans=0.0
2024-09-25 16:37:53,858 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.96 vs. limit=15.0
2024-09-25 16:37:58,579 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.22 vs. limit=15.0
2024-09-25 16:38:24,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=785955.3333333334, ans=0.125
2024-09-25 16:38:32,616 INFO [train.py:1198] (1/4) Epoch 44, batch 900, loss[loss=0.1937, ctc_loss=0.1227, cr_loss=0.3548, over 17168.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1214, cr_loss=0.339, over 3320993.08 frames. ], batch size: 45, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:38:35,285 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=5.11 vs. limit=15.0
2024-09-25 16:38:36,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=786002.0, ans=0.125
2024-09-25 16:38:41,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=786002.0, ans=0.0
2024-09-25 16:38:41,731 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.91 vs. limit=12.0
2024-09-25 16:39:15,049 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.314e+02 1.378e+02 1.470e+02 1.852e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 16:39:15,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=786095.3333333334, ans=0.1
2024-09-25 16:39:42,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=786188.6666666666, ans=0.125
2024-09-25 16:39:48,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=786188.6666666666, ans=0.0
2024-09-25 16:39:56,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=786235.3333333334, ans=0.125
2024-09-25 16:39:58,020 INFO [train.py:1198] (1/4) Epoch 44, batch 950, loss[loss=0.1747, ctc_loss=0.1077, cr_loss=0.3349, over 16962.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1213, cr_loss=0.3388, over 3336301.26 frames. ], batch size: 42, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:39:59,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=786235.3333333334, ans=0.125
2024-09-25 16:40:05,316 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=6.49 vs. limit=15.0
2024-09-25 16:40:10,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=786235.3333333334, ans=0.125
2024-09-25 16:40:33,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=786328.6666666666, ans=0.125
2024-09-25 16:41:18,074 INFO [train.py:1198] (1/4) Epoch 44, batch 1000, loss[loss=0.1771, ctc_loss=0.1129, cr_loss=0.3208, over 17312.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1206, cr_loss=0.3374, over 3341089.44 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:41:27,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=786468.6666666666, ans=0.1
2024-09-25 16:41:45,412 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 16:41:57,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=786562.0, ans=0.0
2024-09-25 16:42:02,020 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.298e+02 1.360e+02 1.466e+02 2.434e+02, threshold=2.721e+02, percent-clipped=0.0
2024-09-25 16:42:37,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=786655.3333333334, ans=0.0
2024-09-25 16:42:40,315 INFO [train.py:1198] (1/4) Epoch 44, batch 1050, loss[loss=0.1897, ctc_loss=0.124, cr_loss=0.3286, over 17021.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.121, cr_loss=0.3377, over 3348758.69 frames. ], batch size: 52, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:42:59,631 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=786748.6666666666, ans=0.07
2024-09-25 16:43:37,292 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 16:43:54,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=786888.6666666666, ans=0.025
2024-09-25 16:44:02,233 INFO [train.py:1198] (1/4) Epoch 44, batch 1100, loss[loss=0.1857, ctc_loss=0.1181, cr_loss=0.3376, over 17267.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1209, cr_loss=0.3371, over 3346280.01 frames. ], batch size: 55, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:44:22,752 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.93 vs. limit=15.0
2024-09-25 16:44:24,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=786982.0, ans=0.125
2024-09-25 16:44:27,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=786982.0, ans=0.1
2024-09-25 16:44:31,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=786982.0, ans=0.125
2024-09-25 16:44:33,727 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.71 vs. limit=15.0
2024-09-25 16:44:34,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=786982.0, ans=0.0
2024-09-25 16:44:36,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=786982.0, ans=0.1
2024-09-25 16:44:48,723 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.281e+02 1.343e+02 1.422e+02 3.447e+02, threshold=2.685e+02, percent-clipped=1.0
2024-09-25 16:45:08,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=787075.3333333334, ans=0.025
2024-09-25 16:45:27,408 INFO [train.py:1198] (1/4) Epoch 44, batch 1150, loss[loss=0.1616, ctc_loss=0.1024, cr_loss=0.2959, over 17109.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1205, cr_loss=0.3371, over 3349267.99 frames. ], batch size: 40, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:45:29,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=787168.6666666666, ans=0.2
2024-09-25 16:46:01,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=787262.0, ans=0.125
2024-09-25 16:46:04,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.11 vs. limit=15.0
2024-09-25 16:46:26,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=787308.6666666666, ans=0.2
2024-09-25 16:46:34,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=787355.3333333334, ans=0.07
2024-09-25 16:46:38,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=787355.3333333334, ans=0.0
2024-09-25 16:46:47,230 INFO [train.py:1198] (1/4) Epoch 44, batch 1200, loss[loss=0.1958, ctc_loss=0.1251, cr_loss=0.3536, over 16995.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1209, cr_loss=0.3374, over 3338519.07 frames. ], batch size: 56, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:46:54,594 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=787402.0, ans=0.0
2024-09-25 16:47:25,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=787495.3333333334, ans=0.125
2024-09-25 16:47:31,123 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.306e+02 1.377e+02 1.476e+02 2.006e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-25 16:47:37,029 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.14 vs. limit=15.0
2024-09-25 16:47:49,569 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=9.33 vs. limit=15.0
2024-09-25 16:47:53,136 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.30 vs. limit=15.0
2024-09-25 16:47:59,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=787588.6666666666, ans=0.125
2024-09-25 16:48:03,108 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=787588.6666666666, ans=0.125
2024-09-25 16:48:06,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=787588.6666666666, ans=0.025
2024-09-25 16:48:11,729 INFO [train.py:1198] (1/4) Epoch 44, batch 1250, loss[loss=0.1567, ctc_loss=0.09935, cr_loss=0.2867, over 16323.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3385, over 3350304.98 frames. ], batch size: 36, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:48:47,684 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=787728.6666666666, ans=0.125
2024-09-25 16:48:55,805 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=787728.6666666666, ans=0.125
2024-09-25 16:48:57,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=787728.6666666666, ans=0.125
2024-09-25 16:49:04,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=787775.3333333334, ans=0.0
2024-09-25 16:49:37,342 INFO [train.py:1198] (1/4) Epoch 44, batch 1300, loss[loss=0.2004, ctc_loss=0.1279, cr_loss=0.3624, over 17212.00 frames. ], tot_loss[loss=0.1894, ctc_loss=0.1215, cr_loss=0.3394, over 3347297.42 frames. ], batch size: 47, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:49:40,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=787868.6666666666, ans=0.0
2024-09-25 16:49:51,023 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.08 vs. limit=22.5
2024-09-25 16:50:01,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=787915.3333333334, ans=0.125
2024-09-25 16:50:11,171 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=787962.0, ans=0.125
2024-09-25 16:50:11,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=787962.0, ans=0.035
2024-09-25 16:50:18,914 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.318e+02 1.377e+02 1.473e+02 1.934e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-25 16:50:34,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=788008.6666666666, ans=0.125
2024-09-25 16:50:51,128 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 16:50:57,003 INFO [train.py:1198] (1/4) Epoch 44, batch 1350, loss[loss=0.16, ctc_loss=0.1012, cr_loss=0.2945, over 16951.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1212, cr_loss=0.3388, over 3352385.53 frames. ], batch size: 42, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:51:03,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=788102.0, ans=10.0
2024-09-25 16:51:08,438 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=788102.0, ans=0.2
2024-09-25 16:51:10,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=788102.0, ans=0.125
2024-09-25 16:51:13,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=788148.6666666666, ans=0.0
2024-09-25 16:52:03,677 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=788288.6666666666, ans=0.0
2024-09-25 16:52:08,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=788288.6666666666, ans=0.0
2024-09-25 16:52:16,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=788288.6666666666, ans=0.125
2024-09-25 16:52:19,417 INFO [train.py:1198] (1/4) Epoch 44, batch 1400, loss[loss=0.2105, ctc_loss=0.1339, cr_loss=0.3832, over 15938.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1207, cr_loss=0.3379, over 3347410.99 frames. ], batch size: 74, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:52:39,905 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.57 vs. limit=22.5
2024-09-25 16:52:55,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=788428.6666666666, ans=0.1
2024-09-25 16:53:01,170 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.288e+02 1.388e+02 1.492e+02 2.105e+02, threshold=2.776e+02, percent-clipped=0.0
2024-09-25 16:53:04,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=788428.6666666666, ans=0.0
2024-09-25 16:53:42,246 INFO [train.py:1198] (1/4) Epoch 44, batch 1450, loss[loss=0.1674, ctc_loss=0.1055, cr_loss=0.3096, over 17248.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1209, cr_loss=0.3379, over 3344540.34 frames. ], batch size: 44, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:54:57,087 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.46 vs. limit=22.5
2024-09-25 16:55:07,624 INFO [train.py:1198] (1/4) Epoch 44, batch 1500, loss[loss=0.1902, ctc_loss=0.1241, cr_loss=0.3308, over 17016.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1207, cr_loss=0.3372, over 3339249.93 frames. ], batch size: 53, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:55:20,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=788802.0, ans=0.025
2024-09-25 16:55:35,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=788848.6666666666, ans=0.1
2024-09-25 16:55:37,376 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.63 vs. limit=15.0
2024-09-25 16:55:50,910 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.297e+02 1.379e+02 1.448e+02 1.999e+02, threshold=2.757e+02, percent-clipped=0.0
2024-09-25 16:56:01,168 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.56 vs. limit=22.5
2024-09-25 16:56:02,485 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=788942.0, ans=0.125
2024-09-25 16:56:13,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=788988.6666666666, ans=0.125
2024-09-25 16:56:19,052 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.61 vs. limit=10.0
2024-09-25 16:56:26,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=789035.3333333334, ans=0.0
2024-09-25 16:56:27,694 INFO [train.py:1198] (1/4) Epoch 44, batch 1550, loss[loss=0.1462, ctc_loss=0.08993, cr_loss=0.2813, over 17091.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1203, cr_loss=0.3365, over 3349387.78 frames. ], batch size: 40, lr: 2.70e-03, grad_scale: 16.0
2024-09-25 16:57:49,474 INFO [train.py:1198] (1/4) Epoch 44, batch 1600, loss[loss=0.2184, ctc_loss=0.1498, cr_loss=0.3428, over 11811.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.12, cr_loss=0.3363, over 3359343.44 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:57:57,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=789268.6666666666, ans=0.125
2024-09-25 16:58:00,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=789268.6666666666, ans=0.025
2024-09-25 16:58:08,831 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=789315.3333333334, ans=0.125
2024-09-25 16:58:14,910 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.31 vs. limit=15.0
2024-09-25 16:58:23,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.06 vs. limit=15.0
2024-09-25 16:58:24,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=789362.0, ans=0.025
2024-09-25 16:58:25,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.min_positive, batch_count=789362.0, ans=0.05
2024-09-25 16:58:34,987 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.295e+02 1.391e+02 1.482e+02 2.401e+02, threshold=2.782e+02, percent-clipped=0.0
2024-09-25 16:58:35,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=789362.0, ans=0.5
2024-09-25 16:58:38,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=789408.6666666666, ans=0.0
2024-09-25 16:58:46,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=789408.6666666666, ans=0.125
2024-09-25 16:58:54,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=789455.3333333334, ans=0.1
2024-09-25 16:59:11,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=789455.3333333334, ans=0.025
2024-09-25 16:59:14,303 INFO [train.py:1198] (1/4) Epoch 44, batch 1650, loss[loss=0.1887, ctc_loss=0.122, cr_loss=0.3338, over 17277.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1211, cr_loss=0.338, over 3355349.02 frames. ], batch size: 46, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 16:59:29,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=789502.0, ans=0.025
2024-09-25 16:59:36,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=789548.6666666666, ans=0.2
2024-09-25 16:59:50,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=789595.3333333334, ans=0.0
2024-09-25 17:00:33,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=789688.6666666666, ans=0.125
2024-09-25 17:00:36,856 INFO [train.py:1198] (1/4) Epoch 44, batch 1700, loss[loss=0.1993, ctc_loss=0.1286, cr_loss=0.3536, over 17187.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.121, cr_loss=0.3378, over 3354605.74 frames. ], batch size: 45, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 17:00:51,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=789782.0, ans=0.025
2024-09-25 17:01:17,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=789828.6666666666, ans=0.2
2024-09-25 17:01:20,178 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.326e+02 1.402e+02 1.495e+02 1.823e+02, threshold=2.804e+02, percent-clipped=0.0
2024-09-25 17:01:24,073 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.50 vs. limit=15.0
2024-09-25 17:01:30,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=789875.3333333334, ans=0.125
2024-09-25 17:01:31,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=789875.3333333334, ans=0.125
2024-09-25 17:01:31,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=789875.3333333334, ans=0.0
2024-09-25 17:01:51,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=789922.0, ans=0.125
2024-09-25 17:01:59,534 INFO [train.py:1198] (1/4) Epoch 44, batch 1750, loss[loss=0.2025, ctc_loss=0.1328, cr_loss=0.3483, over 15930.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1202, cr_loss=0.3368, over 3361530.04 frames. ], batch size: 74, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 17:02:22,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=790015.3333333334, ans=0.1
2024-09-25 17:02:35,136 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:02:38,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=790062.0, ans=0.0
2024-09-25 17:02:41,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=790062.0, ans=0.0
2024-09-25 17:03:09,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=790155.3333333334, ans=0.0
2024-09-25 17:03:22,076 INFO [train.py:1198] (1/4) Epoch 44, batch 1800, loss[loss=0.2239, ctc_loss=0.1485, cr_loss=0.3771, over 16085.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1209, cr_loss=0.3379, over 3349534.29 frames. ], batch size: 74, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 17:03:31,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.47 vs. limit=12.0
2024-09-25 17:03:51,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=790248.6666666666, ans=0.0
2024-09-25 17:04:08,009 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.325e+02 1.389e+02 1.497e+02 2.037e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-25 17:04:08,302 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=790295.3333333334, ans=0.0
2024-09-25 17:04:18,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=790342.0, ans=0.0
2024-09-25 17:04:23,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=790342.0, ans=0.0
2024-09-25 17:04:31,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=790388.6666666666, ans=0.0
2024-09-25 17:04:47,702 INFO [train.py:1198] (1/4) Epoch 44, batch 1850, loss[loss=0.2032, ctc_loss=0.1319, cr_loss=0.3566, over 15131.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3382, over 3344783.12 frames. ], batch size: 89, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 17:04:52,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=790435.3333333334, ans=0.0
2024-09-25 17:05:07,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=790482.0, ans=0.1
2024-09-25 17:05:18,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=790528.6666666666, ans=0.0
2024-09-25 17:05:55,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=790622.0, ans=0.125
2024-09-25 17:06:02,775 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.81 vs. limit=12.0
2024-09-25 17:06:08,120 INFO [train.py:1198] (1/4) Epoch 44, batch 1900, loss[loss=0.2145, ctc_loss=0.1391, cr_loss=0.3773, over 17017.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1213, cr_loss=0.3384, over 3351035.00 frames. ], batch size: 56, lr: 2.70e-03, grad_scale: 32.0
2024-09-25 17:06:11,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=790668.6666666666, ans=0.125
2024-09-25 17:06:14,755 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=790668.6666666666, ans=0.025
2024-09-25 17:06:46,415 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=790762.0, ans=0.125
2024-09-25 17:06:46,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=790762.0, ans=0.125
2024-09-25 17:06:48,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=790762.0, ans=0.1
2024-09-25 17:06:54,117 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.292e+02 1.379e+02 1.443e+02 2.422e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-25 17:06:54,529 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:06:54,583 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=790762.0, ans=0.0
2024-09-25 17:06:59,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=790808.6666666666, ans=0.0
2024-09-25 17:07:31,338 INFO [train.py:1198] (1/4) Epoch 44, batch 1950, loss[loss=0.1777, ctc_loss=0.1118, cr_loss=0.3295, over 17207.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1211, cr_loss=0.3381, over 3356518.75 frames. ], batch size: 50, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:08:02,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=790995.3333333334, ans=0.2
2024-09-25 17:08:02,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=790995.3333333334, ans=0.0
2024-09-25 17:08:17,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=790995.3333333334, ans=0.2
2024-09-25 17:08:27,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=791042.0, ans=0.125
2024-09-25 17:08:56,404 INFO [train.py:1198] (1/4) Epoch 44, batch 2000, loss[loss=0.1704, ctc_loss=0.1076, cr_loss=0.3143, over 17299.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1209, cr_loss=0.3377, over 3350530.98 frames. ], batch size: 51, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:09:02,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=791135.3333333334, ans=0.07
2024-09-25 17:09:10,908 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=791182.0, ans=0.0
2024-09-25 17:09:13,174 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.06 vs. limit=12.0
2024-09-25 17:09:20,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=791182.0, ans=0.07
2024-09-25 17:09:43,343 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.332e+02 1.440e+02 1.511e+02 2.187e+02, threshold=2.879e+02, percent-clipped=0.0
2024-09-25 17:09:50,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=9.19 vs. limit=15.0
2024-09-25 17:09:55,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=791275.3333333334, ans=0.2
2024-09-25 17:09:58,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=791275.3333333334, ans=0.125
2024-09-25 17:10:15,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=791322.0, ans=0.125
2024-09-25 17:10:17,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=791368.6666666666, ans=0.2
2024-09-25 17:10:18,723 INFO [train.py:1198] (1/4) Epoch 44, batch 2050, loss[loss=0.1817, ctc_loss=0.1154, cr_loss=0.3317, over 17100.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1202, cr_loss=0.3367, over 3354224.59 frames. ], batch size: 43, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:10:31,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=791368.6666666666, ans=0.04949747468305833
2024-09-25 17:10:41,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=791415.3333333334, ans=0.125
2024-09-25 17:10:57,356 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=791462.0, ans=0.0
2024-09-25 17:11:11,933 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=13.35 vs. limit=22.5
2024-09-25 17:11:38,300 INFO [train.py:1198] (1/4) Epoch 44, batch 2100, loss[loss=0.1429, ctc_loss=0.08768, cr_loss=0.2759, over 17034.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.12, cr_loss=0.3361, over 3361760.37 frames. ], batch size: 39, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:11:50,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=791602.0, ans=0.0
2024-09-25 17:11:56,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=791648.6666666666, ans=0.2
2024-09-25 17:12:22,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=791695.3333333334, ans=0.2
2024-09-25 17:12:25,395 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.332e+02 1.398e+02 1.467e+02 2.660e+02, threshold=2.796e+02, percent-clipped=0.0
2024-09-25 17:13:00,854 INFO [train.py:1198] (1/4) Epoch 44, batch 2150, loss[loss=0.1934, ctc_loss=0.1235, cr_loss=0.3493, over 16740.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1197, cr_loss=0.3357, over 3361197.00 frames. ], batch size: 61, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:13:02,832 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=791835.3333333334, ans=0.0
2024-09-25 17:13:26,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=791882.0, ans=0.015
2024-09-25 17:13:31,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=791882.0, ans=0.125
2024-09-25 17:14:20,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=792022.0, ans=0.125
2024-09-25 17:14:28,853 INFO [train.py:1198] (1/4) Epoch 44, batch 2200, loss[loss=0.1976, ctc_loss=0.1281, cr_loss=0.3476, over 16527.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1194, cr_loss=0.3355, over 3366609.21 frames. ], batch size: 66, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:14:35,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=792068.6666666666, ans=0.125
2024-09-25 17:14:40,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=792068.6666666666, ans=0.0
2024-09-25 17:14:55,297 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.99 vs. limit=6.0
2024-09-25 17:15:13,651 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.270e+02 1.367e+02 1.447e+02 1.926e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-25 17:15:48,885 INFO [train.py:1198] (1/4) Epoch 44, batch 2250, loss[loss=0.1856, ctc_loss=0.1167, cr_loss=0.3446, over 17214.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1199, cr_loss=0.3367, over 3364548.62 frames. ], batch size: 55, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:16:01,227 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.74 vs. limit=12.0
2024-09-25 17:16:16,441 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:16:16,851 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.56 vs. limit=15.0
2024-09-25 17:17:11,219 INFO [train.py:1198] (1/4) Epoch 44, batch 2300, loss[loss=0.2031, ctc_loss=0.1293, cr_loss=0.3688, over 17220.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1194, cr_loss=0.3354, over 3366018.13 frames. ], batch size: 47, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:17:21,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=792535.3333333334, ans=0.1
2024-09-25 17:17:43,282 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=792628.6666666666, ans=0.025
2024-09-25 17:17:51,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=792628.6666666666, ans=0.125
2024-09-25 17:17:55,694 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.302e+02 1.387e+02 1.511e+02 2.811e+02, threshold=2.774e+02, percent-clipped=1.0
2024-09-25 17:18:22,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=792722.0, ans=0.125
2024-09-25 17:18:27,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=792722.0, ans=0.2
2024-09-25 17:18:28,412 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=5.01 vs. limit=10.0
2024-09-25 17:18:33,646 INFO [train.py:1198] (1/4) Epoch 44, batch 2350, loss[loss=0.1544, ctc_loss=0.09519, cr_loss=0.2961, over 17018.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.1192, cr_loss=0.3347, over 3367183.15 frames. ], batch size: 39, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:19:16,588 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:19:35,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=792908.6666666666, ans=0.125
2024-09-25 17:19:40,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=792908.6666666666, ans=0.2
2024-09-25 17:19:48,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=792955.3333333334, ans=0.125
2024-09-25 17:19:59,162 INFO [train.py:1198] (1/4) Epoch 44, batch 2400, loss[loss=0.1761, ctc_loss=0.1116, cr_loss=0.3223, over 17020.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1189, cr_loss=0.3345, over 3361529.38 frames. ], batch size: 44, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:20:06,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=793002.0, ans=0.1
2024-09-25 17:20:12,573 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.03 vs. limit=22.5
2024-09-25 17:20:26,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=793048.6666666666, ans=0.125
2024-09-25 17:20:45,389 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.312e+02 1.417e+02 1.545e+02 2.964e+02, threshold=2.835e+02, percent-clipped=1.0
2024-09-25 17:20:51,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=793142.0, ans=0.015
2024-09-25 17:21:01,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=793188.6666666666, ans=0.125
2024-09-25 17:21:09,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=793188.6666666666, ans=0.2
2024-09-25 17:21:19,082 INFO [train.py:1198] (1/4) Epoch 44, batch 2450, loss[loss=0.2277, ctc_loss=0.1465, cr_loss=0.406, over 14911.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1194, cr_loss=0.3358, over 3365355.76 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:21:40,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=793282.0, ans=0.0
2024-09-25 17:22:07,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=793328.6666666666, ans=0.0
2024-09-25 17:22:42,508 INFO [train.py:1198] (1/4) Epoch 44, batch 2500, loss[loss=0.1704, ctc_loss=0.1092, cr_loss=0.3061, over 17027.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1193, cr_loss=0.3358, over 3362855.77 frames. ], batch size: 39, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:22:57,978 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.83 vs. limit=15.0
2024-09-25 17:23:14,471 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:23:31,901 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.312e+02 1.368e+02 1.453e+02 1.982e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 17:23:49,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=793655.3333333334, ans=0.125
2024-09-25 17:23:58,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=793655.3333333334, ans=0.0
2024-09-25 17:24:08,164 INFO [train.py:1198] (1/4) Epoch 44, batch 2550, loss[loss=0.1772, ctc_loss=0.1134, cr_loss=0.3187, over 17025.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1197, cr_loss=0.3358, over 3358227.78 frames. ], batch size: 51, lr: 2.69e-03, grad_scale: 16.0
2024-09-25 17:24:16,434 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:25:00,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=793842.0, ans=0.0
2024-09-25 17:25:16,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=793888.6666666666, ans=0.0
2024-09-25 17:25:31,105 INFO [train.py:1198] (1/4) Epoch 44, batch 2600, loss[loss=0.2063, ctc_loss=0.1309, cr_loss=0.377, over 16872.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1197, cr_loss=0.3355, over 3363196.56 frames. ], batch size: 58, lr: 2.69e-03, grad_scale: 16.0
2024-09-25 17:25:36,909 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.34 vs. limit=15.0
2024-09-25 17:25:45,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=793982.0, ans=0.125
2024-09-25 17:25:45,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=793982.0, ans=0.0
2024-09-25 17:26:07,370 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.95 vs. limit=15.0
2024-09-25 17:26:19,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.289e+02 1.404e+02 1.511e+02 2.276e+02, threshold=2.809e+02, percent-clipped=0.0
2024-09-25 17:26:54,091 INFO [train.py:1198] (1/4) Epoch 44, batch 2650, loss[loss=0.1928, ctc_loss=0.1278, cr_loss=0.3246, over 11930.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1202, cr_loss=0.3364, over 3359715.19 frames. ], batch size: 124, lr: 2.69e-03, grad_scale: 16.0
2024-09-25 17:26:59,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=794168.6666666666, ans=0.015
2024-09-25 17:27:02,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=794168.6666666666, ans=0.125
2024-09-25 17:28:15,225 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.47 vs. limit=22.5
2024-09-25 17:28:17,365 INFO [train.py:1198] (1/4) Epoch 44, batch 2700, loss[loss=0.2234, ctc_loss=0.1492, cr_loss=0.3707, over 14904.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1204, cr_loss=0.3365, over 3362870.68 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 16.0
2024-09-25 17:29:07,857 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.291e+02 1.350e+02 1.441e+02 1.690e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-25 17:29:18,081 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=794542.0, ans=0.025
2024-09-25 17:29:27,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=794588.6666666666, ans=0.025
2024-09-25 17:29:29,432 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=11.51 vs. limit=22.5
2024-09-25 17:29:32,592 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.73 vs. limit=15.0
2024-09-25 17:29:43,111 INFO [train.py:1198] (1/4) Epoch 44, batch 2750, loss[loss=0.1777, ctc_loss=0.1147, cr_loss=0.3152, over 16888.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1203, cr_loss=0.3365, over 3360096.68 frames. ], batch size: 58, lr: 2.69e-03, grad_scale: 16.0
2024-09-25 17:30:21,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=794728.6666666666, ans=0.125
2024-09-25 17:30:26,674 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.84 vs. limit=15.0
2024-09-25 17:30:34,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=794775.3333333334, ans=0.125
2024-09-25 17:31:02,586 INFO [train.py:1198] (1/4) Epoch 44, batch 2800, loss[loss=0.1976, ctc_loss=0.1256, cr_loss=0.3602, over 16744.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1198, cr_loss=0.3363, over 3367256.82 frames. ], batch size: 61, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:31:29,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=794915.3333333334, ans=0.1
2024-09-25 17:31:31,950 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.62 vs. limit=22.5
2024-09-25 17:31:45,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=794962.0, ans=0.125
2024-09-25 17:31:47,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.01 vs. limit=22.5
2024-09-25 17:31:52,769 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.331e+02 1.405e+02 1.538e+02 1.952e+02, threshold=2.809e+02, percent-clipped=0.0
2024-09-25 17:32:03,166 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.48 vs. limit=22.5
2024-09-25 17:32:06,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=795008.6666666666, ans=0.025
2024-09-25 17:32:13,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=795055.3333333334, ans=0.125
2024-09-25 17:32:24,904 INFO [train.py:1198] (1/4) Epoch 44, batch 2850, loss[loss=0.1932, ctc_loss=0.1239, cr_loss=0.3468, over 17217.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1198, cr_loss=0.3366, over 3356837.80 frames. ], batch size: 50, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:32:41,436 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=795148.6666666666, ans=0.0
2024-09-25 17:32:49,532 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:32:53,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer_na.min_abs, batch_count=795148.6666666666, ans=0.02
2024-09-25 17:33:27,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=795242.0, ans=0.0
2024-09-25 17:33:47,967 INFO [train.py:1198] (1/4) Epoch 44, batch 2900, loss[loss=0.213, ctc_loss=0.1384, cr_loss=0.3734, over 16730.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1202, cr_loss=0.3368, over 3347200.16 frames. ], batch size: 61, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:33:48,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=795335.3333333334, ans=0.125
2024-09-25 17:34:37,767 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.00 vs. limit=12.0
2024-09-25 17:34:41,409 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.299e+02 1.362e+02 1.425e+02 2.572e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-25 17:35:03,070 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=13.30 vs. limit=22.5
2024-09-25 17:35:05,644 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=795522.0, ans=0.125
2024-09-25 17:35:08,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=795522.0, ans=0.125
2024-09-25 17:35:12,173 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=795568.6666666666, ans=0.025
2024-09-25 17:35:13,453 INFO [train.py:1198] (1/4) Epoch 44, batch 2950, loss[loss=0.1737, ctc_loss=0.109, cr_loss=0.3236, over 17135.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1215, cr_loss=0.3393, over 3350906.73 frames. ], batch size: 48, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:35:46,563 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.28 vs. limit=15.0
2024-09-25 17:36:01,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=795708.6666666666, ans=0.1
2024-09-25 17:36:29,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=795755.3333333334, ans=0.125
2024-09-25 17:36:31,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=795802.0, ans=0.0
2024-09-25 17:36:32,577 INFO [train.py:1198] (1/4) Epoch 44, batch 3000, loss[loss=0.1861, ctc_loss=0.1188, cr_loss=0.3363, over 17258.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1216, cr_loss=0.3396, over 3351272.85 frames. ], batch size: 44, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:36:32,577 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 17:36:47,863 INFO [train.py:1230] (1/4) Epoch 44, validation: loss=0.03521, ctc_loss=0.03521, cr_loss=1.022e-14, over 944034.00 frames. 
2024-09-25 17:36:47,863 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 17:37:00,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=795802.0, ans=0.0
2024-09-25 17:37:19,952 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=4.87 vs. limit=15.0
2024-09-25 17:37:34,778 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.305e+02 1.379e+02 1.486e+02 1.924e+02, threshold=2.758e+02, percent-clipped=0.0
2024-09-25 17:38:01,941 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.99 vs. limit=15.0
2024-09-25 17:38:03,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=795988.6666666666, ans=0.125
2024-09-25 17:38:06,191 INFO [train.py:1198] (1/4) Epoch 44, batch 3050, loss[loss=0.2019, ctc_loss=0.1289, cr_loss=0.365, over 17040.00 frames. ], tot_loss[loss=0.19, ctc_loss=0.1219, cr_loss=0.3407, over 3356042.18 frames. ], batch size: 52, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:38:39,551 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=796128.6666666666, ans=0.125
2024-09-25 17:38:45,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=796128.6666666666, ans=0.1
2024-09-25 17:39:25,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=796268.6666666666, ans=0.0
2024-09-25 17:39:26,748 INFO [train.py:1198] (1/4) Epoch 44, batch 3100, loss[loss=0.1553, ctc_loss=0.09541, cr_loss=0.2997, over 16957.00 frames. ], tot_loss[loss=0.1903, ctc_loss=0.1222, cr_loss=0.3408, over 3354787.85 frames. ], batch size: 42, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:39:27,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.51 vs. limit=5.0
2024-09-25 17:39:48,980 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=796315.3333333334, ans=0.125
2024-09-25 17:40:02,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=796362.0, ans=0.125
2024-09-25 17:40:03,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=796362.0, ans=0.025
2024-09-25 17:40:13,596 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.310e+02 1.386e+02 1.469e+02 1.981e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 17:40:32,747 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=796455.3333333334, ans=0.125
2024-09-25 17:40:44,923 INFO [train.py:1198] (1/4) Epoch 44, batch 3150, loss[loss=0.1793, ctc_loss=0.1137, cr_loss=0.3279, over 17297.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1215, cr_loss=0.3397, over 3351949.02 frames. ], batch size: 51, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:40:45,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=796502.0, ans=0.0
2024-09-25 17:41:12,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=796548.6666666666, ans=0.125
2024-09-25 17:41:36,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=796642.0, ans=0.035
2024-09-25 17:42:08,192 INFO [train.py:1198] (1/4) Epoch 44, batch 3200, loss[loss=0.1907, ctc_loss=0.1215, cr_loss=0.3462, over 17008.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1211, cr_loss=0.3388, over 3345752.43 frames. ], batch size: 53, lr: 2.69e-03, grad_scale: 32.0
2024-09-25 17:42:19,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=796735.3333333334, ans=0.125
2024-09-25 17:42:40,523 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.92 vs. limit=10.0
2024-09-25 17:42:57,898 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.350e+02 1.424e+02 1.561e+02 1.915e+02, threshold=2.848e+02, percent-clipped=0.0
2024-09-25 17:43:01,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=796875.3333333334, ans=0.125
2024-09-25 17:43:10,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=796922.0, ans=0.0
2024-09-25 17:43:26,125 INFO [train.py:1198] (1/4) Epoch 44, batch 3250, loss[loss=0.1526, ctc_loss=0.09716, cr_loss=0.2773, over 17084.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3383, over 3347301.76 frames. ], batch size: 40, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:43:27,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=796968.6666666666, ans=0.1
2024-09-25 17:44:01,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=797062.0, ans=0.025
2024-09-25 17:44:02,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=797062.0, ans=0.1
2024-09-25 17:44:42,346 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=7.82 vs. limit=15.0
2024-09-25 17:44:45,004 INFO [train.py:1198] (1/4) Epoch 44, batch 3300, loss[loss=0.1808, ctc_loss=0.1175, cr_loss=0.3165, over 17302.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1206, cr_loss=0.3375, over 3353038.40 frames. ], batch size: 49, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:45:00,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=797248.6666666666, ans=0.1
2024-09-25 17:45:13,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=797248.6666666666, ans=0.125
2024-09-25 17:45:25,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=797295.3333333334, ans=0.07
2024-09-25 17:45:27,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=797295.3333333334, ans=0.0
2024-09-25 17:45:34,580 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.296e+02 1.381e+02 1.496e+02 2.395e+02, threshold=2.762e+02, percent-clipped=0.0
2024-09-25 17:45:34,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=797342.0, ans=0.125
2024-09-25 17:45:47,373 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=797388.6666666666, ans=0.0
2024-09-25 17:45:53,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=797388.6666666666, ans=0.125
2024-09-25 17:46:01,481 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:46:02,721 INFO [train.py:1198] (1/4) Epoch 44, batch 3350, loss[loss=0.2533, ctc_loss=0.1674, cr_loss=0.4291, over 14952.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1211, cr_loss=0.3388, over 3348918.96 frames. ], batch size: 89, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:46:05,078 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.25 vs. limit=10.0
2024-09-25 17:46:29,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=797482.0, ans=0.0
2024-09-25 17:46:59,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=797575.3333333334, ans=0.1
2024-09-25 17:47:02,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=797575.3333333334, ans=0.0
2024-09-25 17:47:21,459 INFO [train.py:1198] (1/4) Epoch 44, batch 3400, loss[loss=0.1953, ctc_loss=0.1244, cr_loss=0.3545, over 17221.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1214, cr_loss=0.3394, over 3342194.57 frames. ], batch size: 50, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:47:49,282 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:48:12,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=797808.6666666666, ans=0.125
2024-09-25 17:48:13,774 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.291e+02 1.376e+02 1.452e+02 2.263e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-25 17:48:42,146 INFO [train.py:1198] (1/4) Epoch 44, batch 3450, loss[loss=0.1844, ctc_loss=0.1194, cr_loss=0.3249, over 17324.00 frames. ], tot_loss[loss=0.1886, ctc_loss=0.121, cr_loss=0.3384, over 3352487.94 frames. ], batch size: 51, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:49:02,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=797948.6666666666, ans=0.0
2024-09-25 17:49:13,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=797995.3333333334, ans=0.025
2024-09-25 17:49:34,436 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.23 vs. limit=15.0
2024-09-25 17:49:47,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=798088.6666666666, ans=0.125
2024-09-25 17:49:52,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=798088.6666666666, ans=0.0
2024-09-25 17:50:00,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=798135.3333333334, ans=0.125
2024-09-25 17:50:02,278 INFO [train.py:1198] (1/4) Epoch 44, batch 3500, loss[loss=0.1837, ctc_loss=0.1184, cr_loss=0.3265, over 17213.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1208, cr_loss=0.3381, over 3347775.75 frames. ], batch size: 50, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:50:23,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=798182.0, ans=0.2
2024-09-25 17:50:23,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=798182.0, ans=0.125
2024-09-25 17:50:48,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=798275.3333333334, ans=0.125
2024-09-25 17:50:52,796 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.299e+02 1.358e+02 1.430e+02 3.438e+02, threshold=2.715e+02, percent-clipped=1.0
2024-09-25 17:51:06,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=798322.0, ans=0.125
2024-09-25 17:51:16,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=798322.0, ans=0.0
2024-09-25 17:51:16,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=798322.0, ans=0.125
2024-09-25 17:51:18,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=798322.0, ans=0.125
2024-09-25 17:51:20,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=798322.0, ans=0.125
2024-09-25 17:51:22,862 INFO [train.py:1198] (1/4) Epoch 44, batch 3550, loss[loss=0.1449, ctc_loss=0.0888, cr_loss=0.2805, over 16300.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1212, cr_loss=0.3391, over 3342469.73 frames. ], batch size: 36, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:51:34,441 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.81 vs. limit=15.0
2024-09-25 17:51:42,372 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.09 vs. limit=15.0
2024-09-25 17:51:58,761 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.94 vs. limit=22.5
2024-09-25 17:52:20,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=798508.6666666666, ans=0.0
2024-09-25 17:52:42,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=798602.0, ans=0.025
2024-09-25 17:52:43,564 INFO [train.py:1198] (1/4) Epoch 44, batch 3600, loss[loss=0.2088, ctc_loss=0.1343, cr_loss=0.3724, over 16539.00 frames. ], tot_loss[loss=0.1897, ctc_loss=0.1217, cr_loss=0.3398, over 3344486.48 frames. ], batch size: 66, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:52:46,208 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.83 vs. limit=10.0
2024-09-25 17:52:54,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=798602.0, ans=0.125
2024-09-25 17:53:07,569 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.38 vs. limit=15.0
2024-09-25 17:53:09,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=798648.6666666666, ans=0.0
2024-09-25 17:53:16,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=798695.3333333334, ans=0.125
2024-09-25 17:53:17,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=798695.3333333334, ans=0.1
2024-09-25 17:53:34,701 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.315e+02 1.375e+02 1.458e+02 2.973e+02, threshold=2.750e+02, percent-clipped=1.0
2024-09-25 17:53:47,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=798788.6666666666, ans=0.125
2024-09-25 17:54:01,420 INFO [train.py:1198] (1/4) Epoch 44, batch 3650, loss[loss=0.1902, ctc_loss=0.1216, cr_loss=0.3428, over 17207.00 frames. ], tot_loss[loss=0.1898, ctc_loss=0.1218, cr_loss=0.3402, over 3349643.66 frames. ], batch size: 50, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:54:36,781 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.85 vs. limit=6.0
2024-09-25 17:54:41,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=798928.6666666666, ans=0.1
2024-09-25 17:55:11,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=799022.0, ans=0.2
2024-09-25 17:55:20,636 INFO [train.py:1198] (1/4) Epoch 44, batch 3700, loss[loss=0.1789, ctc_loss=0.1163, cr_loss=0.3126, over 17141.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1212, cr_loss=0.3388, over 3352484.66 frames. ], batch size: 48, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:55:22,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=799068.6666666666, ans=0.125
2024-09-25 17:55:27,602 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.39 vs. limit=15.0
2024-09-25 17:55:42,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=799115.3333333334, ans=0.125
2024-09-25 17:56:11,868 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.294e+02 1.350e+02 1.442e+02 2.627e+02, threshold=2.701e+02, percent-clipped=0.0
2024-09-25 17:56:38,700 INFO [train.py:1198] (1/4) Epoch 44, batch 3750, loss[loss=0.1667, ctc_loss=0.1037, cr_loss=0.3152, over 17190.00 frames. ], tot_loss[loss=0.1892, ctc_loss=0.1214, cr_loss=0.339, over 3344933.30 frames. ], batch size: 41, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:56:54,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=799348.6666666666, ans=0.2
2024-09-25 17:56:56,274 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 17:57:17,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=799395.3333333334, ans=0.5
2024-09-25 17:57:24,899 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=799442.0, ans=0.0
2024-09-25 17:57:57,078 INFO [train.py:1198] (1/4) Epoch 44, batch 3800, loss[loss=0.2229, ctc_loss=0.1429, cr_loss=0.3997, over 16505.00 frames. ], tot_loss[loss=0.1895, ctc_loss=0.1216, cr_loss=0.3392, over 3315859.27 frames. ], batch size: 66, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:58:01,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=799535.3333333334, ans=0.2
2024-09-25 17:58:44,639 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.32 vs. limit=15.0
2024-09-25 17:58:47,998 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 1.317e+02 1.409e+02 1.513e+02 1.849e+02, threshold=2.818e+02, percent-clipped=0.0
2024-09-25 17:58:52,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=799675.3333333334, ans=0.125
2024-09-25 17:58:54,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=799675.3333333334, ans=0.125
2024-09-25 17:58:58,976 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=799722.0, ans=0.2
2024-09-25 17:59:03,645 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.17 vs. limit=10.0
2024-09-25 17:59:07,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=799722.0, ans=0.0
2024-09-25 17:59:14,335 INFO [train.py:1198] (1/4) Epoch 44, batch 3850, loss[loss=0.2178, ctc_loss=0.1459, cr_loss=0.3597, over 11579.00 frames. ], tot_loss[loss=0.1931, ctc_loss=0.1244, cr_loss=0.3434, over 3267890.05 frames. ], batch size: 123, lr: 2.68e-03, grad_scale: 16.0
2024-09-25 17:59:34,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=799815.3333333334, ans=0.125
2024-09-25 18:00:11,205 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=799908.6666666666, ans=0.125
2024-09-25 18:00:17,309 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:00:20,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=799955.3333333334, ans=0.125
2024-09-25 18:01:14,622 INFO [train.py:1198] (1/4) Epoch 45, batch 0, loss[loss=0.2037, ctc_loss=0.1319, cr_loss=0.3592, over 17339.00 frames. ], tot_loss[loss=0.2037, ctc_loss=0.1319, cr_loss=0.3592, over 17339.00 frames. ], batch size: 48, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:01:14,623 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 18:01:29,794 INFO [train.py:1230] (1/4) Epoch 45, validation: loss=0.03539, ctc_loss=0.03539, cr_loss=1.113e-14, over 944034.00 frames. 
2024-09-25 18:01:29,795 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 18:02:16,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=800076.6666666666, ans=0.2
2024-09-25 18:02:18,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=800076.6666666666, ans=0.125
2024-09-25 18:02:21,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=800123.3333333334, ans=0.025
2024-09-25 18:02:31,990 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.406e+02 1.547e+02 1.686e+02 2.322e+02, threshold=3.093e+02, percent-clipped=0.0
2024-09-25 18:02:49,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=800170.0, ans=0.125
2024-09-25 18:02:52,919 INFO [train.py:1198] (1/4) Epoch 45, batch 50, loss[loss=0.1876, ctc_loss=0.1179, cr_loss=0.3486, over 17073.00 frames. ], tot_loss[loss=0.1905, ctc_loss=0.1223, cr_loss=0.3409, over 762657.63 frames. ], batch size: 46, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:02:59,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=800216.6666666666, ans=0.0
2024-09-25 18:03:10,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=800263.3333333334, ans=0.04949747468305833
2024-09-25 18:03:10,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=800263.3333333334, ans=0.2
2024-09-25 18:03:39,585 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=800356.6666666666, ans=0.125
2024-09-25 18:03:44,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=800356.6666666666, ans=0.125
2024-09-25 18:04:13,066 INFO [train.py:1198] (1/4) Epoch 45, batch 100, loss[loss=0.1767, ctc_loss=0.1114, cr_loss=0.3268, over 17009.00 frames. ], tot_loss[loss=0.191, ctc_loss=0.1226, cr_loss=0.3423, over 1335976.66 frames. ], batch size: 44, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:04:21,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=800450.0, ans=0.2
2024-09-25 18:04:24,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=800450.0, ans=0.125
2024-09-25 18:04:55,105 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=800543.3333333334, ans=0.2
2024-09-25 18:05:12,263 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.307e+02 1.359e+02 1.469e+02 2.520e+02, threshold=2.719e+02, percent-clipped=0.0
2024-09-25 18:05:35,981 INFO [train.py:1198] (1/4) Epoch 45, batch 150, loss[loss=0.1971, ctc_loss=0.1259, cr_loss=0.356, over 17100.00 frames. ], tot_loss[loss=0.1916, ctc_loss=0.123, cr_loss=0.343, over 1771607.63 frames. ], batch size: 49, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:05:48,703 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=800683.3333333334, ans=0.125
2024-09-25 18:05:58,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=800730.0, ans=0.125
2024-09-25 18:06:09,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=800776.6666666666, ans=0.0
2024-09-25 18:06:35,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=800823.3333333334, ans=0.125
2024-09-25 18:06:43,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=800870.0, ans=0.125
2024-09-25 18:06:50,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=800870.0, ans=0.0
2024-09-25 18:07:02,958 INFO [train.py:1198] (1/4) Epoch 45, batch 200, loss[loss=0.1865, ctc_loss=0.1225, cr_loss=0.3204, over 17313.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1212, cr_loss=0.3388, over 2113628.82 frames. ], batch size: 51, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:07:19,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.59 vs. limit=22.5
2024-09-25 18:07:24,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=800963.3333333334, ans=0.2
2024-09-25 18:07:25,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=800963.3333333334, ans=0.125
2024-09-25 18:07:28,799 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=800963.3333333334, ans=0.125
2024-09-25 18:07:36,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=801010.0, ans=0.0
2024-09-25 18:07:44,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=801010.0, ans=0.1
2024-09-25 18:07:51,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=801056.6666666666, ans=0.125
2024-09-25 18:08:02,078 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.278e+02 1.376e+02 1.483e+02 1.957e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-25 18:08:08,318 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.79 vs. limit=15.0
2024-09-25 18:08:23,394 INFO [train.py:1198] (1/4) Epoch 45, batch 250, loss[loss=0.1649, ctc_loss=0.1025, cr_loss=0.3118, over 17166.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.12, cr_loss=0.3359, over 2397338.37 frames. ], batch size: 41, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:08:29,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=801150.0, ans=0.1
2024-09-25 18:08:42,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=801196.6666666666, ans=0.0
2024-09-25 18:08:58,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=801243.3333333334, ans=0.125
2024-09-25 18:09:26,005 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=801336.6666666666, ans=0.07
2024-09-25 18:09:38,528 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=801336.6666666666, ans=0.125
2024-09-25 18:09:43,107 INFO [train.py:1198] (1/4) Epoch 45, batch 300, loss[loss=0.2143, ctc_loss=0.1387, cr_loss=0.3781, over 16968.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1201, cr_loss=0.3365, over 2608932.15 frames. ], batch size: 58, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:10:05,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=801430.0, ans=0.0
2024-09-25 18:10:23,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=801476.6666666666, ans=0.2
2024-09-25 18:10:23,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=801476.6666666666, ans=0.125
2024-09-25 18:10:48,157 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.282e+02 1.376e+02 1.463e+02 2.041e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-25 18:11:09,114 INFO [train.py:1198] (1/4) Epoch 45, batch 350, loss[loss=0.1953, ctc_loss=0.1246, cr_loss=0.3531, over 17358.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1204, cr_loss=0.3372, over 2773661.29 frames. ], batch size: 48, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:11:27,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=801663.3333333334, ans=0.125
2024-09-25 18:11:37,392 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.80 vs. limit=22.5
2024-09-25 18:11:45,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=801710.0, ans=0.0
2024-09-25 18:11:45,725 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=14.34 vs. limit=15.0
2024-09-25 18:11:47,224 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.24 vs. limit=15.0
2024-09-25 18:12:14,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=801756.6666666666, ans=0.07
2024-09-25 18:12:34,306 INFO [train.py:1198] (1/4) Epoch 45, batch 400, loss[loss=0.1632, ctc_loss=0.1018, cr_loss=0.307, over 17291.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1206, cr_loss=0.3382, over 2902008.91 frames. ], batch size: 46, lr: 2.65e-03, grad_scale: 32.0
2024-09-25 18:12:39,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=801850.0, ans=0.125
2024-09-25 18:12:51,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=801896.6666666666, ans=0.125
2024-09-25 18:13:31,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=801990.0, ans=0.125
2024-09-25 18:13:34,523 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.278e+02 1.369e+02 1.473e+02 1.980e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-25 18:13:35,274 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.98 vs. limit=10.0
2024-09-25 18:13:53,909 INFO [train.py:1198] (1/4) Epoch 45, batch 450, loss[loss=0.1997, ctc_loss=0.1285, cr_loss=0.3558, over 17218.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.12, cr_loss=0.3367, over 3006873.51 frames. ], batch size: 50, lr: 2.65e-03, grad_scale: 16.0
2024-09-25 18:14:05,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=802083.3333333334, ans=0.0
2024-09-25 18:14:07,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=802083.3333333334, ans=0.125
2024-09-25 18:14:16,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=802130.0, ans=0.1
2024-09-25 18:14:26,136 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=802176.6666666666, ans=0.2
2024-09-25 18:14:29,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=802176.6666666666, ans=0.125
2024-09-25 18:14:29,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.22 vs. limit=15.0
2024-09-25 18:14:34,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=802176.6666666666, ans=0.1
2024-09-25 18:14:39,495 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.97 vs. limit=15.0
2024-09-25 18:14:42,233 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:14:43,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=802223.3333333334, ans=0.125
2024-09-25 18:15:16,371 INFO [train.py:1198] (1/4) Epoch 45, batch 500, loss[loss=0.1851, ctc_loss=0.1183, cr_loss=0.3337, over 17174.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1196, cr_loss=0.3363, over 3090685.76 frames. ], batch size: 45, lr: 2.65e-03, grad_scale: 16.0
2024-09-25 18:15:18,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=802316.6666666666, ans=0.1
2024-09-25 18:15:21,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=802316.6666666666, ans=0.95
2024-09-25 18:15:24,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=802316.6666666666, ans=0.0
2024-09-25 18:15:43,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=802363.3333333334, ans=0.125
2024-09-25 18:16:22,896 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.274e+02 1.342e+02 1.426e+02 1.767e+02, threshold=2.684e+02, percent-clipped=0.0
2024-09-25 18:16:26,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=802503.3333333334, ans=0.1
2024-09-25 18:16:42,132 INFO [train.py:1198] (1/4) Epoch 45, batch 550, loss[loss=0.1532, ctc_loss=0.09713, cr_loss=0.2802, over 16675.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1193, cr_loss=0.3362, over 3148208.02 frames. ], batch size: 37, lr: 2.65e-03, grad_scale: 16.0
2024-09-25 18:16:44,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=802550.0, ans=0.2
2024-09-25 18:17:01,042 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=802596.6666666666, ans=0.2
2024-09-25 18:17:03,253 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.31 vs. limit=6.0
2024-09-25 18:17:20,815 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.18 vs. limit=15.0
2024-09-25 18:17:58,216 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.50 vs. limit=15.0
2024-09-25 18:18:02,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=802736.6666666666, ans=0.0
2024-09-25 18:18:06,652 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=10.20 vs. limit=10.0
2024-09-25 18:18:06,931 INFO [train.py:1198] (1/4) Epoch 45, batch 600, loss[loss=0.159, ctc_loss=0.09956, cr_loss=0.2973, over 17063.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1196, cr_loss=0.3363, over 3186275.19 frames. ], batch size: 39, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:18:12,746 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.61 vs. limit=15.0
2024-09-25 18:18:18,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=802783.3333333334, ans=0.1
2024-09-25 18:18:40,614 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:19:07,371 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.313e+02 1.405e+02 1.513e+02 2.621e+02, threshold=2.810e+02, percent-clipped=0.0
2024-09-25 18:19:14,658 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.22 vs. limit=15.0
2024-09-25 18:19:26,370 INFO [train.py:1198] (1/4) Epoch 45, batch 650, loss[loss=0.2212, ctc_loss=0.1453, cr_loss=0.3792, over 16978.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1196, cr_loss=0.336, over 3218278.04 frames. ], batch size: 53, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:19:36,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=803016.6666666666, ans=0.0
2024-09-25 18:19:36,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=803016.6666666666, ans=0.1
2024-09-25 18:19:36,648 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=6.30 vs. limit=15.0
2024-09-25 18:19:48,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=803063.3333333334, ans=0.0
2024-09-25 18:19:48,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=803063.3333333334, ans=0.0
2024-09-25 18:19:53,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=803063.3333333334, ans=0.125
2024-09-25 18:20:33,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=803203.3333333334, ans=0.1
2024-09-25 18:20:52,179 INFO [train.py:1198] (1/4) Epoch 45, batch 700, loss[loss=0.203, ctc_loss=0.1305, cr_loss=0.3623, over 17294.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.12, cr_loss=0.3364, over 3257898.07 frames. ], batch size: 51, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:21:11,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=803296.6666666666, ans=0.1
2024-09-25 18:21:24,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=803296.6666666666, ans=0.0
2024-09-25 18:21:30,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=803343.3333333334, ans=0.1
2024-09-25 18:21:41,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=803390.0, ans=0.125
2024-09-25 18:21:55,975 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.305e+02 1.379e+02 1.482e+02 2.055e+02, threshold=2.759e+02, percent-clipped=0.0
2024-09-25 18:22:18,092 INFO [train.py:1198] (1/4) Epoch 45, batch 750, loss[loss=0.1771, ctc_loss=0.1135, cr_loss=0.3177, over 17143.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1204, cr_loss=0.3371, over 3280118.86 frames. ], batch size: 48, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:22:22,359 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.36 vs. limit=15.0
2024-09-25 18:22:49,004 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:23:35,850 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:23:38,653 INFO [train.py:1198] (1/4) Epoch 45, batch 800, loss[loss=0.1765, ctc_loss=0.1122, cr_loss=0.3218, over 17015.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1204, cr_loss=0.3374, over 3297397.58 frames. ], batch size: 51, lr: 2.64e-03, grad_scale: 32.0
2024-09-25 18:23:42,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=803716.6666666666, ans=0.125
2024-09-25 18:23:58,059 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=803763.3333333334, ans=0.125
2024-09-25 18:24:02,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=803763.3333333334, ans=0.0
2024-09-25 18:24:15,501 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=803810.0, ans=0.0
2024-09-25 18:24:29,540 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.24 vs. limit=22.5
2024-09-25 18:24:30,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=803856.6666666666, ans=0.2
2024-09-25 18:24:30,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=803856.6666666666, ans=0.2
2024-09-25 18:24:39,374 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.279e+02 1.371e+02 1.474e+02 1.917e+02, threshold=2.742e+02, percent-clipped=0.0
2024-09-25 18:24:40,062 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.83 vs. limit=15.0
2024-09-25 18:24:58,299 INFO [train.py:1198] (1/4) Epoch 45, batch 850, loss[loss=0.1808, ctc_loss=0.1151, cr_loss=0.329, over 17221.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1203, cr_loss=0.3374, over 3324152.66 frames. ], batch size: 47, lr: 2.64e-03, grad_scale: 32.0
2024-09-25 18:25:15,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=803996.6666666666, ans=0.0
2024-09-25 18:25:39,043 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=6.299e-02
2024-09-25 18:26:20,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=804136.6666666666, ans=0.0
2024-09-25 18:26:22,075 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=804136.6666666666, ans=0.0
2024-09-25 18:26:26,523 INFO [train.py:1198] (1/4) Epoch 45, batch 900, loss[loss=0.1615, ctc_loss=0.1027, cr_loss=0.2938, over 16949.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1202, cr_loss=0.3374, over 3340020.42 frames. ], batch size: 42, lr: 2.64e-03, grad_scale: 32.0
2024-09-25 18:27:14,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=804276.6666666666, ans=0.125
2024-09-25 18:27:23,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=804323.3333333334, ans=0.0
2024-09-25 18:27:25,631 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.44 vs. limit=12.0
2024-09-25 18:27:29,591 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.299e+02 1.367e+02 1.436e+02 2.740e+02, threshold=2.735e+02, percent-clipped=0.0
2024-09-25 18:27:31,497 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.prob, batch_count=804370.0, ans=0.125
2024-09-25 18:27:48,683 INFO [train.py:1198] (1/4) Epoch 45, batch 950, loss[loss=0.2095, ctc_loss=0.1383, cr_loss=0.3559, over 11461.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1194, cr_loss=0.3353, over 3337296.04 frames. ], batch size: 124, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:27:57,561 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.31 vs. limit=15.0
2024-09-25 18:27:58,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=804416.6666666666, ans=0.125
2024-09-25 18:27:59,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=804416.6666666666, ans=0.1
2024-09-25 18:28:04,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=804463.3333333334, ans=0.2
2024-09-25 18:28:09,345 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=804463.3333333334, ans=0.0
2024-09-25 18:28:11,705 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.32 vs. limit=15.0
2024-09-25 18:28:17,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=804463.3333333334, ans=0.125
2024-09-25 18:28:34,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=804556.6666666666, ans=0.025
2024-09-25 18:28:41,039 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:29:07,946 INFO [train.py:1198] (1/4) Epoch 45, batch 1000, loss[loss=0.1957, ctc_loss=0.1231, cr_loss=0.3629, over 17306.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.12, cr_loss=0.3366, over 3340247.17 frames. ], batch size: 49, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:29:41,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=804743.3333333334, ans=0.125
2024-09-25 18:29:51,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=804743.3333333334, ans=0.0
2024-09-25 18:29:56,870 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.25 vs. limit=15.0
2024-09-25 18:30:08,065 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.44 vs. limit=22.5
2024-09-25 18:30:13,227 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.285e+02 1.391e+02 1.486e+02 1.776e+02, threshold=2.782e+02, percent-clipped=0.0
2024-09-25 18:30:21,956 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.41 vs. limit=15.0
2024-09-25 18:30:33,440 INFO [train.py:1198] (1/4) Epoch 45, batch 1050, loss[loss=0.1952, ctc_loss=0.126, cr_loss=0.3458, over 16853.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1208, cr_loss=0.3379, over 3340888.34 frames. ], batch size: 58, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:30:50,330 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.17 vs. limit=15.0
2024-09-25 18:31:30,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=805023.3333333334, ans=0.125
2024-09-25 18:31:35,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=805023.3333333334, ans=0.0
2024-09-25 18:31:58,555 INFO [train.py:1198] (1/4) Epoch 45, batch 1100, loss[loss=0.1872, ctc_loss=0.1191, cr_loss=0.3406, over 17308.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3383, over 3341553.68 frames. ], batch size: 49, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:32:22,618 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=805163.3333333334, ans=0.125
2024-09-25 18:32:46,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=805256.6666666666, ans=0.125
2024-09-25 18:32:56,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=805256.6666666666, ans=0.125
2024-09-25 18:33:00,831 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.299e+02 1.379e+02 1.493e+02 1.866e+02, threshold=2.759e+02, percent-clipped=0.0
2024-09-25 18:33:04,796 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.71 vs. limit=22.5
2024-09-25 18:33:18,415 INFO [train.py:1198] (1/4) Epoch 45, batch 1150, loss[loss=0.1894, ctc_loss=0.1211, cr_loss=0.3419, over 17023.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.121, cr_loss=0.3384, over 3352886.92 frames. ], batch size: 44, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:33:26,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=805350.0, ans=0.0
2024-09-25 18:33:36,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=805396.6666666666, ans=0.0
2024-09-25 18:34:09,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=805490.0, ans=0.125
2024-09-25 18:34:15,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.52 vs. limit=15.0
2024-09-25 18:34:38,827 INFO [train.py:1198] (1/4) Epoch 45, batch 1200, loss[loss=0.1689, ctc_loss=0.1057, cr_loss=0.3161, over 17034.00 frames. ], tot_loss[loss=0.188, ctc_loss=0.1204, cr_loss=0.3379, over 3363961.70 frames. ], batch size: 39, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:34:45,959 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.24 vs. limit=6.0
2024-09-25 18:34:54,119 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.07 vs. limit=15.0
2024-09-25 18:35:11,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=805676.6666666666, ans=0.125
2024-09-25 18:35:15,150 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.prob, batch_count=805676.6666666666, ans=0.125
2024-09-25 18:35:16,881 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=805676.6666666666, ans=0.125
2024-09-25 18:35:27,304 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=805676.6666666666, ans=0.0
2024-09-25 18:35:32,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=805723.3333333334, ans=0.05
2024-09-25 18:35:46,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=805770.0, ans=0.0
2024-09-25 18:35:47,625 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.307e+02 1.404e+02 1.482e+02 2.313e+02, threshold=2.807e+02, percent-clipped=0.0
2024-09-25 18:36:06,488 INFO [train.py:1198] (1/4) Epoch 45, batch 1250, loss[loss=0.1938, ctc_loss=0.1238, cr_loss=0.3499, over 16515.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1206, cr_loss=0.3381, over 3362027.38 frames. ], batch size: 66, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:36:08,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.86 vs. limit=22.5
2024-09-25 18:36:52,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.46 vs. limit=15.0
2024-09-25 18:36:57,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=805956.6666666666, ans=0.0
2024-09-25 18:36:58,863 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=805956.6666666666, ans=0.125
2024-09-25 18:37:10,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=805956.6666666666, ans=0.2
2024-09-25 18:37:16,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=806003.3333333334, ans=0.125
2024-09-25 18:37:21,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=806003.3333333334, ans=0.0
2024-09-25 18:37:29,080 INFO [train.py:1198] (1/4) Epoch 45, batch 1300, loss[loss=0.2039, ctc_loss=0.1323, cr_loss=0.3579, over 17307.00 frames. ], tot_loss[loss=0.1886, ctc_loss=0.1209, cr_loss=0.3386, over 3358454.90 frames. ], batch size: 49, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:37:53,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=806096.6666666666, ans=0.125
2024-09-25 18:38:01,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=806143.3333333334, ans=0.125
2024-09-25 18:38:17,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=806190.0, ans=0.5
2024-09-25 18:38:31,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=806236.6666666666, ans=0.125
2024-09-25 18:38:32,667 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.317e+02 1.369e+02 1.447e+02 1.900e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-25 18:38:34,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=806236.6666666666, ans=0.1
2024-09-25 18:38:48,676 INFO [train.py:1198] (1/4) Epoch 45, batch 1350, loss[loss=0.1809, ctc_loss=0.1132, cr_loss=0.3386, over 17035.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1202, cr_loss=0.3375, over 3366028.30 frames. ], batch size: 56, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:38:53,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=806283.3333333334, ans=0.0
2024-09-25 18:38:59,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=806283.3333333334, ans=0.0
2024-09-25 18:39:05,667 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.09 vs. limit=15.0
2024-09-25 18:39:19,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=806376.6666666666, ans=0.2
2024-09-25 18:39:39,507 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.19 vs. limit=15.0
2024-09-25 18:39:50,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.16 vs. limit=22.5
2024-09-25 18:39:57,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=806470.0, ans=0.125
2024-09-25 18:40:14,349 INFO [train.py:1198] (1/4) Epoch 45, batch 1400, loss[loss=0.2124, ctc_loss=0.1349, cr_loss=0.3876, over 17289.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.12, cr_loss=0.3374, over 3362350.32 frames. ], batch size: 49, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:40:46,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=806610.0, ans=0.2
2024-09-25 18:41:00,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_abs, batch_count=806610.0, ans=0.5
2024-09-25 18:41:21,112 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.028e+02 1.293e+02 1.370e+02 1.466e+02 2.131e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-25 18:41:37,441 INFO [train.py:1198] (1/4) Epoch 45, batch 1450, loss[loss=0.1889, ctc_loss=0.1203, cr_loss=0.3433, over 17027.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1197, cr_loss=0.3364, over 3358443.73 frames. ], batch size: 51, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:41:37,766 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=806750.0, ans=0.125
2024-09-25 18:41:48,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=806750.0, ans=0.125
2024-09-25 18:41:52,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=806750.0, ans=0.2
2024-09-25 18:42:41,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=806890.0, ans=0.0
2024-09-25 18:42:51,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=806936.6666666666, ans=0.125
2024-09-25 18:43:00,438 INFO [train.py:1198] (1/4) Epoch 45, batch 1500, loss[loss=0.1887, ctc_loss=0.1199, cr_loss=0.3441, over 17028.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1191, cr_loss=0.3354, over 3360040.57 frames. ], batch size: 44, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:43:04,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=806983.3333333334, ans=0.125
2024-09-25 18:43:08,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=806983.3333333334, ans=0.0
2024-09-25 18:43:15,686 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=10.67 vs. limit=15.0
2024-09-25 18:44:01,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=807123.3333333334, ans=0.125
2024-09-25 18:44:04,286 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.307e+02 1.374e+02 1.496e+02 1.969e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-25 18:44:07,952 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=807170.0, ans=0.0
2024-09-25 18:44:09,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=807170.0, ans=0.125
2024-09-25 18:44:20,507 INFO [train.py:1198] (1/4) Epoch 45, batch 1550, loss[loss=0.248, ctc_loss=0.168, cr_loss=0.3999, over 11445.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1188, cr_loss=0.3347, over 3354481.52 frames. ], batch size: 123, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:44:30,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=807216.6666666666, ans=0.0
2024-09-25 18:45:15,758 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:45:23,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=807356.6666666666, ans=0.125
2024-09-25 18:45:23,457 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=807356.6666666666, ans=0.0
2024-09-25 18:45:39,912 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.68 vs. limit=15.0
2024-09-25 18:45:44,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=807450.0, ans=0.125
2024-09-25 18:45:45,593 INFO [train.py:1198] (1/4) Epoch 45, batch 1600, loss[loss=0.2145, ctc_loss=0.1424, cr_loss=0.3602, over 11623.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1184, cr_loss=0.3339, over 3353949.56 frames. ], batch size: 124, lr: 2.64e-03, grad_scale: 32.0
2024-09-25 18:46:50,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.51 vs. limit=15.0
2024-09-25 18:46:56,353 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.296e+02 1.378e+02 1.483e+02 2.434e+02, threshold=2.757e+02, percent-clipped=0.0
2024-09-25 18:47:10,854 INFO [train.py:1198] (1/4) Epoch 45, batch 1650, loss[loss=0.1581, ctc_loss=0.0992, cr_loss=0.2944, over 16967.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1186, cr_loss=0.3347, over 3352051.91 frames. ], batch size: 42, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:47:46,708 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.08 vs. limit=22.5
2024-09-25 18:47:54,412 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:47:59,477 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=9.33 vs. limit=12.0
2024-09-25 18:48:28,388 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.04 vs. limit=15.0
2024-09-25 18:48:31,083 INFO [train.py:1198] (1/4) Epoch 45, batch 1700, loss[loss=0.148, ctc_loss=0.09283, cr_loss=0.276, over 17283.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1188, cr_loss=0.3346, over 3346952.67 frames. ], batch size: 42, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:48:39,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=807916.6666666666, ans=0.05
2024-09-25 18:48:43,281 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=8.16 vs. limit=10.0
2024-09-25 18:48:52,140 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=807963.3333333334, ans=0.2
2024-09-25 18:48:58,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=807963.3333333334, ans=0.125
2024-09-25 18:49:23,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=808056.6666666666, ans=0.125
2024-09-25 18:49:36,102 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.276e+02 1.351e+02 1.443e+02 2.943e+02, threshold=2.702e+02, percent-clipped=1.0
2024-09-25 18:49:39,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=808103.3333333334, ans=0.125
2024-09-25 18:49:50,471 INFO [train.py:1198] (1/4) Epoch 45, batch 1750, loss[loss=0.1772, ctc_loss=0.1139, cr_loss=0.3165, over 17281.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1199, cr_loss=0.3371, over 3356180.83 frames. ], batch size: 51, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:50:01,600 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten.whitening_limit, batch_count=808150.0, ans=15.0
2024-09-25 18:50:50,490 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:51:18,075 INFO [train.py:1198] (1/4) Epoch 45, batch 1800, loss[loss=0.159, ctc_loss=0.09924, cr_loss=0.2987, over 17272.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1198, cr_loss=0.3367, over 3361338.28 frames. ], batch size: 42, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:51:42,831 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.69 vs. limit=15.0
2024-09-25 18:52:09,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=808523.3333333334, ans=0.07
2024-09-25 18:52:22,189 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.87 vs. limit=6.0
2024-09-25 18:52:23,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=808570.0, ans=0.2
2024-09-25 18:52:24,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=808570.0, ans=0.125
2024-09-25 18:52:26,148 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.286e+02 1.376e+02 1.494e+02 2.508e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-25 18:52:26,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=808570.0, ans=0.125
2024-09-25 18:52:31,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=808570.0, ans=0.2
2024-09-25 18:52:31,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=808570.0, ans=0.0
2024-09-25 18:52:37,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=808570.0, ans=0.025
2024-09-25 18:52:40,620 INFO [train.py:1198] (1/4) Epoch 45, batch 1850, loss[loss=0.1898, ctc_loss=0.1246, cr_loss=0.3262, over 17207.00 frames. ], tot_loss[loss=0.187, ctc_loss=0.1197, cr_loss=0.3368, over 3369233.06 frames. ], batch size: 47, lr: 2.64e-03, grad_scale: 16.0
2024-09-25 18:52:50,491 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=808616.6666666666, ans=0.1
2024-09-25 18:52:58,538 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=808663.3333333334, ans=0.0
2024-09-25 18:53:32,143 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=808756.6666666666, ans=0.125
2024-09-25 18:53:40,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=808756.6666666666, ans=10.0
2024-09-25 18:53:43,428 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=808803.3333333334, ans=0.125
2024-09-25 18:53:48,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=808803.3333333334, ans=0.0
2024-09-25 18:54:00,630 INFO [train.py:1198] (1/4) Epoch 45, batch 1900, loss[loss=0.2059, ctc_loss=0.1342, cr_loss=0.3587, over 16037.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.12, cr_loss=0.3369, over 3364161.14 frames. ], batch size: 75, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 18:54:07,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=808850.0, ans=0.2
2024-09-25 18:54:14,046 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.46 vs. limit=15.0
2024-09-25 18:54:14,590 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=4.94 vs. limit=12.0
2024-09-25 18:54:43,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=512, metric=12.50 vs. limit=22.5
2024-09-25 18:55:07,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=808990.0, ans=0.0
2024-09-25 18:55:12,117 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.309e+02 1.390e+02 1.516e+02 1.973e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-25 18:55:14,587 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.82 vs. limit=6.0
2024-09-25 18:55:26,421 INFO [train.py:1198] (1/4) Epoch 45, batch 1950, loss[loss=0.1802, ctc_loss=0.113, cr_loss=0.3364, over 17284.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1199, cr_loss=0.3369, over 3367850.42 frames. ], batch size: 42, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 18:55:42,770 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=809130.0, ans=0.0
2024-09-25 18:55:44,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=809130.0, ans=0.1
2024-09-25 18:55:52,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=809130.0, ans=0.1
2024-09-25 18:56:11,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=809176.6666666666, ans=0.0
2024-09-25 18:56:23,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.51 vs. limit=6.0
2024-09-25 18:56:49,742 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.94 vs. limit=22.5
2024-09-25 18:56:52,060 INFO [train.py:1198] (1/4) Epoch 45, batch 2000, loss[loss=0.1695, ctc_loss=0.1082, cr_loss=0.3062, over 17153.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1201, cr_loss=0.3372, over 3368832.62 frames. ], batch size: 41, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 18:57:23,079 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten.whitening_limit, batch_count=809410.0, ans=15.0
2024-09-25 18:57:30,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=809410.0, ans=0.5
2024-09-25 18:57:46,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=809456.6666666666, ans=0.125
2024-09-25 18:57:57,598 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.313e+02 1.395e+02 1.516e+02 2.306e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-25 18:58:12,072 INFO [train.py:1198] (1/4) Epoch 45, batch 2050, loss[loss=0.1557, ctc_loss=0.09857, cr_loss=0.2859, over 17089.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1192, cr_loss=0.3352, over 3373799.35 frames. ], batch size: 43, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 18:58:28,453 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 18:58:52,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=809643.3333333334, ans=0.0
2024-09-25 18:59:08,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=809690.0, ans=0.2
2024-09-25 18:59:18,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=809736.6666666666, ans=0.1
2024-09-25 18:59:18,676 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.42 vs. limit=15.0
2024-09-25 18:59:29,835 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.72 vs. limit=12.0
2024-09-25 18:59:32,218 INFO [train.py:1198] (1/4) Epoch 45, batch 2100, loss[loss=0.1758, ctc_loss=0.1111, cr_loss=0.3236, over 17007.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1186, cr_loss=0.3343, over 3378697.33 frames. ], batch size: 44, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 18:59:43,772 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=809783.3333333334, ans=0.2
2024-09-25 19:00:02,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=809830.0, ans=0.0
2024-09-25 19:00:10,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=809876.6666666666, ans=0.125
2024-09-25 19:00:19,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=809876.6666666666, ans=0.125
2024-09-25 19:00:19,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=809876.6666666666, ans=0.1
2024-09-25 19:00:21,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=809923.3333333334, ans=0.0
2024-09-25 19:00:24,915 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=3.84 vs. limit=15.0
2024-09-25 19:00:35,852 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=809923.3333333334, ans=0.0
2024-09-25 19:00:40,214 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.255e+02 1.338e+02 1.412e+02 1.754e+02, threshold=2.676e+02, percent-clipped=0.0
2024-09-25 19:00:48,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=809970.0, ans=0.125
2024-09-25 19:00:57,180 INFO [train.py:1198] (1/4) Epoch 45, batch 2150, loss[loss=0.1933, ctc_loss=0.1234, cr_loss=0.3493, over 17278.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1193, cr_loss=0.3354, over 3364009.28 frames. ], batch size: 46, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:01:11,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=810063.3333333334, ans=0.125
2024-09-25 19:01:12,087 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.45 vs. limit=12.0
2024-09-25 19:01:23,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=810063.3333333334, ans=0.05
2024-09-25 19:01:28,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=810063.3333333334, ans=22.5
2024-09-25 19:01:46,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer2.prob, batch_count=810156.6666666666, ans=0.125
2024-09-25 19:01:56,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=810156.6666666666, ans=0.125
2024-09-25 19:02:01,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=810156.6666666666, ans=0.5
2024-09-25 19:02:20,041 INFO [train.py:1198] (1/4) Epoch 45, batch 2200, loss[loss=0.2318, ctc_loss=0.1485, cr_loss=0.4165, over 16863.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1197, cr_loss=0.3356, over 3346512.01 frames. ], batch size: 58, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:02:31,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=810250.0, ans=0.2
2024-09-25 19:02:34,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=810296.6666666666, ans=0.1
2024-09-25 19:02:39,832 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.22 vs. limit=12.0
2024-09-25 19:02:59,576 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.66 vs. limit=5.0
2024-09-25 19:03:08,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=810390.0, ans=0.125
2024-09-25 19:03:09,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=810390.0, ans=0.025
2024-09-25 19:03:25,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=810436.6666666666, ans=0.125
2024-09-25 19:03:27,150 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.297e+02 1.368e+02 1.487e+02 2.152e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 19:03:29,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=810436.6666666666, ans=0.125
2024-09-25 19:03:40,105 INFO [train.py:1198] (1/4) Epoch 45, batch 2250, loss[loss=0.1494, ctc_loss=0.09418, cr_loss=0.2762, over 16788.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1196, cr_loss=0.335, over 3347857.30 frames. ], batch size: 37, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:03:43,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=810483.3333333334, ans=0.125
2024-09-25 19:04:28,420 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=810623.3333333334, ans=0.1
2024-09-25 19:05:02,934 INFO [train.py:1198] (1/4) Epoch 45, batch 2300, loss[loss=0.1756, ctc_loss=0.111, cr_loss=0.3231, over 17063.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1201, cr_loss=0.3359, over 3341368.02 frames. ], batch size: 46, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:05:51,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=810856.6666666666, ans=0.0
2024-09-25 19:06:09,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=810903.3333333334, ans=0.025
2024-09-25 19:06:10,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=810903.3333333334, ans=0.1
2024-09-25 19:06:12,003 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.306e+02 1.375e+02 1.455e+02 1.967e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-25 19:06:12,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=810903.3333333334, ans=0.035
2024-09-25 19:06:24,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=810903.3333333334, ans=0.1
2024-09-25 19:06:27,347 INFO [train.py:1198] (1/4) Epoch 45, batch 2350, loss[loss=0.1979, ctc_loss=0.128, cr_loss=0.3496, over 16807.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1203, cr_loss=0.3366, over 3349038.40 frames. ], batch size: 61, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:06:49,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=810996.6666666666, ans=0.125
2024-09-25 19:07:12,365 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=811043.3333333334, ans=0.2
2024-09-25 19:07:24,166 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.62 vs. limit=6.0
2024-09-25 19:07:24,893 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=811090.0, ans=0.025
2024-09-25 19:07:46,977 INFO [train.py:1198] (1/4) Epoch 45, batch 2400, loss[loss=0.1899, ctc_loss=0.1207, cr_loss=0.346, over 17035.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1203, cr_loss=0.3368, over 3358418.23 frames. ], batch size: 56, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:07:48,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=811183.3333333334, ans=0.0
2024-09-25 19:07:48,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=811183.3333333334, ans=0.125
2024-09-25 19:07:55,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=811183.3333333334, ans=0.125
2024-09-25 19:08:43,289 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=811323.3333333334, ans=0.125
2024-09-25 19:08:51,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=811370.0, ans=0.025
2024-09-25 19:08:55,592 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.054e+02 1.292e+02 1.391e+02 1.494e+02 2.070e+02, threshold=2.781e+02, percent-clipped=0.0
2024-09-25 19:09:02,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=811370.0, ans=0.0
2024-09-25 19:09:07,045 INFO [train.py:1198] (1/4) Epoch 45, batch 2450, loss[loss=0.2177, ctc_loss=0.1394, cr_loss=0.3915, over 17326.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1204, cr_loss=0.3369, over 3358037.90 frames. ], batch size: 51, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:09:18,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=811416.6666666666, ans=0.0
2024-09-25 19:09:33,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=811463.3333333334, ans=0.125
2024-09-25 19:09:45,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=811510.0, ans=0.125
2024-09-25 19:10:05,299 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=811556.6666666666, ans=0.125
2024-09-25 19:10:32,451 INFO [train.py:1198] (1/4) Epoch 45, batch 2500, loss[loss=0.1949, ctc_loss=0.1236, cr_loss=0.3565, over 17063.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1204, cr_loss=0.337, over 3352583.93 frames. ], batch size: 46, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:10:53,554 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.05 vs. limit=10.0
2024-09-25 19:11:32,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=811790.0, ans=0.0
2024-09-25 19:11:46,431 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.307e+02 1.405e+02 1.520e+02 3.997e+02, threshold=2.811e+02, percent-clipped=1.0
2024-09-25 19:11:57,470 INFO [train.py:1198] (1/4) Epoch 45, batch 2550, loss[loss=0.2007, ctc_loss=0.1295, cr_loss=0.356, over 17088.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1208, cr_loss=0.338, over 3346984.14 frames. ], batch size: 49, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:12:46,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=812023.3333333334, ans=0.125
2024-09-25 19:12:49,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=812023.3333333334, ans=0.125
2024-09-25 19:12:54,684 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.30 vs. limit=15.0
2024-09-25 19:13:18,002 INFO [train.py:1198] (1/4) Epoch 45, batch 2600, loss[loss=0.1894, ctc_loss=0.1191, cr_loss=0.3518, over 17023.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1211, cr_loss=0.3391, over 3350911.49 frames. ], batch size: 52, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:13:23,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=812116.6666666666, ans=0.0
2024-09-25 19:13:50,437 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=812210.0, ans=0.025
2024-09-25 19:14:00,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=812210.0, ans=0.0
2024-09-25 19:14:04,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=812256.6666666666, ans=0.0
2024-09-25 19:14:06,229 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=812256.6666666666, ans=0.1
2024-09-25 19:14:26,619 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.298e+02 1.365e+02 1.508e+02 3.211e+02, threshold=2.730e+02, percent-clipped=1.0
2024-09-25 19:14:30,094 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:14:33,188 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=812303.3333333334, ans=0.125
2024-09-25 19:14:37,603 INFO [train.py:1198] (1/4) Epoch 45, batch 2650, loss[loss=0.1617, ctc_loss=0.1017, cr_loss=0.2999, over 16942.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1212, cr_loss=0.3393, over 3348777.15 frames. ], batch size: 42, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:14:53,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.23 vs. limit=15.0
2024-09-25 19:15:20,957 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=4.75 vs. limit=15.0
2024-09-25 19:15:26,616 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:15:31,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=812490.0, ans=0.125
2024-09-25 19:15:56,023 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=812536.6666666666, ans=0.125
2024-09-25 19:16:05,228 INFO [train.py:1198] (1/4) Epoch 45, batch 2700, loss[loss=0.1681, ctc_loss=0.1074, cr_loss=0.3035, over 17287.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.121, cr_loss=0.3389, over 3352835.23 frames. ], batch size: 49, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:16:05,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=812583.3333333334, ans=0.0
2024-09-25 19:16:11,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=812583.3333333334, ans=0.2
2024-09-25 19:17:00,673 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=812723.3333333334, ans=0.1
2024-09-25 19:17:08,690 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:17:08,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=812723.3333333334, ans=0.1
2024-09-25 19:17:16,507 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.332e+02 1.395e+02 1.480e+02 1.969e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-25 19:17:23,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=812770.0, ans=0.125
2024-09-25 19:17:27,700 INFO [train.py:1198] (1/4) Epoch 45, batch 2750, loss[loss=0.1976, ctc_loss=0.1296, cr_loss=0.34, over 16106.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1213, cr_loss=0.3389, over 3351200.37 frames. ], batch size: 74, lr: 2.63e-03, grad_scale: 16.0
2024-09-25 19:17:31,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=812816.6666666666, ans=0.1
2024-09-25 19:17:51,640 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=812863.3333333334, ans=0.125
2024-09-25 19:17:55,633 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.74 vs. limit=6.0
2024-09-25 19:17:58,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=812910.0, ans=0.0
2024-09-25 19:18:01,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=812910.0, ans=0.125
2024-09-25 19:18:19,364 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=18.46 vs. limit=22.5
2024-09-25 19:18:44,814 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=813003.3333333334, ans=0.0
2024-09-25 19:18:47,783 INFO [train.py:1198] (1/4) Epoch 45, batch 2800, loss[loss=0.207, ctc_loss=0.1334, cr_loss=0.3679, over 17228.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1213, cr_loss=0.3387, over 3348991.88 frames. ], batch size: 55, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:18:54,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=813050.0, ans=0.0
2024-09-25 19:18:58,313 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.07 vs. limit=10.0
2024-09-25 19:19:04,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=813096.6666666666, ans=0.09899494936611666
2024-09-25 19:19:07,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=813096.6666666666, ans=0.0
2024-09-25 19:19:18,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=813143.3333333334, ans=0.2
2024-09-25 19:20:01,231 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.322e+02 1.415e+02 1.536e+02 2.357e+02, threshold=2.830e+02, percent-clipped=0.0
2024-09-25 19:20:12,478 INFO [train.py:1198] (1/4) Epoch 45, batch 2850, loss[loss=0.1967, ctc_loss=0.1288, cr_loss=0.3393, over 17213.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1209, cr_loss=0.338, over 3345302.47 frames. ], batch size: 50, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:20:15,938 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=813283.3333333334, ans=0.1
2024-09-25 19:21:00,448 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.14 vs. limit=15.0
2024-09-25 19:21:04,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=813423.3333333334, ans=0.0
2024-09-25 19:21:13,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=813423.3333333334, ans=0.125
2024-09-25 19:21:14,596 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.50 vs. limit=12.0
2024-09-25 19:21:32,234 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.59 vs. limit=15.0
2024-09-25 19:21:33,534 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.79 vs. limit=15.0
2024-09-25 19:21:37,529 INFO [train.py:1198] (1/4) Epoch 45, batch 2900, loss[loss=0.1797, ctc_loss=0.1147, cr_loss=0.3252, over 17001.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1199, cr_loss=0.3368, over 3353703.20 frames. ], batch size: 51, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:21:44,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=813516.6666666666, ans=0.025
2024-09-25 19:21:52,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=813563.3333333334, ans=0.1
2024-09-25 19:22:07,253 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=8.72 vs. limit=15.0
2024-09-25 19:22:09,670 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=813610.0, ans=0.0
2024-09-25 19:22:13,250 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.14 vs. limit=10.0
2024-09-25 19:22:28,139 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.69 vs. limit=6.0
2024-09-25 19:22:29,425 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.97 vs. limit=6.0
2024-09-25 19:22:46,286 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.314e+02 1.430e+02 1.564e+02 2.364e+02, threshold=2.859e+02, percent-clipped=0.0
2024-09-25 19:22:51,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=813703.3333333334, ans=0.125
2024-09-25 19:22:57,460 INFO [train.py:1198] (1/4) Epoch 45, batch 2950, loss[loss=0.207, ctc_loss=0.1353, cr_loss=0.3582, over 17066.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1196, cr_loss=0.336, over 3349655.16 frames. ], batch size: 56, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:23:05,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=813750.0, ans=0.025
2024-09-25 19:23:33,135 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=813843.3333333334, ans=0.1
2024-09-25 19:24:09,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=813936.6666666666, ans=0.125
2024-09-25 19:24:16,611 INFO [train.py:1198] (1/4) Epoch 45, batch 3000, loss[loss=0.2008, ctc_loss=0.1356, cr_loss=0.3261, over 11233.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1202, cr_loss=0.3368, over 3336479.13 frames. ], batch size: 123, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:24:16,612 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 19:24:32,374 INFO [train.py:1230] (1/4) Epoch 45, validation: loss=0.03541, ctc_loss=0.03541, cr_loss=1.054e-14, over 944034.00 frames. 
2024-09-25 19:24:32,375 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 19:24:48,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=814030.0, ans=0.0
2024-09-25 19:24:51,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=814030.0, ans=0.0
2024-09-25 19:24:53,642 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.81 vs. limit=15.0
2024-09-25 19:24:54,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer_ff3.min_abs, batch_count=814030.0, ans=0.2
2024-09-25 19:24:56,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=814030.0, ans=0.125
2024-09-25 19:25:00,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=814030.0, ans=0.125
2024-09-25 19:25:02,957 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=18.20 vs. limit=22.5
2024-09-25 19:25:27,073 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.10 vs. limit=15.0
2024-09-25 19:25:32,855 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=814123.3333333334, ans=0.125
2024-09-25 19:25:33,174 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=9.10 vs. limit=15.0
2024-09-25 19:25:44,657 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.310e+02 1.382e+02 1.500e+02 2.246e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-25 19:25:44,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=814170.0, ans=0.125
2024-09-25 19:25:55,842 INFO [train.py:1198] (1/4) Epoch 45, batch 3050, loss[loss=0.1598, ctc_loss=0.1, cr_loss=0.2987, over 17277.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1205, cr_loss=0.3376, over 3339523.73 frames. ], batch size: 42, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:26:10,104 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=814263.3333333334, ans=0.95
2024-09-25 19:26:12,067 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.63 vs. limit=12.0
2024-09-25 19:26:31,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=814310.0, ans=0.125
2024-09-25 19:26:53,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=814356.6666666666, ans=0.125
2024-09-25 19:27:13,877 INFO [train.py:1198] (1/4) Epoch 45, batch 3100, loss[loss=0.1854, ctc_loss=0.1207, cr_loss=0.3236, over 15965.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1197, cr_loss=0.336, over 3352188.74 frames. ], batch size: 74, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:27:18,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=814450.0, ans=0.125
2024-09-25 19:27:19,394 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.99 vs. limit=15.0
2024-09-25 19:28:05,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=814590.0, ans=0.125
2024-09-25 19:28:19,425 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=814636.6666666666, ans=0.0
2024-09-25 19:28:23,832 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.327e+02 1.418e+02 1.524e+02 1.856e+02, threshold=2.835e+02, percent-clipped=0.0
2024-09-25 19:28:36,920 INFO [train.py:1198] (1/4) Epoch 45, batch 3150, loss[loss=0.1553, ctc_loss=0.09695, cr_loss=0.2917, over 17202.00 frames. ], tot_loss[loss=0.187, ctc_loss=0.1197, cr_loss=0.3367, over 3359613.87 frames. ], batch size: 41, lr: 2.63e-03, grad_scale: 32.0
2024-09-25 19:28:41,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=814683.3333333334, ans=0.125
2024-09-25 19:28:43,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=814683.3333333334, ans=0.025
2024-09-25 19:29:03,883 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=814730.0, ans=0.125
2024-09-25 19:29:29,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=814823.3333333334, ans=0.05
2024-09-25 19:29:37,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=814823.3333333334, ans=0.0
2024-09-25 19:29:55,901 INFO [train.py:1198] (1/4) Epoch 45, batch 3200, loss[loss=0.1863, ctc_loss=0.1189, cr_loss=0.3369, over 17093.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1195, cr_loss=0.3362, over 3357032.24 frames. ], batch size: 49, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:29:56,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=814916.6666666666, ans=0.1
2024-09-25 19:29:59,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=814916.6666666666, ans=0.025
2024-09-25 19:30:24,439 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:30:27,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=815010.0, ans=0.04949747468305833
2024-09-25 19:31:03,190 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.319e+02 1.396e+02 1.488e+02 2.041e+02, threshold=2.793e+02, percent-clipped=0.0
2024-09-25 19:31:14,382 INFO [train.py:1198] (1/4) Epoch 45, batch 3250, loss[loss=0.197, ctc_loss=0.1268, cr_loss=0.3512, over 16051.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.119, cr_loss=0.3356, over 3367858.80 frames. ], batch size: 74, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:31:29,012 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=815196.6666666666, ans=0.125
2024-09-25 19:31:34,485 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.24 vs. limit=12.0
2024-09-25 19:31:55,704 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:32:19,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=815336.6666666666, ans=0.125
2024-09-25 19:32:22,388 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=815336.6666666666, ans=0.025
2024-09-25 19:32:33,038 INFO [train.py:1198] (1/4) Epoch 45, batch 3300, loss[loss=0.2171, ctc_loss=0.1406, cr_loss=0.3828, over 17143.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.119, cr_loss=0.3356, over 3367976.91 frames. ], batch size: 48, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:32:33,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=815383.3333333334, ans=0.125
2024-09-25 19:32:37,277 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=12.22 vs. limit=15.0
2024-09-25 19:32:45,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=815383.3333333334, ans=0.0
2024-09-25 19:32:47,935 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.15 vs. limit=12.0
2024-09-25 19:33:03,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=815476.6666666666, ans=0.125
2024-09-25 19:33:18,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=815523.3333333334, ans=0.125
2024-09-25 19:33:28,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=815523.3333333334, ans=0.1
2024-09-25 19:33:40,404 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.311e+02 1.407e+02 1.511e+02 1.886e+02, threshold=2.815e+02, percent-clipped=0.0
2024-09-25 19:33:46,918 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=815570.0, ans=0.1
2024-09-25 19:33:47,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=815570.0, ans=0.1
2024-09-25 19:33:51,418 INFO [train.py:1198] (1/4) Epoch 45, batch 3350, loss[loss=0.1477, ctc_loss=0.0912, cr_loss=0.2824, over 16284.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1192, cr_loss=0.3351, over 3352198.42 frames. ], batch size: 36, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:34:04,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.89 vs. limit=6.0
2024-09-25 19:34:10,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=815663.3333333334, ans=0.0
2024-09-25 19:34:29,607 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=815710.0, ans=0.0
2024-09-25 19:34:34,453 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=815710.0, ans=0.0
2024-09-25 19:34:41,234 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=16.44 vs. limit=22.5
2024-09-25 19:34:46,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=815756.6666666666, ans=0.125
2024-09-25 19:34:56,507 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=815803.3333333334, ans=0.0
2024-09-25 19:35:10,347 INFO [train.py:1198] (1/4) Epoch 45, batch 3400, loss[loss=0.2012, ctc_loss=0.1303, cr_loss=0.3541, over 16907.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1188, cr_loss=0.3337, over 3346200.88 frames. ], batch size: 58, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:35:38,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=815896.6666666666, ans=0.125
2024-09-25 19:36:02,944 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=815990.0, ans=0.125
2024-09-25 19:36:19,819 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.333e+02 1.429e+02 1.523e+02 2.294e+02, threshold=2.857e+02, percent-clipped=0.0
2024-09-25 19:36:30,574 INFO [train.py:1198] (1/4) Epoch 45, batch 3450, loss[loss=0.1713, ctc_loss=0.1061, cr_loss=0.3259, over 17023.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.119, cr_loss=0.334, over 3350737.22 frames. ], batch size: 44, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:37:01,901 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.max_abs, batch_count=816176.6666666666, ans=10.0
2024-09-25 19:37:05,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=816176.6666666666, ans=0.0
2024-09-25 19:37:08,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=816176.6666666666, ans=0.0
2024-09-25 19:37:27,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=816223.3333333334, ans=0.125
2024-09-25 19:37:50,525 INFO [train.py:1198] (1/4) Epoch 45, batch 3500, loss[loss=0.1759, ctc_loss=0.1107, cr_loss=0.326, over 17105.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1187, cr_loss=0.3339, over 3354482.42 frames. ], batch size: 49, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:38:00,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=816316.6666666666, ans=0.125
2024-09-25 19:38:09,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=816363.3333333334, ans=0.125
2024-09-25 19:38:12,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=816363.3333333334, ans=0.1
2024-09-25 19:38:23,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=816410.0, ans=0.2
2024-09-25 19:38:25,095 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=816410.0, ans=0.0
2024-09-25 19:38:47,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=816456.6666666666, ans=0.125
2024-09-25 19:38:50,622 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.32 vs. limit=15.0
2024-09-25 19:38:55,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=816503.3333333334, ans=0.025
2024-09-25 19:38:56,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=816503.3333333334, ans=0.125
2024-09-25 19:38:59,811 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=816503.3333333334, ans=0.025
2024-09-25 19:39:01,034 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.307e+02 1.374e+02 1.461e+02 1.986e+02, threshold=2.748e+02, percent-clipped=0.0
2024-09-25 19:39:10,423 INFO [train.py:1198] (1/4) Epoch 45, batch 3550, loss[loss=0.182, ctc_loss=0.1176, cr_loss=0.3221, over 17079.00 frames. ], tot_loss[loss=0.1853, ctc_loss=0.1187, cr_loss=0.3334, over 3358197.54 frames. ], batch size: 43, lr: 2.62e-03, grad_scale: 16.0
2024-09-25 19:39:11,461 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.69 vs. limit=15.0
2024-09-25 19:39:21,862 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=816550.0, ans=0.1
2024-09-25 19:39:23,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=816550.0, ans=0.0
2024-09-25 19:39:39,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=816596.6666666666, ans=0.125
2024-09-25 19:39:39,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=816596.6666666666, ans=0.1
2024-09-25 19:39:54,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=4.72 vs. limit=15.0
2024-09-25 19:40:05,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=816690.0, ans=0.125
2024-09-25 19:40:08,833 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:40:22,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=816736.6666666666, ans=0.125
2024-09-25 19:40:28,818 INFO [train.py:1198] (1/4) Epoch 45, batch 3600, loss[loss=0.1967, ctc_loss=0.1252, cr_loss=0.3574, over 17295.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1191, cr_loss=0.3343, over 3352289.11 frames. ], batch size: 46, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:40:32,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=816783.3333333334, ans=0.0
2024-09-25 19:40:44,754 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=816830.0, ans=0.0
2024-09-25 19:41:37,495 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.193e+02 1.298e+02 1.374e+02 1.488e+02 2.136e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-25 19:41:46,855 INFO [train.py:1198] (1/4) Epoch 45, batch 3650, loss[loss=0.1868, ctc_loss=0.1214, cr_loss=0.3266, over 17353.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1193, cr_loss=0.3345, over 3362028.99 frames. ], batch size: 48, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:41:53,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=817016.6666666666, ans=0.125
2024-09-25 19:42:28,837 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.42 vs. limit=6.0
2024-09-25 19:42:31,534 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=817110.0, ans=0.125
2024-09-25 19:42:39,540 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:42:54,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=817203.3333333334, ans=0.125
2024-09-25 19:43:01,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=817203.3333333334, ans=0.2
2024-09-25 19:43:05,643 INFO [train.py:1198] (1/4) Epoch 45, batch 3700, loss[loss=0.2438, ctc_loss=0.16, cr_loss=0.419, over 15138.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1202, cr_loss=0.3364, over 3356840.86 frames. ], batch size: 89, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:43:37,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=817343.3333333334, ans=0.1
2024-09-25 19:43:40,320 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:43:49,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=817343.3333333334, ans=0.025
2024-09-25 19:43:53,481 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.86 vs. limit=12.0
2024-09-25 19:44:00,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=817390.0, ans=0.025
2024-09-25 19:44:03,969 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=817390.0, ans=0.1
2024-09-25 19:44:14,698 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.280e+02 1.340e+02 1.437e+02 1.807e+02, threshold=2.680e+02, percent-clipped=0.0
2024-09-25 19:44:24,087 INFO [train.py:1198] (1/4) Epoch 45, batch 3750, loss[loss=0.1985, ctc_loss=0.1294, cr_loss=0.345, over 17000.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1206, cr_loss=0.3372, over 3356263.60 frames. ], batch size: 56, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:44:24,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=817483.3333333334, ans=0.025
2024-09-25 19:44:29,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=817483.3333333334, ans=10.0
2024-09-25 19:45:08,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=817576.6666666666, ans=0.125
2024-09-25 19:45:12,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=817623.3333333334, ans=0.125
2024-09-25 19:45:21,587 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=817623.3333333334, ans=0.125
2024-09-25 19:45:21,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=817623.3333333334, ans=0.125
2024-09-25 19:45:26,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=817623.3333333334, ans=0.1
2024-09-25 19:45:42,770 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.12 vs. limit=15.0
2024-09-25 19:45:44,856 INFO [train.py:1198] (1/4) Epoch 45, batch 3800, loss[loss=0.2337, ctc_loss=0.1537, cr_loss=0.4, over 15169.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1204, cr_loss=0.3364, over 3338870.59 frames. ], batch size: 89, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:46:00,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=817763.3333333334, ans=0.05
2024-09-25 19:46:05,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=817763.3333333334, ans=0.125
2024-09-25 19:46:54,350 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.363e+02 1.462e+02 1.575e+02 2.339e+02, threshold=2.925e+02, percent-clipped=0.0
2024-09-25 19:47:01,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=817903.3333333334, ans=0.125
2024-09-25 19:47:03,900 INFO [train.py:1198] (1/4) Epoch 45, batch 3850, loss[loss=0.164, ctc_loss=0.1049, cr_loss=0.2956, over 17148.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1216, cr_loss=0.3378, over 3296684.52 frames. ], batch size: 41, lr: 2.62e-03, grad_scale: 32.0
2024-09-25 19:47:44,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=818043.3333333334, ans=0.025
2024-09-25 19:47:47,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=818043.3333333334, ans=0.2
2024-09-25 19:48:02,963 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=818090.0, ans=0.025
2024-09-25 19:49:07,311 INFO [train.py:1198] (1/4) Epoch 46, batch 0, loss[loss=0.195, ctc_loss=0.1245, cr_loss=0.3526, over 17081.00 frames. ], tot_loss[loss=0.195, ctc_loss=0.1245, cr_loss=0.3526, over 17081.00 frames. ], batch size: 49, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 19:49:07,311 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 19:49:21,486 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([2.0372, 3.2268, 3.0902, 3.2979, 3.2847, 2.7331, 3.0582, 2.1288],
       device='cuda:1')
2024-09-25 19:49:22,412 INFO [train.py:1230] (1/4) Epoch 46, validation: loss=0.03502, ctc_loss=0.03502, cr_loss=1.054e-14, over 944034.00 frames. 
2024-09-25 19:49:22,413 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 19:49:27,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=818164.6666666666, ans=0.0
2024-09-25 19:49:27,936 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.20 vs. limit=15.0
2024-09-25 19:49:35,458 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 19:49:40,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=818211.3333333334, ans=0.1
2024-09-25 19:50:04,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=818258.0, ans=0.125
2024-09-25 19:50:17,268 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=818304.6666666666, ans=0.125
2024-09-25 19:50:40,884 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.337e+02 1.469e+02 1.679e+02 2.405e+02, threshold=2.939e+02, percent-clipped=0.0
2024-09-25 19:50:44,076 INFO [train.py:1198] (1/4) Epoch 46, batch 50, loss[loss=0.1773, ctc_loss=0.1135, cr_loss=0.3188, over 17088.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1214, cr_loss=0.3373, over 747984.81 frames. ], batch size: 49, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 19:51:19,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=19.77 vs. limit=22.5
2024-09-25 19:52:08,977 INFO [train.py:1198] (1/4) Epoch 46, batch 100, loss[loss=0.2026, ctc_loss=0.1322, cr_loss=0.3516, over 16778.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1199, cr_loss=0.3358, over 1330360.81 frames. ], batch size: 61, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 19:52:33,586 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=818678.0, ans=0.1
2024-09-25 19:53:03,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=818771.3333333334, ans=0.035
2024-09-25 19:53:12,309 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.39 vs. limit=15.0
2024-09-25 19:53:27,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=818818.0, ans=0.125
2024-09-25 19:53:28,956 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.282e+02 1.339e+02 1.432e+02 3.555e+02, threshold=2.678e+02, percent-clipped=2.0
2024-09-25 19:53:32,231 INFO [train.py:1198] (1/4) Epoch 46, batch 150, loss[loss=0.2106, ctc_loss=0.1368, cr_loss=0.3689, over 17214.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1199, cr_loss=0.337, over 1789348.83 frames. ], batch size: 55, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 19:53:32,808 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.22 vs. limit=22.5
2024-09-25 19:53:39,555 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.10 vs. limit=15.0
2024-09-25 19:53:56,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=818911.3333333334, ans=0.2
2024-09-25 19:54:22,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.62 vs. limit=22.5
2024-09-25 19:54:39,329 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=819051.3333333334, ans=0.125
2024-09-25 19:54:48,808 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=819051.3333333334, ans=0.09899494936611666
2024-09-25 19:54:51,561 INFO [train.py:1198] (1/4) Epoch 46, batch 200, loss[loss=0.2068, ctc_loss=0.1332, cr_loss=0.368, over 17192.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1195, cr_loss=0.336, over 2142491.52 frames. ], batch size: 55, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 19:54:53,916 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.91 vs. limit=22.5
2024-09-25 19:55:19,045 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=819144.6666666666, ans=0.125
2024-09-25 19:55:38,020 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.07 vs. limit=15.0
2024-09-25 19:56:13,857 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.302e+02 1.353e+02 1.428e+02 2.054e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-25 19:56:17,269 INFO [train.py:1198] (1/4) Epoch 46, batch 250, loss[loss=0.2114, ctc_loss=0.135, cr_loss=0.3821, over 16976.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1203, cr_loss=0.3377, over 2414085.85 frames. ], batch size: 53, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 19:56:48,145 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=819378.0, ans=0.0
2024-09-25 19:56:50,162 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=4.84 vs. limit=15.0
2024-09-25 19:57:20,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=819471.3333333334, ans=0.125
2024-09-25 19:57:23,635 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=819518.0, ans=0.0
2024-09-25 19:57:28,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=819518.0, ans=0.1
2024-09-25 19:57:40,839 INFO [train.py:1198] (1/4) Epoch 46, batch 300, loss[loss=0.2039, ctc_loss=0.1313, cr_loss=0.3631, over 17016.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1205, cr_loss=0.3376, over 2623560.50 frames. ], batch size: 52, lr: 2.59e-03, grad_scale: 16.0
2024-09-25 19:57:49,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=819564.6666666666, ans=0.0
2024-09-25 19:57:57,376 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=819611.3333333334, ans=0.0
2024-09-25 19:57:57,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=819611.3333333334, ans=0.2
2024-09-25 19:58:43,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=819704.6666666666, ans=0.025
2024-09-25 19:59:02,222 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.300e+02 1.374e+02 1.446e+02 2.011e+02, threshold=2.748e+02, percent-clipped=0.0
2024-09-25 19:59:03,734 INFO [train.py:1198] (1/4) Epoch 46, batch 350, loss[loss=0.188, ctc_loss=0.121, cr_loss=0.335, over 17202.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1206, cr_loss=0.3377, over 2783723.81 frames. ], batch size: 47, lr: 2.59e-03, grad_scale: 16.0
2024-09-25 19:59:07,211 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=819798.0, ans=0.2
2024-09-25 19:59:15,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=819798.0, ans=0.1
2024-09-25 19:59:27,828 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=819844.6666666666, ans=0.05
2024-09-25 19:59:42,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=819891.3333333334, ans=0.125
2024-09-25 19:59:42,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=819891.3333333334, ans=0.125
2024-09-25 19:59:57,111 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=7.16 vs. limit=15.0
2024-09-25 20:00:22,798 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=9.22 vs. limit=10.0
2024-09-25 20:00:23,195 INFO [train.py:1198] (1/4) Epoch 46, batch 400, loss[loss=0.1824, ctc_loss=0.1134, cr_loss=0.3445, over 17164.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1193, cr_loss=0.3359, over 2915374.83 frames. ], batch size: 41, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:01:20,477 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=820171.3333333334, ans=0.125
2024-09-25 20:01:46,653 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=4.98 vs. limit=15.0
2024-09-25 20:01:47,487 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=820218.0, ans=0.09899494936611666
2024-09-25 20:01:50,262 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.300e+02 1.386e+02 1.492e+02 1.969e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-25 20:01:52,002 INFO [train.py:1198] (1/4) Epoch 46, batch 450, loss[loss=0.1704, ctc_loss=0.1086, cr_loss=0.309, over 17300.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1194, cr_loss=0.3363, over 3015162.43 frames. ], batch size: 46, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:02:13,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=820311.3333333334, ans=0.125
2024-09-25 20:02:26,157 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=820358.0, ans=0.1
2024-09-25 20:02:34,828 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.58 vs. limit=15.0
2024-09-25 20:02:37,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=820358.0, ans=0.0
2024-09-25 20:02:39,021 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=820404.6666666666, ans=0.125
2024-09-25 20:02:42,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=820404.6666666666, ans=0.2
2024-09-25 20:02:45,418 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 20:03:07,279 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=820451.3333333334, ans=0.125
2024-09-25 20:03:14,978 INFO [train.py:1198] (1/4) Epoch 46, batch 500, loss[loss=0.1872, ctc_loss=0.1195, cr_loss=0.3387, over 17269.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1199, cr_loss=0.3369, over 3089960.78 frames. ], batch size: 46, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:03:26,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=820498.0, ans=0.0
2024-09-25 20:03:34,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=820544.6666666666, ans=0.0
2024-09-25 20:03:47,996 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.19 vs. limit=6.0
2024-09-25 20:03:49,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=820591.3333333334, ans=0.125
2024-09-25 20:03:50,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=820591.3333333334, ans=0.125
2024-09-25 20:04:16,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=820638.0, ans=0.1
2024-09-25 20:04:25,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=820684.6666666666, ans=0.0
2024-09-25 20:04:33,504 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.336e+02 1.423e+02 1.544e+02 3.291e+02, threshold=2.846e+02, percent-clipped=2.0
2024-09-25 20:04:35,166 INFO [train.py:1198] (1/4) Epoch 46, batch 550, loss[loss=0.2405, ctc_loss=0.158, cr_loss=0.4128, over 15092.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1208, cr_loss=0.3386, over 3144940.20 frames. ], batch size: 89, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:04:48,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=820731.3333333334, ans=0.0
2024-09-25 20:05:10,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.attention_skip_rate, batch_count=820824.6666666666, ans=0.0
2024-09-25 20:05:19,751 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=820824.6666666666, ans=0.125
2024-09-25 20:05:34,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=820871.3333333334, ans=0.125
2024-09-25 20:05:42,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=820918.0, ans=0.0
2024-09-25 20:05:56,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=820964.6666666666, ans=0.2
2024-09-25 20:06:00,562 INFO [train.py:1198] (1/4) Epoch 46, batch 600, loss[loss=0.1975, ctc_loss=0.1267, cr_loss=0.3542, over 17299.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1212, cr_loss=0.3392, over 3198804.03 frames. ], batch size: 49, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:06:20,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=14.54 vs. limit=22.5
2024-09-25 20:06:30,022 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.82 vs. limit=10.0
2024-09-25 20:06:36,043 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.78 vs. limit=15.0
2024-09-25 20:06:54,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=821104.6666666666, ans=0.125
2024-09-25 20:07:09,857 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=8.50 vs. limit=15.0
2024-09-25 20:07:12,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=821151.3333333334, ans=0.125
2024-09-25 20:07:21,441 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.282e+02 1.361e+02 1.501e+02 2.285e+02, threshold=2.723e+02, percent-clipped=0.0
2024-09-25 20:07:23,100 INFO [train.py:1198] (1/4) Epoch 46, batch 650, loss[loss=0.1516, ctc_loss=0.09529, cr_loss=0.2816, over 17197.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1201, cr_loss=0.3368, over 3233162.89 frames. ], batch size: 41, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:07:28,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.const_attention_rate, batch_count=821198.0, ans=0.025
2024-09-25 20:07:29,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=821198.0, ans=0.1
2024-09-25 20:07:34,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=821198.0, ans=0.125
2024-09-25 20:07:47,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=821244.6666666666, ans=0.2
2024-09-25 20:07:52,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=821244.6666666666, ans=0.025
2024-09-25 20:07:52,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=821244.6666666666, ans=0.125
2024-09-25 20:08:45,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=821384.6666666666, ans=0.125
2024-09-25 20:08:48,494 INFO [train.py:1198] (1/4) Epoch 46, batch 700, loss[loss=0.2071, ctc_loss=0.1345, cr_loss=0.3629, over 14866.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1203, cr_loss=0.3373, over 3253385.58 frames. ], batch size: 89, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:08:48,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=821431.3333333334, ans=0.1
2024-09-25 20:08:50,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=821431.3333333334, ans=0.0
2024-09-25 20:08:56,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=821431.3333333334, ans=0.5
2024-09-25 20:09:00,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=821431.3333333334, ans=0.05
2024-09-25 20:09:04,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=821478.0, ans=10.0
2024-09-25 20:09:14,848 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.44 vs. limit=15.0
2024-09-25 20:09:23,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=821524.6666666666, ans=0.125
2024-09-25 20:09:31,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=821524.6666666666, ans=0.2
2024-09-25 20:10:06,893 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.305e+02 1.403e+02 1.487e+02 1.713e+02, threshold=2.806e+02, percent-clipped=0.0
2024-09-25 20:10:08,541 INFO [train.py:1198] (1/4) Epoch 46, batch 750, loss[loss=0.2311, ctc_loss=0.1514, cr_loss=0.3985, over 17054.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1208, cr_loss=0.3382, over 3282378.86 frames. ], batch size: 52, lr: 2.59e-03, grad_scale: 32.0
2024-09-25 20:10:09,235 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.33 vs. limit=22.5
2024-09-25 20:10:12,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=821664.6666666666, ans=0.125
2024-09-25 20:10:29,337 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=821711.3333333334, ans=0.0
2024-09-25 20:10:40,423 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=821711.3333333334, ans=0.1
2024-09-25 20:10:51,232 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=821758.0, ans=0.125
2024-09-25 20:11:08,904 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.72 vs. limit=12.0
2024-09-25 20:11:27,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=821851.3333333334, ans=0.0
2024-09-25 20:11:34,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=821898.0, ans=0.125
2024-09-25 20:11:36,191 INFO [train.py:1198] (1/4) Epoch 46, batch 800, loss[loss=0.1751, ctc_loss=0.1086, cr_loss=0.3326, over 17001.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1209, cr_loss=0.3388, over 3300700.00 frames. ], batch size: 39, lr: 2.58e-03, grad_scale: 32.0
2024-09-25 20:12:40,464 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=822084.6666666666, ans=0.0
2024-09-25 20:12:43,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=822084.6666666666, ans=0.125
2024-09-25 20:12:46,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=822084.6666666666, ans=0.1
2024-09-25 20:12:58,584 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.283e+02 1.367e+02 1.457e+02 2.182e+02, threshold=2.735e+02, percent-clipped=0.0
2024-09-25 20:12:58,608 INFO [train.py:1198] (1/4) Epoch 46, batch 850, loss[loss=0.1813, ctc_loss=0.1105, cr_loss=0.3537, over 17254.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1211, cr_loss=0.3395, over 3309442.46 frames. ], batch size: 44, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:12:59,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=822131.3333333334, ans=0.0
2024-09-25 20:13:50,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.min_positive, batch_count=822271.3333333334, ans=0.05
2024-09-25 20:14:14,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=822318.0, ans=0.1
2024-09-25 20:14:18,697 INFO [train.py:1198] (1/4) Epoch 46, batch 900, loss[loss=0.1853, ctc_loss=0.1172, cr_loss=0.3407, over 17222.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1212, cr_loss=0.3395, over 3320784.09 frames. ], batch size: 50, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:14:33,456 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=822411.3333333334, ans=0.1
2024-09-25 20:14:35,916 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.52 vs. limit=5.0
2024-09-25 20:14:49,676 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.79 vs. limit=10.0
2024-09-25 20:14:51,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=822458.0, ans=0.125
2024-09-25 20:15:41,536 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.288e+02 1.364e+02 1.452e+02 2.497e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-25 20:15:41,561 INFO [train.py:1198] (1/4) Epoch 46, batch 950, loss[loss=0.2005, ctc_loss=0.1273, cr_loss=0.3659, over 17286.00 frames. ], tot_loss[loss=0.1893, ctc_loss=0.1214, cr_loss=0.3399, over 3333183.89 frames. ], batch size: 49, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:15:45,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=822598.0, ans=0.2
2024-09-25 20:16:03,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.54 vs. limit=12.0
2024-09-25 20:16:16,806 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=822691.3333333334, ans=0.125
2024-09-25 20:16:18,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=822691.3333333334, ans=0.125
2024-09-25 20:16:27,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=822691.3333333334, ans=0.07
2024-09-25 20:16:40,645 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=9.96 vs. limit=22.5
2024-09-25 20:17:07,491 INFO [train.py:1198] (1/4) Epoch 46, batch 1000, loss[loss=0.1898, ctc_loss=0.1201, cr_loss=0.3485, over 17310.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1203, cr_loss=0.3376, over 3344044.55 frames. ], batch size: 49, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:17:17,630 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=822831.3333333334, ans=0.1
2024-09-25 20:17:48,512 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.97 vs. limit=15.0
2024-09-25 20:17:55,638 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=822924.6666666666, ans=0.025
2024-09-25 20:18:14,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=823018.0, ans=0.125
2024-09-25 20:18:27,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=823018.0, ans=0.015
2024-09-25 20:18:27,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=823018.0, ans=0.0
2024-09-25 20:18:30,593 INFO [train.py:1198] (1/4) Epoch 46, batch 1050, loss[loss=0.1995, ctc_loss=0.1298, cr_loss=0.3488, over 17230.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1202, cr_loss=0.3381, over 3349973.93 frames. ], batch size: 50, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:18:32,115 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.054e+02 1.303e+02 1.373e+02 1.497e+02 1.983e+02, threshold=2.746e+02, percent-clipped=0.0
2024-09-25 20:18:39,174 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.99 vs. limit=6.0
2024-09-25 20:19:13,699 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=823158.0, ans=0.125
2024-09-25 20:19:21,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=823204.6666666666, ans=0.0
2024-09-25 20:19:23,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=823204.6666666666, ans=0.125
2024-09-25 20:19:34,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=823251.3333333334, ans=0.2
2024-09-25 20:19:39,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=823251.3333333334, ans=0.0
2024-09-25 20:19:44,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=823251.3333333334, ans=0.0
2024-09-25 20:19:50,554 INFO [train.py:1198] (1/4) Epoch 46, batch 1100, loss[loss=0.152, ctc_loss=0.09466, cr_loss=0.2867, over 16943.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1202, cr_loss=0.3376, over 3356162.09 frames. ], batch size: 42, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:19:53,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=823298.0, ans=0.0
2024-09-25 20:19:59,302 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.49 vs. limit=22.5
2024-09-25 20:20:07,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=823344.6666666666, ans=0.0
2024-09-25 20:20:17,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=823344.6666666666, ans=0.125
2024-09-25 20:20:25,351 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=823391.3333333334, ans=0.125
2024-09-25 20:20:30,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=823391.3333333334, ans=0.125
2024-09-25 20:20:51,708 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.85 vs. limit=15.0
2024-09-25 20:21:03,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=823484.6666666666, ans=0.0
2024-09-25 20:21:15,712 INFO [train.py:1198] (1/4) Epoch 46, batch 1150, loss[loss=0.1982, ctc_loss=0.1265, cr_loss=0.3585, over 16984.00 frames. ], tot_loss[loss=0.1891, ctc_loss=0.1212, cr_loss=0.3396, over 3361077.28 frames. ], batch size: 53, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:21:17,339 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.324e+02 1.372e+02 1.463e+02 5.655e+02, threshold=2.745e+02, percent-clipped=1.0
2024-09-25 20:21:33,485 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.59 vs. limit=15.0
2024-09-25 20:21:44,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=823578.0, ans=0.0
2024-09-25 20:22:02,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=823624.6666666666, ans=0.125
2024-09-25 20:22:05,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=823671.3333333334, ans=0.125
2024-09-25 20:22:38,928 INFO [train.py:1198] (1/4) Epoch 46, batch 1200, loss[loss=0.1976, ctc_loss=0.1277, cr_loss=0.3495, over 16767.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1206, cr_loss=0.3386, over 3367642.83 frames. ], batch size: 61, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:23:01,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=823811.3333333334, ans=0.125
2024-09-25 20:23:13,608 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=823858.0, ans=0.125
2024-09-25 20:23:20,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=823858.0, ans=0.0
2024-09-25 20:23:26,476 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.93 vs. limit=10.0
2024-09-25 20:23:50,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=823951.3333333334, ans=0.0
2024-09-25 20:23:50,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=823951.3333333334, ans=0.125
2024-09-25 20:24:00,928 INFO [train.py:1198] (1/4) Epoch 46, batch 1250, loss[loss=0.1817, ctc_loss=0.1151, cr_loss=0.3329, over 17032.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.12, cr_loss=0.3373, over 3367432.79 frames. ], batch size: 44, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:24:02,523 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.285e+02 1.375e+02 1.486e+02 1.915e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-25 20:24:27,578 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.42 vs. limit=8.0
2024-09-25 20:24:51,110 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 20:25:22,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=824231.3333333334, ans=0.1
2024-09-25 20:25:23,678 INFO [train.py:1198] (1/4) Epoch 46, batch 1300, loss[loss=0.1712, ctc_loss=0.1093, cr_loss=0.3095, over 17081.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1194, cr_loss=0.3363, over 3373423.76 frames. ], batch size: 43, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:25:33,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_positive, batch_count=824231.3333333334, ans=0.05
2024-09-25 20:25:54,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=824324.6666666666, ans=0.125
2024-09-25 20:26:02,461 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=3.36 vs. limit=15.0
2024-09-25 20:26:33,103 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=824418.0, ans=0.0
2024-09-25 20:26:36,138 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=824418.0, ans=0.2
2024-09-25 20:26:48,556 INFO [train.py:1198] (1/4) Epoch 46, batch 1350, loss[loss=0.2223, ctc_loss=0.144, cr_loss=0.3912, over 14998.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1194, cr_loss=0.3364, over 3376330.89 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:26:51,683 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.295e+02 1.386e+02 1.472e+02 1.767e+02, threshold=2.771e+02, percent-clipped=0.0
2024-09-25 20:26:57,363 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.44 vs. limit=15.0
2024-09-25 20:27:40,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=824604.6666666666, ans=0.125
2024-09-25 20:27:43,435 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=824604.6666666666, ans=0.2
2024-09-25 20:27:58,665 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=824651.3333333334, ans=0.125
2024-09-25 20:28:11,415 INFO [train.py:1198] (1/4) Epoch 46, batch 1400, loss[loss=0.2214, ctc_loss=0.144, cr_loss=0.3872, over 17039.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1201, cr_loss=0.3377, over 3367907.28 frames. ], batch size: 52, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:28:19,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=824698.0, ans=0.0
2024-09-25 20:28:29,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=824744.6666666666, ans=0.1
2024-09-25 20:29:04,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=824838.0, ans=0.07
2024-09-25 20:29:15,382 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.53 vs. limit=22.5
2024-09-25 20:29:21,641 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.79 vs. limit=15.0
2024-09-25 20:29:31,892 INFO [train.py:1198] (1/4) Epoch 46, batch 1450, loss[loss=0.1723, ctc_loss=0.109, cr_loss=0.3166, over 17165.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1206, cr_loss=0.3386, over 3373456.82 frames. ], batch size: 45, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:29:35,104 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.313e+02 1.376e+02 1.492e+02 1.838e+02, threshold=2.751e+02, percent-clipped=0.0
2024-09-25 20:29:40,411 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 20:30:13,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=825024.6666666666, ans=0.2
2024-09-25 20:30:13,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=825024.6666666666, ans=0.125
2024-09-25 20:30:19,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=825024.6666666666, ans=0.125
2024-09-25 20:30:42,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=825118.0, ans=0.125
2024-09-25 20:30:54,859 INFO [train.py:1198] (1/4) Epoch 46, batch 1500, loss[loss=0.2074, ctc_loss=0.1332, cr_loss=0.3712, over 16786.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1202, cr_loss=0.3376, over 3369355.12 frames. ], batch size: 61, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:31:15,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=825211.3333333334, ans=0.1
2024-09-25 20:32:03,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=825351.3333333334, ans=0.125
2024-09-25 20:32:20,344 INFO [train.py:1198] (1/4) Epoch 46, batch 1550, loss[loss=0.1871, ctc_loss=0.1185, cr_loss=0.3431, over 17074.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1192, cr_loss=0.3357, over 3377661.01 frames. ], batch size: 46, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:32:23,492 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.290e+02 1.399e+02 1.518e+02 4.516e+02, threshold=2.798e+02, percent-clipped=1.0
2024-09-25 20:32:43,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=825444.6666666666, ans=0.125
2024-09-25 20:33:16,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=825538.0, ans=0.2
2024-09-25 20:33:43,585 INFO [train.py:1198] (1/4) Epoch 46, batch 1600, loss[loss=0.1977, ctc_loss=0.1281, cr_loss=0.3481, over 17221.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1194, cr_loss=0.3355, over 3370687.25 frames. ], batch size: 50, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:33:48,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=825631.3333333334, ans=0.125
2024-09-25 20:33:57,397 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=4.35 vs. limit=15.0
2024-09-25 20:34:06,989 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.28 vs. limit=22.5
2024-09-25 20:34:21,210 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=14.61 vs. limit=22.5
2024-09-25 20:34:31,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.attention_skip_rate, batch_count=825771.3333333334, ans=0.0
2024-09-25 20:34:35,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=825771.3333333334, ans=0.1
2024-09-25 20:34:49,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=825818.0, ans=0.1
2024-09-25 20:35:04,110 INFO [train.py:1198] (1/4) Epoch 46, batch 1650, loss[loss=0.1587, ctc_loss=0.09885, cr_loss=0.2991, over 16969.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1194, cr_loss=0.336, over 3368066.96 frames. ], batch size: 42, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:35:07,356 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.315e+02 1.405e+02 1.548e+02 2.408e+02, threshold=2.810e+02, percent-clipped=0.0
2024-09-25 20:35:20,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=825864.6666666666, ans=0.2
2024-09-25 20:35:28,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=825911.3333333334, ans=0.035
2024-09-25 20:35:44,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=825958.0, ans=0.2
2024-09-25 20:36:10,377 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=21.35 vs. limit=22.5
2024-09-25 20:36:11,322 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=826004.6666666666, ans=0.1
2024-09-25 20:36:15,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=826051.3333333334, ans=0.1
2024-09-25 20:36:26,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=826051.3333333334, ans=0.1
2024-09-25 20:36:32,222 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.35 vs. limit=15.0
2024-09-25 20:36:32,697 INFO [train.py:1198] (1/4) Epoch 46, batch 1700, loss[loss=0.2039, ctc_loss=0.1295, cr_loss=0.3719, over 16909.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1205, cr_loss=0.3379, over 3361627.44 frames. ], batch size: 58, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:37:52,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=826284.6666666666, ans=0.125
2024-09-25 20:37:52,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=826284.6666666666, ans=0.125
2024-09-25 20:37:55,295 INFO [train.py:1198] (1/4) Epoch 46, batch 1750, loss[loss=0.2153, ctc_loss=0.1394, cr_loss=0.3793, over 16536.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1201, cr_loss=0.3369, over 3348790.04 frames. ], batch size: 66, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:38:00,304 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.299e+02 1.380e+02 1.488e+02 2.427e+02, threshold=2.759e+02, percent-clipped=0.0
2024-09-25 20:38:04,382 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.06 vs. limit=15.0
2024-09-25 20:38:24,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=826378.0, ans=0.125
2024-09-25 20:38:50,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=826471.3333333334, ans=0.125
2024-09-25 20:39:02,958 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=826518.0, ans=0.0
2024-09-25 20:39:15,274 INFO [train.py:1198] (1/4) Epoch 46, batch 1800, loss[loss=0.1669, ctc_loss=0.1031, cr_loss=0.319, over 16973.00 frames. ], tot_loss[loss=0.188, ctc_loss=0.1204, cr_loss=0.338, over 3350563.54 frames. ], batch size: 42, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:39:20,851 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.52 vs. limit=6.0
2024-09-25 20:39:25,116 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=826564.6666666666, ans=0.025
2024-09-25 20:39:36,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=826611.3333333334, ans=0.125
2024-09-25 20:40:00,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=826658.0, ans=0.2
2024-09-25 20:40:13,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=826704.6666666666, ans=0.125
2024-09-25 20:40:38,768 INFO [train.py:1198] (1/4) Epoch 46, batch 1850, loss[loss=0.1745, ctc_loss=0.1113, cr_loss=0.3163, over 17353.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1202, cr_loss=0.3374, over 3347545.54 frames. ], batch size: 48, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:40:42,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=826798.0, ans=0.125
2024-09-25 20:40:43,542 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.312e+02 1.384e+02 1.504e+02 2.324e+02, threshold=2.767e+02, percent-clipped=0.0
2024-09-25 20:40:59,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=826844.6666666666, ans=0.0
2024-09-25 20:41:49,873 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=826984.6666666666, ans=0.125
2024-09-25 20:41:50,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.66 vs. limit=22.5
2024-09-25 20:42:01,116 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 20:42:03,933 INFO [train.py:1198] (1/4) Epoch 46, batch 1900, loss[loss=0.2004, ctc_loss=0.1287, cr_loss=0.3587, over 17018.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1195, cr_loss=0.3356, over 3345698.59 frames. ], batch size: 56, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:42:04,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=827031.3333333334, ans=0.2
2024-09-25 20:42:05,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=827031.3333333334, ans=0.07
2024-09-25 20:42:21,071 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=10.57 vs. limit=15.0
2024-09-25 20:42:23,396 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=827078.0, ans=0.125
2024-09-25 20:43:01,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=827171.3333333334, ans=0.0
2024-09-25 20:43:04,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.attention_skip_rate, batch_count=827171.3333333334, ans=0.0
2024-09-25 20:43:12,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=827218.0, ans=0.0
2024-09-25 20:43:15,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=827218.0, ans=0.125
2024-09-25 20:43:17,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=827218.0, ans=0.0
2024-09-25 20:43:26,921 INFO [train.py:1198] (1/4) Epoch 46, batch 1950, loss[loss=0.2076, ctc_loss=0.1352, cr_loss=0.3622, over 17048.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1195, cr_loss=0.3348, over 3329319.67 frames. ], batch size: 52, lr: 2.58e-03, grad_scale: 8.0
2024-09-25 20:43:31,731 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.295e+02 1.354e+02 1.509e+02 3.056e+02, threshold=2.707e+02, percent-clipped=1.0
2024-09-25 20:43:44,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=827311.3333333334, ans=0.025
2024-09-25 20:44:18,430 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.93 vs. limit=6.0
2024-09-25 20:44:21,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=827404.6666666666, ans=0.0
2024-09-25 20:44:46,531 INFO [train.py:1198] (1/4) Epoch 46, batch 2000, loss[loss=0.2053, ctc_loss=0.1308, cr_loss=0.372, over 16998.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1195, cr_loss=0.3353, over 3340628.84 frames. ], batch size: 56, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:44:47,293 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.67 vs. limit=10.0
2024-09-25 20:44:54,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=827498.0, ans=0.1
2024-09-25 20:45:21,153 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=827591.3333333334, ans=0.2
2024-09-25 20:45:21,243 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 20:45:26,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=827591.3333333334, ans=0.125
2024-09-25 20:45:35,500 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=827638.0, ans=0.125
2024-09-25 20:45:40,151 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=827638.0, ans=0.125
2024-09-25 20:46:05,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=827684.6666666666, ans=0.125
2024-09-25 20:46:11,124 INFO [train.py:1198] (1/4) Epoch 46, batch 2050, loss[loss=0.1973, ctc_loss=0.1276, cr_loss=0.3483, over 17003.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1209, cr_loss=0.3378, over 3323619.06 frames. ], batch size: 53, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:46:15,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=827731.3333333334, ans=0.0
2024-09-25 20:46:18,714 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.306e+02 1.402e+02 1.494e+02 2.074e+02, threshold=2.805e+02, percent-clipped=0.0
2024-09-25 20:47:04,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=827871.3333333334, ans=0.1
2024-09-25 20:47:15,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=827871.3333333334, ans=0.125
2024-09-25 20:47:20,217 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=827918.0, ans=0.1
2024-09-25 20:47:34,393 INFO [train.py:1198] (1/4) Epoch 46, batch 2100, loss[loss=0.178, ctc_loss=0.1135, cr_loss=0.3224, over 17212.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1208, cr_loss=0.3381, over 3330913.11 frames. ], batch size: 47, lr: 2.58e-03, grad_scale: 16.0
2024-09-25 20:47:37,998 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-25 20:47:39,965 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=9.28 vs. limit=15.0
2024-09-25 20:48:06,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.44 vs. limit=6.0
2024-09-25 20:48:08,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=828058.0, ans=0.125
2024-09-25 20:48:17,396 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.79 vs. limit=15.0
2024-09-25 20:48:49,230 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=828151.3333333334, ans=0.0
2024-09-25 20:48:56,845 INFO [train.py:1198] (1/4) Epoch 46, batch 2150, loss[loss=0.1672, ctc_loss=0.1061, cr_loss=0.3056, over 17259.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1186, cr_loss=0.334, over 3349572.40 frames. ], batch size: 44, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 20:48:57,082 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=828198.0, ans=0.125
2024-09-25 20:48:57,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=828198.0, ans=0.125
2024-09-25 20:49:01,529 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.324e+02 1.399e+02 1.517e+02 1.900e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-25 20:49:31,496 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=3.88 vs. limit=15.0
2024-09-25 20:49:41,997 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=828291.3333333334, ans=0.0
2024-09-25 20:49:51,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=828338.0, ans=0.125
2024-09-25 20:49:53,300 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=828338.0, ans=0.125
2024-09-25 20:49:57,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=828338.0, ans=0.125
2024-09-25 20:50:02,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=828384.6666666666, ans=0.0
2024-09-25 20:50:19,237 INFO [train.py:1198] (1/4) Epoch 46, batch 2200, loss[loss=0.2157, ctc_loss=0.1423, cr_loss=0.3669, over 15276.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1196, cr_loss=0.3354, over 3339028.84 frames. ], batch size: 89, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 20:50:19,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=828431.3333333334, ans=0.04949747468305833
2024-09-25 20:50:45,682 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=828478.0, ans=0.125
2024-09-25 20:50:55,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.33 vs. limit=15.0
2024-09-25 20:51:44,712 INFO [train.py:1198] (1/4) Epoch 46, batch 2250, loss[loss=0.2202, ctc_loss=0.1419, cr_loss=0.3915, over 16141.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1198, cr_loss=0.3358, over 3335215.83 frames. ], batch size: 74, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 20:51:49,434 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.294e+02 1.364e+02 1.489e+02 2.080e+02, threshold=2.728e+02, percent-clipped=0.0
2024-09-25 20:52:05,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=828711.3333333334, ans=0.125
2024-09-25 20:52:05,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=828711.3333333334, ans=0.125
2024-09-25 20:52:16,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=828758.0, ans=0.015
2024-09-25 20:52:28,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=828758.0, ans=0.125
2024-09-25 20:52:45,223 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=828804.6666666666, ans=0.2
2024-09-25 20:53:04,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=828851.3333333334, ans=0.125
2024-09-25 20:53:07,440 INFO [train.py:1198] (1/4) Epoch 46, batch 2300, loss[loss=0.2066, ctc_loss=0.1354, cr_loss=0.3558, over 16014.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1196, cr_loss=0.3357, over 3339343.68 frames. ], batch size: 74, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 20:53:10,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=828898.0, ans=0.0
2024-09-25 20:53:57,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=829038.0, ans=0.125
2024-09-25 20:53:58,044 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.90 vs. limit=10.0
2024-09-25 20:54:05,389 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.53 vs. limit=12.0
2024-09-25 20:54:08,563 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=829038.0, ans=0.0
2024-09-25 20:54:14,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=829084.6666666666, ans=0.0
2024-09-25 20:54:27,182 INFO [train.py:1198] (1/4) Epoch 46, batch 2350, loss[loss=0.1906, ctc_loss=0.1196, cr_loss=0.3551, over 17171.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1192, cr_loss=0.3348, over 3343101.09 frames. ], batch size: 45, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 20:54:31,906 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.291e+02 1.368e+02 1.435e+02 1.902e+02, threshold=2.737e+02, percent-clipped=0.0
2024-09-25 20:54:35,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=829131.3333333334, ans=0.0
2024-09-25 20:54:56,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=829178.0, ans=0.125
2024-09-25 20:55:28,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=829271.3333333334, ans=0.1
2024-09-25 20:55:33,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=829318.0, ans=0.05
2024-09-25 20:55:45,940 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=829318.0, ans=0.0
2024-09-25 20:55:50,418 INFO [train.py:1198] (1/4) Epoch 46, batch 2400, loss[loss=0.1926, ctc_loss=0.124, cr_loss=0.3429, over 17096.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1192, cr_loss=0.3348, over 3335933.99 frames. ], batch size: 43, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 20:55:54,707 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.05 vs. limit=22.5
2024-09-25 20:56:29,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=829458.0, ans=0.0
2024-09-25 20:57:07,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=829551.3333333334, ans=0.0
2024-09-25 20:57:15,351 INFO [train.py:1198] (1/4) Epoch 46, batch 2450, loss[loss=0.2102, ctc_loss=0.1338, cr_loss=0.3819, over 16990.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1194, cr_loss=0.3349, over 3337375.77 frames. ], batch size: 53, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 20:57:20,236 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.311e+02 1.415e+02 1.511e+02 3.334e+02, threshold=2.830e+02, percent-clipped=1.0
2024-09-25 20:57:20,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=829598.0, ans=0.1
2024-09-25 20:57:42,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=829644.6666666666, ans=0.04949747468305833
2024-09-25 20:58:24,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=829784.6666666666, ans=0.0
2024-09-25 20:58:24,406 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.17 vs. limit=15.0
2024-09-25 20:58:38,038 INFO [train.py:1198] (1/4) Epoch 46, batch 2500, loss[loss=0.2175, ctc_loss=0.142, cr_loss=0.3776, over 17235.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1194, cr_loss=0.3352, over 3344770.12 frames. ], batch size: 55, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 20:58:44,845 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=829831.3333333334, ans=0.125
2024-09-25 20:58:51,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=829831.3333333334, ans=0.1
2024-09-25 20:59:00,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=829878.0, ans=0.5
2024-09-25 20:59:29,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=829971.3333333334, ans=0.125
2024-09-25 20:59:48,293 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=7.97 vs. limit=15.0
2024-09-25 20:59:50,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=830018.0, ans=0.125
2024-09-25 20:59:58,772 INFO [train.py:1198] (1/4) Epoch 46, batch 2550, loss[loss=0.178, ctc_loss=0.1128, cr_loss=0.3262, over 17140.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1193, cr_loss=0.3352, over 3353409.61 frames. ], batch size: 48, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 20:59:59,643 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.26 vs. limit=22.5
2024-09-25 21:00:06,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.313e+02 1.390e+02 1.516e+02 1.832e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-25 21:00:13,450 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.02 vs. limit=6.0
2024-09-25 21:01:26,888 INFO [train.py:1198] (1/4) Epoch 46, batch 2600, loss[loss=0.2046, ctc_loss=0.1306, cr_loss=0.3699, over 17294.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1197, cr_loss=0.336, over 3352991.43 frames. ], batch size: 49, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:01:32,761 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.49 vs. limit=10.0
2024-09-25 21:02:01,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=830391.3333333334, ans=0.0
2024-09-25 21:02:03,078 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=830391.3333333334, ans=0.125
2024-09-25 21:02:04,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=830391.3333333334, ans=0.125
2024-09-25 21:02:04,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=830391.3333333334, ans=0.1
2024-09-25 21:02:50,768 INFO [train.py:1198] (1/4) Epoch 46, batch 2650, loss[loss=0.1799, ctc_loss=0.1162, cr_loss=0.3186, over 17313.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1201, cr_loss=0.337, over 3357192.80 frames. ], batch size: 51, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:02:55,451 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.287e+02 1.345e+02 1.480e+02 2.035e+02, threshold=2.690e+02, percent-clipped=0.0
2024-09-25 21:03:13,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=830578.0, ans=0.5
2024-09-25 21:03:40,837 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.22 vs. limit=15.0
2024-09-25 21:03:45,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=830671.3333333334, ans=0.1
2024-09-25 21:03:50,174 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=830671.3333333334, ans=0.0
2024-09-25 21:04:03,580 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.29 vs. limit=15.0
2024-09-25 21:04:09,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward2.hidden_balancer.prob, batch_count=830764.6666666666, ans=0.125
2024-09-25 21:04:10,490 INFO [train.py:1198] (1/4) Epoch 46, batch 2700, loss[loss=0.1941, ctc_loss=0.1241, cr_loss=0.3502, over 17319.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.1201, cr_loss=0.3373, over 3355527.19 frames. ], batch size: 51, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:04:15,307 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=830764.6666666666, ans=0.0
2024-09-25 21:04:17,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=830764.6666666666, ans=0.125
2024-09-25 21:04:17,959 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.22 vs. limit=22.5
2024-09-25 21:04:31,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=830811.3333333334, ans=0.0
2024-09-25 21:05:06,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=830904.6666666666, ans=0.125
2024-09-25 21:05:11,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=830904.6666666666, ans=0.125
2024-09-25 21:05:13,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=830904.6666666666, ans=0.025
2024-09-25 21:05:27,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=830951.3333333334, ans=0.0
2024-09-25 21:05:32,279 INFO [train.py:1198] (1/4) Epoch 46, batch 2750, loss[loss=0.1994, ctc_loss=0.1289, cr_loss=0.3526, over 17296.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.1198, cr_loss=0.3369, over 3361320.59 frames. ], batch size: 49, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:05:32,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=830998.0, ans=0.0
2024-09-25 21:05:35,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=830998.0, ans=0.1
2024-09-25 21:05:37,114 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.184e+02 1.313e+02 1.406e+02 1.516e+02 1.975e+02, threshold=2.813e+02, percent-clipped=0.0
2024-09-25 21:05:38,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=830998.0, ans=0.1
2024-09-25 21:06:16,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=831091.3333333334, ans=0.035
2024-09-25 21:06:30,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=831138.0, ans=0.125
2024-09-25 21:06:35,407 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=831138.0, ans=0.0
2024-09-25 21:06:38,588 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=831138.0, ans=0.125
2024-09-25 21:06:57,840 INFO [train.py:1198] (1/4) Epoch 46, batch 2800, loss[loss=0.1782, ctc_loss=0.1114, cr_loss=0.3342, over 17300.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1196, cr_loss=0.3364, over 3357960.48 frames. ], batch size: 46, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:07:12,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=831278.0, ans=0.0
2024-09-25 21:07:16,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=831278.0, ans=0.0
2024-09-25 21:08:15,379 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.04 vs. limit=15.0
2024-09-25 21:08:20,587 INFO [train.py:1198] (1/4) Epoch 46, batch 2850, loss[loss=0.1866, ctc_loss=0.1181, cr_loss=0.3426, over 17153.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1198, cr_loss=0.3369, over 3353982.67 frames. ], batch size: 45, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:08:20,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=831464.6666666666, ans=0.2
2024-09-25 21:08:26,959 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.275e+02 1.372e+02 1.492e+02 2.089e+02, threshold=2.744e+02, percent-clipped=0.0
2024-09-25 21:08:35,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=831511.3333333334, ans=0.0
2024-09-25 21:08:57,798 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=831558.0, ans=0.1
2024-09-25 21:09:18,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=831604.6666666666, ans=0.1
2024-09-25 21:09:40,971 INFO [train.py:1198] (1/4) Epoch 46, batch 2900, loss[loss=0.1934, ctc_loss=0.1227, cr_loss=0.3531, over 17094.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1195, cr_loss=0.3362, over 3356199.29 frames. ], batch size: 49, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:09:47,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=831698.0, ans=0.035
2024-09-25 21:09:52,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=831698.0, ans=0.2
2024-09-25 21:09:59,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=831744.6666666666, ans=0.025
2024-09-25 21:10:04,815 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=831744.6666666666, ans=0.125
2024-09-25 21:10:11,776 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.01 vs. limit=15.0
2024-09-25 21:10:45,372 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.78 vs. limit=12.0
2024-09-25 21:10:48,301 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=3.65 vs. limit=10.0
2024-09-25 21:11:08,572 INFO [train.py:1198] (1/4) Epoch 46, batch 2950, loss[loss=0.1551, ctc_loss=0.09617, cr_loss=0.2946, over 17176.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1194, cr_loss=0.336, over 3357433.60 frames. ], batch size: 41, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:11:10,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=831931.3333333334, ans=0.125
2024-09-25 21:11:16,442 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.273e+02 1.341e+02 1.442e+02 4.466e+02, threshold=2.682e+02, percent-clipped=1.0
2024-09-25 21:11:21,696 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.prob, batch_count=831931.3333333334, ans=0.125
2024-09-25 21:11:31,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=831978.0, ans=0.0
2024-09-25 21:11:56,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=832071.3333333334, ans=0.125
2024-09-25 21:12:04,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=832071.3333333334, ans=0.04949747468305833
2024-09-25 21:12:27,906 INFO [train.py:1198] (1/4) Epoch 46, batch 3000, loss[loss=0.1866, ctc_loss=0.1202, cr_loss=0.3321, over 17303.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1204, cr_loss=0.3372, over 3357006.96 frames. ], batch size: 49, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:12:27,907 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 21:12:43,455 INFO [train.py:1230] (1/4) Epoch 46, validation: loss=0.03583, ctc_loss=0.03583, cr_loss=1.006e-14, over 944034.00 frames. 
2024-09-25 21:12:43,456 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 21:12:51,686 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=832164.6666666666, ans=0.125
2024-09-25 21:13:02,814 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 21:13:06,739 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=5.13 vs. limit=15.0
2024-09-25 21:13:12,261 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=832211.3333333334, ans=0.1
2024-09-25 21:13:35,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=832304.6666666666, ans=0.025
2024-09-25 21:13:38,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=832304.6666666666, ans=0.125
2024-09-25 21:13:43,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=832304.6666666666, ans=0.2
2024-09-25 21:13:46,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=832351.3333333334, ans=0.0
2024-09-25 21:13:46,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=832351.3333333334, ans=0.2
2024-09-25 21:14:01,693 INFO [train.py:1198] (1/4) Epoch 46, batch 3050, loss[loss=0.1501, ctc_loss=0.09416, cr_loss=0.2797, over 16310.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1213, cr_loss=0.3388, over 3363701.47 frames. ], batch size: 36, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:14:09,458 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.315e+02 1.415e+02 1.524e+02 2.421e+02, threshold=2.829e+02, percent-clipped=0.0
2024-09-25 21:14:33,681 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=10.86 vs. limit=22.5
2024-09-25 21:14:36,387 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=832491.3333333334, ans=0.025
2024-09-25 21:15:20,313 INFO [train.py:1198] (1/4) Epoch 46, batch 3100, loss[loss=0.1468, ctc_loss=0.09022, cr_loss=0.2831, over 17022.00 frames. ], tot_loss[loss=0.1889, ctc_loss=0.1211, cr_loss=0.3386, over 3360735.80 frames. ], batch size: 39, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:15:45,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=832678.0, ans=0.2
2024-09-25 21:16:03,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=832724.6666666666, ans=0.0
2024-09-25 21:16:06,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=832771.3333333334, ans=0.125
2024-09-25 21:16:09,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=832771.3333333334, ans=0.2
2024-09-25 21:16:38,987 INFO [train.py:1198] (1/4) Epoch 46, batch 3150, loss[loss=0.1306, ctc_loss=0.0793, cr_loss=0.2563, over 16277.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1196, cr_loss=0.3364, over 3365164.84 frames. ], batch size: 36, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:16:46,804 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.285e+02 1.343e+02 1.430e+02 2.452e+02, threshold=2.687e+02, percent-clipped=0.0
2024-09-25 21:16:47,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=832864.6666666666, ans=0.0
2024-09-25 21:17:10,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=832958.0, ans=0.0
2024-09-25 21:17:15,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=832958.0, ans=0.1
2024-09-25 21:17:24,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=833004.6666666666, ans=0.125
2024-09-25 21:17:53,591 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=833051.3333333334, ans=0.125
2024-09-25 21:17:59,371 INFO [train.py:1198] (1/4) Epoch 46, batch 3200, loss[loss=0.1798, ctc_loss=0.1149, cr_loss=0.3243, over 16845.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1196, cr_loss=0.3358, over 3364640.22 frames. ], batch size: 58, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:18:16,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=833144.6666666666, ans=0.125
2024-09-25 21:19:17,329 INFO [train.py:1198] (1/4) Epoch 46, batch 3250, loss[loss=0.1638, ctc_loss=0.1027, cr_loss=0.3056, over 16960.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.119, cr_loss=0.334, over 3365447.74 frames. ], batch size: 42, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:19:25,151 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.305e+02 1.369e+02 1.483e+02 2.240e+02, threshold=2.739e+02, percent-clipped=0.0
2024-09-25 21:19:52,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=833424.6666666666, ans=0.04949747468305833
2024-09-25 21:19:56,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=833424.6666666666, ans=0.2
2024-09-25 21:20:17,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=833471.3333333334, ans=0.09899494936611666
2024-09-25 21:20:40,789 INFO [train.py:1198] (1/4) Epoch 46, batch 3300, loss[loss=0.2274, ctc_loss=0.1502, cr_loss=0.3863, over 15068.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1193, cr_loss=0.3354, over 3364747.16 frames. ], batch size: 89, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:20:53,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=833564.6666666666, ans=0.125
2024-09-25 21:21:17,692 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.77 vs. limit=12.0
2024-09-25 21:21:58,829 INFO [train.py:1198] (1/4) Epoch 46, batch 3350, loss[loss=0.2077, ctc_loss=0.1372, cr_loss=0.3529, over 16774.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.12, cr_loss=0.3374, over 3367761.33 frames. ], batch size: 61, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:22:06,614 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.301e+02 1.404e+02 1.465e+02 2.030e+02, threshold=2.807e+02, percent-clipped=0.0
2024-09-25 21:22:08,484 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=833798.0, ans=0.2
2024-09-25 21:22:33,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=833891.3333333334, ans=0.1
2024-09-25 21:22:50,736 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.72 vs. limit=15.0
2024-09-25 21:22:56,630 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=833938.0, ans=0.125
2024-09-25 21:23:02,104 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=10.16 vs. limit=10.0
2024-09-25 21:23:16,305 INFO [train.py:1198] (1/4) Epoch 46, batch 3400, loss[loss=0.2297, ctc_loss=0.1516, cr_loss=0.3905, over 15091.00 frames. ], tot_loss[loss=0.1882, ctc_loss=0.1206, cr_loss=0.3384, over 3360455.15 frames. ], batch size: 89, lr: 2.57e-03, grad_scale: 32.0
2024-09-25 21:23:16,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=834031.3333333334, ans=0.125
2024-09-25 21:24:14,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=834171.3333333334, ans=0.2
2024-09-25 21:24:17,153 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.91 vs. limit=15.0
2024-09-25 21:24:19,470 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=834218.0, ans=0.125
2024-09-25 21:24:25,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=834218.0, ans=0.1
2024-09-25 21:24:36,359 INFO [train.py:1198] (1/4) Epoch 46, batch 3450, loss[loss=0.1849, ctc_loss=0.1203, cr_loss=0.3229, over 17363.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.12, cr_loss=0.3373, over 3364207.02 frames. ], batch size: 48, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:24:43,689 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.94 vs. limit=15.0
2024-09-25 21:24:45,571 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.317e+02 1.418e+02 1.501e+02 3.351e+02, threshold=2.836e+02, percent-clipped=1.0
2024-09-25 21:24:49,387 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.75 vs. limit=6.0
2024-09-25 21:24:58,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=834311.3333333334, ans=0.125
2024-09-25 21:25:18,557 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=834358.0, ans=0.0
2024-09-25 21:25:25,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=834404.6666666666, ans=0.125
2024-09-25 21:25:33,517 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.15 vs. limit=22.5
2024-09-25 21:25:35,843 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=2.502e-03
2024-09-25 21:25:42,627 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.73 vs. limit=6.0
2024-09-25 21:25:54,706 INFO [train.py:1198] (1/4) Epoch 46, batch 3500, loss[loss=0.151, ctc_loss=0.09261, cr_loss=0.2918, over 17194.00 frames. ], tot_loss[loss=0.1883, ctc_loss=0.1205, cr_loss=0.3388, over 3358781.47 frames. ], batch size: 41, lr: 2.57e-03, grad_scale: 16.0
2024-09-25 21:25:54,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=834498.0, ans=0.2
2024-09-25 21:26:25,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=834591.3333333334, ans=0.1
2024-09-25 21:26:33,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=834591.3333333334, ans=0.0
2024-09-25 21:26:38,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=834591.3333333334, ans=0.025
2024-09-25 21:27:05,981 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn2.whiten, num_groups=1, num_channels=192, metric=8.06 vs. limit=22.5
2024-09-25 21:27:12,618 INFO [train.py:1198] (1/4) Epoch 46, batch 3550, loss[loss=0.1595, ctc_loss=0.09989, cr_loss=0.2981, over 17047.00 frames. ], tot_loss[loss=0.188, ctc_loss=0.1203, cr_loss=0.3387, over 3365116.41 frames. ], batch size: 39, lr: 2.56e-03, grad_scale: 16.0
2024-09-25 21:27:17,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=834731.3333333334, ans=0.0
2024-09-25 21:27:21,484 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=8.65 vs. limit=15.0
2024-09-25 21:27:21,880 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.283e+02 1.360e+02 1.445e+02 2.258e+02, threshold=2.720e+02, percent-clipped=0.0
2024-09-25 21:27:28,344 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=834778.0, ans=0.025
2024-09-25 21:27:36,581 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=834778.0, ans=0.0
2024-09-25 21:27:37,408 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=7.96 vs. limit=15.0
2024-09-25 21:28:25,182 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.92 vs. limit=15.0
2024-09-25 21:28:31,163 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=834964.6666666666, ans=0.025
2024-09-25 21:28:32,453 INFO [train.py:1198] (1/4) Epoch 46, batch 3600, loss[loss=0.1993, ctc_loss=0.1283, cr_loss=0.3549, over 17362.00 frames. ], tot_loss[loss=0.189, ctc_loss=0.1211, cr_loss=0.3395, over 3357695.55 frames. ], batch size: 48, lr: 2.56e-03, grad_scale: 32.0
2024-09-25 21:28:43,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=834964.6666666666, ans=0.125
2024-09-25 21:28:46,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=835011.3333333334, ans=0.2
2024-09-25 21:28:56,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=835011.3333333334, ans=0.1
2024-09-25 21:29:39,878 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.min_positive, batch_count=835151.3333333334, ans=0.025
2024-09-25 21:29:55,383 INFO [train.py:1198] (1/4) Epoch 46, batch 3650, loss[loss=0.1654, ctc_loss=0.1048, cr_loss=0.3033, over 17080.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1203, cr_loss=0.3375, over 3353610.74 frames. ], batch size: 43, lr: 2.56e-03, grad_scale: 32.0
2024-09-25 21:30:04,584 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.303e+02 1.373e+02 1.458e+02 2.085e+02, threshold=2.745e+02, percent-clipped=0.0
2024-09-25 21:30:29,900 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=835291.3333333334, ans=0.0
2024-09-25 21:30:46,553 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=5.10 vs. limit=15.0
2024-09-25 21:30:56,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=835384.6666666666, ans=0.125
2024-09-25 21:31:14,426 INFO [train.py:1198] (1/4) Epoch 46, batch 3700, loss[loss=0.2421, ctc_loss=0.1616, cr_loss=0.4028, over 11681.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1208, cr_loss=0.3381, over 3348388.82 frames. ], batch size: 123, lr: 2.56e-03, grad_scale: 16.0
2024-09-25 21:31:26,406 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=6.02 vs. limit=15.0
2024-09-25 21:31:46,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=835524.6666666666, ans=0.125
2024-09-25 21:32:01,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=835571.3333333334, ans=0.025
2024-09-25 21:32:32,251 INFO [train.py:1198] (1/4) Epoch 46, batch 3750, loss[loss=0.1958, ctc_loss=0.1241, cr_loss=0.3585, over 17263.00 frames. ], tot_loss[loss=0.1894, ctc_loss=0.1215, cr_loss=0.3395, over 3341181.80 frames. ], batch size: 44, lr: 2.56e-03, grad_scale: 16.0
2024-09-25 21:32:32,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=835664.6666666666, ans=0.125
2024-09-25 21:32:35,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=835664.6666666666, ans=0.125
2024-09-25 21:32:43,197 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.294e+02 1.384e+02 1.512e+02 2.088e+02, threshold=2.767e+02, percent-clipped=0.0
2024-09-25 21:32:49,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=835711.3333333334, ans=0.0
2024-09-25 21:33:15,375 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=835758.0, ans=0.1
2024-09-25 21:33:18,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=835804.6666666666, ans=0.2
2024-09-25 21:33:19,947 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=835804.6666666666, ans=0.2
2024-09-25 21:33:37,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=835851.3333333334, ans=0.1
2024-09-25 21:33:50,842 INFO [train.py:1198] (1/4) Epoch 46, batch 3800, loss[loss=0.1652, ctc_loss=0.1048, cr_loss=0.3019, over 17290.00 frames. ], tot_loss[loss=0.1899, ctc_loss=0.1219, cr_loss=0.3399, over 3311850.05 frames. ], batch size: 46, lr: 2.56e-03, grad_scale: 16.0
2024-09-25 21:34:27,326 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.93 vs. limit=6.0
2024-09-25 21:34:48,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=836038.0, ans=0.125
2024-09-25 21:35:01,856 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=10.76 vs. limit=15.0
2024-09-25 21:35:08,993 INFO [train.py:1198] (1/4) Epoch 46, batch 3850, loss[loss=0.2196, ctc_loss=0.1482, cr_loss=0.3567, over 11643.00 frames. ], tot_loss[loss=0.1906, ctc_loss=0.1225, cr_loss=0.3405, over 3284393.94 frames. ], batch size: 123, lr: 2.56e-03, grad_scale: 16.0
2024-09-25 21:35:19,791 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.341e+02 1.442e+02 1.559e+02 2.635e+02, threshold=2.885e+02, percent-clipped=0.0
2024-09-25 21:35:21,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.attention_skip_rate, batch_count=836131.3333333334, ans=0.0
2024-09-25 21:35:23,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=836178.0, ans=0.0
2024-09-25 21:35:24,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=836178.0, ans=0.125
2024-09-25 21:35:47,347 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=836224.6666666666, ans=0.1
2024-09-25 21:35:53,530 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=836271.3333333334, ans=0.05
2024-09-25 21:36:04,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=836271.3333333334, ans=0.125
2024-09-25 21:36:05,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=836271.3333333334, ans=0.1
2024-09-25 21:37:05,580 INFO [train.py:1198] (1/4) Epoch 47, batch 0, loss[loss=0.2103, ctc_loss=0.1373, cr_loss=0.3647, over 16124.00 frames. ], tot_loss[loss=0.2103, ctc_loss=0.1373, cr_loss=0.3647, over 16124.00 frames. ], batch size: 74, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:37:05,581 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 21:37:14,252 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.3.self_attn_weights, attn_weights_entropy = tensor([3.4905, 3.7228, 4.0265, 3.8535, 4.4075, 4.3659, 4.4392, 3.6700],
       device='cuda:1')
2024-09-25 21:37:22,170 INFO [train.py:1230] (1/4) Epoch 47, validation: loss=0.03509, ctc_loss=0.03509, cr_loss=1.062e-14, over 944034.00 frames. 
2024-09-25 21:37:22,171 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 21:37:33,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=836346.0, ans=0.1
2024-09-25 21:37:45,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.92 vs. limit=6.0
2024-09-25 21:38:07,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=836439.3333333334, ans=0.125
2024-09-25 21:38:32,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=836532.6666666666, ans=0.125
2024-09-25 21:38:44,943 INFO [train.py:1198] (1/4) Epoch 47, batch 50, loss[loss=0.1856, ctc_loss=0.1176, cr_loss=0.3404, over 17294.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1178, cr_loss=0.334, over 767008.06 frames. ], batch size: 46, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:38:48,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=836579.3333333334, ans=0.0
2024-09-25 21:38:53,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=836579.3333333334, ans=0.125
2024-09-25 21:39:01,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=836626.0, ans=0.09899494936611666
2024-09-25 21:39:02,615 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.320e+02 1.514e+02 1.646e+02 2.881e+02, threshold=3.028e+02, percent-clipped=0.0
2024-09-25 21:39:09,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=836626.0, ans=0.125
2024-09-25 21:39:14,090 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=836626.0, ans=0.0
2024-09-25 21:39:14,120 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=836626.0, ans=0.125
2024-09-25 21:39:15,846 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 21:39:22,698 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=7.96 vs. limit=15.0
2024-09-25 21:39:40,266 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.42 vs. limit=12.0
2024-09-25 21:39:51,299 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.42 vs. limit=15.0
2024-09-25 21:40:05,067 INFO [train.py:1198] (1/4) Epoch 47, batch 100, loss[loss=0.1965, ctc_loss=0.1273, cr_loss=0.3457, over 17345.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1204, cr_loss=0.3388, over 1340831.18 frames. ], batch size: 48, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:40:10,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=836812.6666666666, ans=0.125
2024-09-25 21:40:13,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward3.hidden_balancer.prob, batch_count=836812.6666666666, ans=0.125
2024-09-25 21:40:14,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=836812.6666666666, ans=0.125
2024-09-25 21:40:14,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=836812.6666666666, ans=0.125
2024-09-25 21:40:19,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=836859.3333333334, ans=0.125
2024-09-25 21:40:23,111 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=836859.3333333334, ans=0.125
2024-09-25 21:40:47,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=836906.0, ans=0.0
2024-09-25 21:40:50,448 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 21:40:56,934 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=836952.6666666666, ans=0.125
2024-09-25 21:41:26,868 INFO [train.py:1198] (1/4) Epoch 47, batch 150, loss[loss=0.1702, ctc_loss=0.108, cr_loss=0.311, over 17021.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1186, cr_loss=0.3352, over 1797735.94 frames. ], batch size: 44, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:41:27,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=837046.0, ans=0.04949747468305833
2024-09-25 21:41:44,260 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.291e+02 1.353e+02 1.432e+02 2.053e+02, threshold=2.706e+02, percent-clipped=0.0
2024-09-25 21:42:09,386 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.23 vs. limit=22.5
2024-09-25 21:42:28,025 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.48 vs. limit=22.5
2024-09-25 21:42:29,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=837186.0, ans=0.125
2024-09-25 21:42:47,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=837232.6666666666, ans=0.125
2024-09-25 21:42:53,060 INFO [train.py:1198] (1/4) Epoch 47, batch 200, loss[loss=0.1627, ctc_loss=0.1038, cr_loss=0.2944, over 17259.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1182, cr_loss=0.3344, over 2150148.00 frames. ], batch size: 44, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:43:07,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=837326.0, ans=0.025
2024-09-25 21:43:09,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=837326.0, ans=0.2
2024-09-25 21:43:13,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=837326.0, ans=0.125
2024-09-25 21:43:14,268 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.98 vs. limit=6.0
2024-09-25 21:43:16,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=837326.0, ans=0.125
2024-09-25 21:43:34,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=837372.6666666666, ans=0.035
2024-09-25 21:43:37,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=837372.6666666666, ans=0.025
2024-09-25 21:43:44,414 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.27 vs. limit=15.0
2024-09-25 21:43:45,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=837419.3333333334, ans=0.125
2024-09-25 21:44:01,918 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.88 vs. limit=6.0
2024-09-25 21:44:15,403 INFO [train.py:1198] (1/4) Epoch 47, batch 250, loss[loss=0.1954, ctc_loss=0.128, cr_loss=0.337, over 17293.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1186, cr_loss=0.3354, over 2418476.09 frames. ], batch size: 46, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:44:15,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=837512.6666666666, ans=0.2
2024-09-25 21:44:32,924 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.332e+02 1.402e+02 1.492e+02 3.444e+02, threshold=2.804e+02, percent-clipped=1.0
2024-09-25 21:44:36,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=837559.3333333334, ans=0.125
2024-09-25 21:45:08,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=837652.6666666666, ans=0.1
2024-09-25 21:45:35,449 INFO [train.py:1198] (1/4) Epoch 47, batch 300, loss[loss=0.188, ctc_loss=0.1198, cr_loss=0.341, over 17102.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1189, cr_loss=0.3357, over 2637226.91 frames. ], batch size: 49, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:46:21,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=837839.3333333334, ans=0.1
2024-09-25 21:46:44,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=837932.6666666666, ans=0.125
2024-09-25 21:47:01,260 INFO [train.py:1198] (1/4) Epoch 47, batch 350, loss[loss=0.172, ctc_loss=0.1081, cr_loss=0.3197, over 17160.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.1181, cr_loss=0.3341, over 2804926.57 frames. ], batch size: 45, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:47:17,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=838026.0, ans=0.2
2024-09-25 21:47:18,802 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.331e+02 1.404e+02 1.493e+02 2.320e+02, threshold=2.808e+02, percent-clipped=0.0
2024-09-25 21:47:49,838 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.90 vs. limit=15.0
2024-09-25 21:47:55,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=838119.3333333334, ans=0.2
2024-09-25 21:48:18,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=838166.0, ans=0.07
2024-09-25 21:48:24,403 INFO [train.py:1198] (1/4) Epoch 47, batch 400, loss[loss=0.1869, ctc_loss=0.1212, cr_loss=0.3286, over 17028.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.118, cr_loss=0.334, over 2933138.17 frames. ], batch size: 51, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:48:24,788 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=838212.6666666666, ans=0.125
2024-09-25 21:48:31,089 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=8.24 vs. limit=15.0
2024-09-25 21:48:40,160 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass_mid.scale_min, batch_count=838212.6666666666, ans=0.2
2024-09-25 21:48:59,266 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=838306.0, ans=0.125
2024-09-25 21:49:33,841 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.91 vs. limit=10.0
2024-09-25 21:49:47,315 INFO [train.py:1198] (1/4) Epoch 47, batch 450, loss[loss=0.1368, ctc_loss=0.08578, cr_loss=0.2551, over 17284.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.118, cr_loss=0.3344, over 3036916.72 frames. ], batch size: 42, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:50:02,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=838492.6666666666, ans=0.125
2024-09-25 21:50:04,976 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.297e+02 1.382e+02 1.499e+02 1.706e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-25 21:51:09,780 INFO [train.py:1198] (1/4) Epoch 47, batch 500, loss[loss=0.1745, ctc_loss=0.109, cr_loss=0.3273, over 17256.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1184, cr_loss=0.3354, over 3108491.16 frames. ], batch size: 44, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:51:13,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=838679.3333333334, ans=0.125
2024-09-25 21:51:16,334 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=838679.3333333334, ans=0.125
2024-09-25 21:51:19,515 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=838679.3333333334, ans=0.125
2024-09-25 21:51:42,350 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.91 vs. limit=15.0
2024-09-25 21:51:43,424 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=838772.6666666666, ans=0.125
2024-09-25 21:51:50,230 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.34 vs. limit=15.0
2024-09-25 21:51:52,896 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=838772.6666666666, ans=0.125
2024-09-25 21:52:05,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=838819.3333333334, ans=0.1
2024-09-25 21:52:35,634 INFO [train.py:1198] (1/4) Epoch 47, batch 550, loss[loss=0.1761, ctc_loss=0.1124, cr_loss=0.3184, over 17265.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1193, cr_loss=0.3366, over 3160321.54 frames. ], batch size: 44, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:52:46,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=838912.6666666666, ans=0.2
2024-09-25 21:52:53,168 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.330e+02 1.398e+02 1.505e+02 2.211e+02, threshold=2.797e+02, percent-clipped=0.0
2024-09-25 21:53:44,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=839099.3333333334, ans=0.125
2024-09-25 21:53:54,167 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=839099.3333333334, ans=0.125
2024-09-25 21:53:58,552 INFO [train.py:1198] (1/4) Epoch 47, batch 600, loss[loss=0.1894, ctc_loss=0.1217, cr_loss=0.3386, over 16009.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1192, cr_loss=0.3357, over 3206133.86 frames. ], batch size: 74, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:55:04,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=839332.6666666666, ans=0.2
2024-09-25 21:55:18,521 INFO [train.py:1198] (1/4) Epoch 47, batch 650, loss[loss=0.2082, ctc_loss=0.1339, cr_loss=0.3712, over 17183.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1183, cr_loss=0.3345, over 3241755.17 frames. ], batch size: 55, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:55:36,234 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.288e+02 1.367e+02 1.485e+02 1.946e+02, threshold=2.733e+02, percent-clipped=0.0
2024-09-25 21:55:39,913 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.67 vs. limit=15.0
2024-09-25 21:56:22,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=839519.3333333334, ans=0.125
2024-09-25 21:56:23,316 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=5.55 vs. limit=12.0
2024-09-25 21:56:33,726 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=839566.0, ans=0.125
2024-09-25 21:56:35,551 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.36 vs. limit=15.0
2024-09-25 21:56:41,550 INFO [train.py:1198] (1/4) Epoch 47, batch 700, loss[loss=0.1673, ctc_loss=0.1075, cr_loss=0.2989, over 17016.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1188, cr_loss=0.3353, over 3267705.50 frames. ], batch size: 39, lr: 2.53e-03, grad_scale: 16.0
2024-09-25 21:56:49,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=839612.6666666666, ans=0.0
2024-09-25 21:56:49,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=839612.6666666666, ans=0.2
2024-09-25 21:56:57,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=839612.6666666666, ans=0.1
2024-09-25 21:57:24,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=839706.0, ans=0.125
2024-09-25 21:57:43,495 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=8.63 vs. limit=15.0
2024-09-25 21:57:52,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.52 vs. limit=22.5
2024-09-25 21:58:06,331 INFO [train.py:1198] (1/4) Epoch 47, batch 750, loss[loss=0.1895, ctc_loss=0.119, cr_loss=0.3529, over 17204.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1197, cr_loss=0.3372, over 3284168.66 frames. ], batch size: 55, lr: 2.53e-03, grad_scale: 16.0
2024-09-25 21:58:27,883 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.312e+02 1.373e+02 1.486e+02 2.253e+02, threshold=2.746e+02, percent-clipped=0.0
2024-09-25 21:58:38,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=839892.6666666666, ans=0.5
2024-09-25 21:59:19,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=840032.6666666666, ans=0.125
2024-09-25 21:59:31,441 INFO [train.py:1198] (1/4) Epoch 47, batch 800, loss[loss=0.2029, ctc_loss=0.1306, cr_loss=0.3613, over 17021.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1192, cr_loss=0.336, over 3302045.31 frames. ], batch size: 52, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 21:59:44,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=840079.3333333334, ans=0.2
2024-09-25 21:59:56,153 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.62 vs. limit=15.0
2024-09-25 22:00:22,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=840219.3333333334, ans=0.125
2024-09-25 22:00:54,191 INFO [train.py:1198] (1/4) Epoch 47, batch 850, loss[loss=0.179, ctc_loss=0.1134, cr_loss=0.328, over 17087.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1189, cr_loss=0.3351, over 3310772.08 frames. ], batch size: 40, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:00:56,377 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.34 vs. limit=15.0
2024-09-25 22:01:00,691 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff3_skip_rate, batch_count=840312.6666666666, ans=0.0
2024-09-25 22:01:02,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=840312.6666666666, ans=0.125
2024-09-25 22:01:05,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=840312.6666666666, ans=0.1
2024-09-25 22:01:13,303 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.297e+02 1.378e+02 1.477e+02 2.622e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 22:01:32,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=840406.0, ans=0.0
2024-09-25 22:02:19,133 INFO [train.py:1198] (1/4) Epoch 47, batch 900, loss[loss=0.1622, ctc_loss=0.102, cr_loss=0.301, over 17096.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1191, cr_loss=0.3358, over 3321504.69 frames. ], batch size: 49, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:02:33,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=840592.6666666666, ans=0.125
2024-09-25 22:03:10,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff3_skip_rate, batch_count=840686.0, ans=0.0
2024-09-25 22:03:41,091 INFO [train.py:1198] (1/4) Epoch 47, batch 950, loss[loss=0.1945, ctc_loss=0.1238, cr_loss=0.3539, over 17205.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1193, cr_loss=0.3363, over 3335578.56 frames. ], batch size: 55, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:03:57,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=840826.0, ans=0.2
2024-09-25 22:04:00,133 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.323e+02 1.399e+02 1.542e+02 2.460e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-25 22:04:05,263 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=840826.0, ans=0.0
2024-09-25 22:04:49,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=840966.0, ans=0.1
2024-09-25 22:05:00,654 INFO [train.py:1198] (1/4) Epoch 47, batch 1000, loss[loss=0.1695, ctc_loss=0.1063, cr_loss=0.316, over 17281.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1198, cr_loss=0.3366, over 3326540.62 frames. ], batch size: 42, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:05:15,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=841059.3333333334, ans=0.2
2024-09-25 22:05:17,035 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=841059.3333333334, ans=0.2
2024-09-25 22:05:51,445 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=841152.6666666666, ans=0.125
2024-09-25 22:06:04,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=841152.6666666666, ans=0.125
2024-09-25 22:06:23,151 INFO [train.py:1198] (1/4) Epoch 47, batch 1050, loss[loss=0.1937, ctc_loss=0.1285, cr_loss=0.326, over 17032.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.1201, cr_loss=0.3373, over 3325001.55 frames. ], batch size: 56, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:06:38,074 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:06:42,387 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.322e+02 1.409e+02 1.520e+02 2.848e+02, threshold=2.818e+02, percent-clipped=1.0
2024-09-25 22:06:44,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=841292.6666666666, ans=0.125
2024-09-25 22:06:44,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=841292.6666666666, ans=0.1
2024-09-25 22:07:03,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=841339.3333333334, ans=0.025
2024-09-25 22:07:17,023 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.72 vs. limit=6.0
2024-09-25 22:07:24,626 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=841386.0, ans=0.1
2024-09-25 22:07:30,977 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=841432.6666666666, ans=0.1
2024-09-25 22:07:32,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=841432.6666666666, ans=0.2
2024-09-25 22:07:47,936 INFO [train.py:1198] (1/4) Epoch 47, batch 1100, loss[loss=0.2256, ctc_loss=0.1473, cr_loss=0.3918, over 16532.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1198, cr_loss=0.3373, over 3334631.63 frames. ], batch size: 66, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:07:48,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=841479.3333333334, ans=0.125
2024-09-25 22:08:22,924 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=841572.6666666666, ans=0.025
2024-09-25 22:08:34,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=841572.6666666666, ans=0.125
2024-09-25 22:09:10,494 INFO [train.py:1198] (1/4) Epoch 47, batch 1150, loss[loss=0.1666, ctc_loss=0.1053, cr_loss=0.3061, over 15854.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1197, cr_loss=0.3377, over 3334363.78 frames. ], batch size: 35, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:09:29,553 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.291e+02 1.365e+02 1.481e+02 2.112e+02, threshold=2.730e+02, percent-clipped=0.0
2024-09-25 22:09:39,467 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=841759.3333333334, ans=0.125
2024-09-25 22:09:39,512 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=841759.3333333334, ans=0.0
2024-09-25 22:09:50,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=841806.0, ans=0.1
2024-09-25 22:10:00,234 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=841852.6666666666, ans=0.125
2024-09-25 22:10:08,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=841852.6666666666, ans=0.0
2024-09-25 22:10:08,622 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.99 vs. limit=15.0
2024-09-25 22:10:32,842 INFO [train.py:1198] (1/4) Epoch 47, batch 1200, loss[loss=0.18, ctc_loss=0.1133, cr_loss=0.3332, over 17034.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1201, cr_loss=0.3379, over 3342406.78 frames. ], batch size: 44, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:10:34,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=841946.0, ans=0.0
2024-09-25 22:10:45,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=841946.0, ans=0.0
2024-09-25 22:10:48,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=841992.6666666666, ans=0.125
2024-09-25 22:11:06,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.min_abs, batch_count=842039.3333333334, ans=0.5
2024-09-25 22:11:18,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=842039.3333333334, ans=0.5
2024-09-25 22:11:40,728 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=842132.6666666666, ans=0.1
2024-09-25 22:11:55,781 INFO [train.py:1198] (1/4) Epoch 47, batch 1250, loss[loss=0.1581, ctc_loss=0.0994, cr_loss=0.2937, over 16688.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1202, cr_loss=0.3377, over 3331220.66 frames. ], batch size: 37, lr: 2.53e-03, grad_scale: 32.0
2024-09-25 22:12:17,775 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.299e+02 1.369e+02 1.454e+02 1.923e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-25 22:12:27,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=842226.0, ans=0.125
2024-09-25 22:12:29,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=842272.6666666666, ans=0.125
2024-09-25 22:12:31,086 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=842272.6666666666, ans=0.0
2024-09-25 22:12:40,582 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=842272.6666666666, ans=0.125
2024-09-25 22:12:43,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=842272.6666666666, ans=0.125
2024-09-25 22:13:14,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=842366.0, ans=0.07
2024-09-25 22:13:17,104 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.58 vs. limit=15.0
2024-09-25 22:13:20,997 INFO [train.py:1198] (1/4) Epoch 47, batch 1300, loss[loss=0.1902, ctc_loss=0.122, cr_loss=0.3413, over 17088.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1204, cr_loss=0.3384, over 3334654.65 frames. ], batch size: 49, lr: 2.53e-03, grad_scale: 16.0
2024-09-25 22:13:26,112 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=842412.6666666666, ans=0.1
2024-09-25 22:13:34,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=842412.6666666666, ans=0.125
2024-09-25 22:13:35,819 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:13:37,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=842459.3333333334, ans=0.2
2024-09-25 22:13:43,734 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=842459.3333333334, ans=0.125
2024-09-25 22:14:09,944 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.69 vs. limit=15.0
2024-09-25 22:14:12,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=842552.6666666666, ans=0.125
2024-09-25 22:14:29,835 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=842599.3333333334, ans=0.125
2024-09-25 22:14:40,803 INFO [train.py:1198] (1/4) Epoch 47, batch 1350, loss[loss=0.1989, ctc_loss=0.128, cr_loss=0.3541, over 17227.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1207, cr_loss=0.3388, over 3343546.17 frames. ], batch size: 50, lr: 2.53e-03, grad_scale: 16.0
2024-09-25 22:14:44,649 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=10.82 vs. limit=15.0
2024-09-25 22:14:45,909 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=842646.0, ans=0.2
2024-09-25 22:14:55,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=842692.6666666666, ans=0.125
2024-09-25 22:15:00,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=842692.6666666666, ans=0.125
2024-09-25 22:15:01,642 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.184e+02 1.286e+02 1.352e+02 1.458e+02 2.037e+02, threshold=2.703e+02, percent-clipped=0.0
2024-09-25 22:15:01,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=842692.6666666666, ans=0.125
2024-09-25 22:15:02,049 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=842692.6666666666, ans=0.0
2024-09-25 22:15:33,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=842786.0, ans=0.0
2024-09-25 22:15:33,616 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.76 vs. limit=10.0
2024-09-25 22:15:39,836 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=842786.0, ans=0.2
2024-09-25 22:15:53,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=842832.6666666666, ans=0.0
2024-09-25 22:16:03,149 INFO [train.py:1198] (1/4) Epoch 47, batch 1400, loss[loss=0.2025, ctc_loss=0.1265, cr_loss=0.3798, over 17224.00 frames. ], tot_loss[loss=0.1881, ctc_loss=0.1205, cr_loss=0.3382, over 3341630.66 frames. ], batch size: 55, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:16:09,827 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=842879.3333333334, ans=0.0
2024-09-25 22:16:19,795 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=256, metric=10.37 vs. limit=22.5
2024-09-25 22:16:22,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=842926.0, ans=0.2
2024-09-25 22:17:27,855 INFO [train.py:1198] (1/4) Epoch 47, batch 1450, loss[loss=0.2201, ctc_loss=0.1431, cr_loss=0.385, over 17038.00 frames. ], tot_loss[loss=0.1884, ctc_loss=0.1207, cr_loss=0.3388, over 3323971.12 frames. ], batch size: 52, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:17:48,452 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.274e+02 1.355e+02 1.464e+02 2.562e+02, threshold=2.711e+02, percent-clipped=0.0
2024-09-25 22:17:56,791 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=843159.3333333334, ans=0.2
2024-09-25 22:18:12,426 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.28 vs. limit=12.0
2024-09-25 22:18:50,192 INFO [train.py:1198] (1/4) Epoch 47, batch 1500, loss[loss=0.1743, ctc_loss=0.1083, cr_loss=0.3301, over 17183.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1202, cr_loss=0.3384, over 3329285.14 frames. ], batch size: 41, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:18:50,503 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.prob, batch_count=843346.0, ans=0.125
2024-09-25 22:19:24,232 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=9.21 vs. limit=15.0
2024-09-25 22:19:34,042 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.99 vs. limit=15.0
2024-09-25 22:19:51,664 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:19:55,649 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=10.42 vs. limit=15.0
2024-09-25 22:20:10,581 INFO [train.py:1198] (1/4) Epoch 47, batch 1550, loss[loss=0.1395, ctc_loss=0.0862, cr_loss=0.2664, over 17084.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.1197, cr_loss=0.337, over 3338648.29 frames. ], batch size: 40, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:20:17,367 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:20:34,259 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.301e+02 1.382e+02 1.468e+02 1.851e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-25 22:20:53,962 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff3_skip_rate, batch_count=843672.6666666666, ans=0.0
2024-09-25 22:21:04,126 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.44 vs. limit=6.0
2024-09-25 22:21:04,491 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.38 vs. limit=15.0
2024-09-25 22:21:23,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=843766.0, ans=0.1
2024-09-25 22:21:34,159 INFO [train.py:1198] (1/4) Epoch 47, batch 1600, loss[loss=0.1545, ctc_loss=0.09571, cr_loss=0.2942, over 17117.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1189, cr_loss=0.3353, over 3350059.78 frames. ], batch size: 40, lr: 2.52e-03, grad_scale: 32.0
2024-09-25 22:21:36,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=843812.6666666666, ans=0.0
2024-09-25 22:21:39,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=843812.6666666666, ans=0.09899494936611666
2024-09-25 22:21:39,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=843812.6666666666, ans=0.0
2024-09-25 22:21:49,065 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=7.46 vs. limit=15.0
2024-09-25 22:22:51,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=843999.3333333334, ans=0.2
2024-09-25 22:22:59,305 INFO [train.py:1198] (1/4) Epoch 47, batch 1650, loss[loss=0.1828, ctc_loss=0.1161, cr_loss=0.3339, over 16896.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1184, cr_loss=0.3343, over 3354376.33 frames. ], batch size: 58, lr: 2.52e-03, grad_scale: 32.0
2024-09-25 22:22:59,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=844046.0, ans=0.09899494936611666
2024-09-25 22:23:21,443 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=844092.6666666666, ans=0.125
2024-09-25 22:23:22,602 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.287e+02 1.363e+02 1.436e+02 1.851e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-25 22:23:27,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=844092.6666666666, ans=0.1
2024-09-25 22:23:38,231 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.48 vs. limit=15.0
2024-09-25 22:23:56,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=844186.0, ans=0.125
2024-09-25 22:23:59,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=844186.0, ans=0.0
2024-09-25 22:24:21,990 INFO [train.py:1198] (1/4) Epoch 47, batch 1700, loss[loss=0.2113, ctc_loss=0.1382, cr_loss=0.3654, over 17342.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1185, cr_loss=0.3345, over 3364882.47 frames. ], batch size: 48, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:24:22,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=844279.3333333334, ans=10.0
2024-09-25 22:24:38,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=844326.0, ans=0.0
2024-09-25 22:25:06,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=844372.6666666666, ans=0.125
2024-09-25 22:25:44,105 INFO [train.py:1198] (1/4) Epoch 47, batch 1750, loss[loss=0.1867, ctc_loss=0.1217, cr_loss=0.3249, over 17301.00 frames. ], tot_loss[loss=0.1853, ctc_loss=0.1186, cr_loss=0.3339, over 3364625.47 frames. ], batch size: 51, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:25:50,821 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.attention_skip_rate, batch_count=844512.6666666666, ans=0.0
2024-09-25 22:25:58,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=844559.3333333334, ans=0.1
2024-09-25 22:26:02,466 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.15 vs. limit=15.0
2024-09-25 22:26:06,156 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.334e+02 1.394e+02 1.508e+02 1.965e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-25 22:26:08,604 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.51 vs. limit=10.0
2024-09-25 22:26:14,652 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=844606.0, ans=0.125
2024-09-25 22:26:16,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=844606.0, ans=0.0
2024-09-25 22:26:37,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=844652.6666666666, ans=0.125
2024-09-25 22:26:50,618 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:27:08,726 INFO [train.py:1198] (1/4) Epoch 47, batch 1800, loss[loss=0.1857, ctc_loss=0.1186, cr_loss=0.3356, over 16609.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.1183, cr_loss=0.3334, over 3356353.07 frames. ], batch size: 66, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:27:17,014 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=844746.0, ans=0.125
2024-09-25 22:27:36,743 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.74 vs. limit=15.0
2024-09-25 22:27:46,468 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.86 vs. limit=12.0
2024-09-25 22:28:17,327 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=844932.6666666666, ans=0.125
2024-09-25 22:28:31,339 INFO [train.py:1198] (1/4) Epoch 47, batch 1850, loss[loss=0.188, ctc_loss=0.12, cr_loss=0.3402, over 17179.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1186, cr_loss=0.3344, over 3358218.75 frames. ], batch size: 45, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:28:37,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=844979.3333333334, ans=0.125
2024-09-25 22:28:43,394 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.48 vs. limit=15.0
2024-09-25 22:28:49,121 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=845026.0, ans=0.125
2024-09-25 22:28:53,365 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.318e+02 1.395e+02 1.509e+02 1.851e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-25 22:28:56,991 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.93 vs. limit=15.0
2024-09-25 22:28:58,881 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=512, metric=3.32 vs. limit=15.0
2024-09-25 22:28:59,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=845026.0, ans=0.125
2024-09-25 22:29:01,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=845072.6666666666, ans=0.125
2024-09-25 22:29:08,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.prob, batch_count=845072.6666666666, ans=0.125
2024-09-25 22:29:21,524 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=7.40 vs. limit=15.0
2024-09-25 22:29:24,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=845119.3333333334, ans=0.05
2024-09-25 22:29:50,498 INFO [train.py:1198] (1/4) Epoch 47, batch 1900, loss[loss=0.164, ctc_loss=0.1039, cr_loss=0.3008, over 17254.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1192, cr_loss=0.3357, over 3351101.60 frames. ], batch size: 44, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:30:26,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=845306.0, ans=0.125
2024-09-25 22:30:38,064 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=845306.0, ans=0.125
2024-09-25 22:30:42,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=845352.6666666666, ans=0.035
2024-09-25 22:30:47,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=845352.6666666666, ans=0.125
2024-09-25 22:31:12,816 INFO [train.py:1198] (1/4) Epoch 47, batch 1950, loss[loss=0.2173, ctc_loss=0.1417, cr_loss=0.378, over 17203.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1195, cr_loss=0.3367, over 3365933.82 frames. ], batch size: 55, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:31:24,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=845446.0, ans=0.0
2024-09-25 22:31:35,175 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.279e+02 1.368e+02 1.443e+02 1.975e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-25 22:31:45,100 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module2.whiten, num_groups=1, num_channels=192, metric=13.09 vs. limit=15.0
2024-09-25 22:31:50,700 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=845539.3333333334, ans=0.125
2024-09-25 22:32:20,063 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.41 vs. limit=15.0
2024-09-25 22:32:38,558 INFO [train.py:1198] (1/4) Epoch 47, batch 2000, loss[loss=0.1682, ctc_loss=0.105, cr_loss=0.3156, over 17241.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.1197, cr_loss=0.3375, over 3360235.18 frames. ], batch size: 44, lr: 2.52e-03, grad_scale: 32.0
2024-09-25 22:32:57,396 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.61 vs. limit=15.0
2024-09-25 22:33:21,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=845772.6666666666, ans=0.1
2024-09-25 22:33:26,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=845772.6666666666, ans=0.0
2024-09-25 22:33:45,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=845866.0, ans=0.125
2024-09-25 22:34:01,510 INFO [train.py:1198] (1/4) Epoch 47, batch 2050, loss[loss=0.1772, ctc_loss=0.1109, cr_loss=0.3316, over 17142.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.12, cr_loss=0.3379, over 3354435.06 frames. ], batch size: 48, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:34:09,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=845912.6666666666, ans=0.0
2024-09-25 22:34:25,350 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.271e+02 1.350e+02 1.448e+02 1.978e+02, threshold=2.700e+02, percent-clipped=0.0
2024-09-25 22:34:27,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=845959.3333333334, ans=0.015
2024-09-25 22:34:28,906 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=845959.3333333334, ans=0.2
2024-09-25 22:34:33,837 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=846006.0, ans=0.125
2024-09-25 22:34:49,693 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=846052.6666666666, ans=0.025
2024-09-25 22:34:58,353 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.75 vs. limit=15.0
2024-09-25 22:34:59,333 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=846052.6666666666, ans=0.125
2024-09-25 22:35:02,558 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=846052.6666666666, ans=0.125
2024-09-25 22:35:23,976 INFO [train.py:1198] (1/4) Epoch 47, batch 2100, loss[loss=0.238, ctc_loss=0.154, cr_loss=0.4201, over 15054.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1192, cr_loss=0.3364, over 3360124.97 frames. ], batch size: 89, lr: 2.52e-03, grad_scale: 8.0
2024-09-25 22:35:43,711 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=846192.6666666666, ans=0.0
2024-09-25 22:36:01,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=846239.3333333334, ans=0.015
2024-09-25 22:36:01,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.min_positive, batch_count=846239.3333333334, ans=0.05
2024-09-25 22:36:21,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=846286.0, ans=0.0
2024-09-25 22:36:25,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=846286.0, ans=0.0
2024-09-25 22:36:25,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=846286.0, ans=0.125
2024-09-25 22:36:36,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=846332.6666666666, ans=0.125
2024-09-25 22:36:46,573 INFO [train.py:1198] (1/4) Epoch 47, batch 2150, loss[loss=0.1874, ctc_loss=0.1188, cr_loss=0.3429, over 17139.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1191, cr_loss=0.3358, over 3356059.56 frames. ], batch size: 45, lr: 2.52e-03, grad_scale: 8.0
2024-09-25 22:36:55,720 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.prob, batch_count=846379.3333333334, ans=0.125
2024-09-25 22:37:14,615 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.323e+02 1.402e+02 1.477e+02 1.932e+02, threshold=2.805e+02, percent-clipped=0.0
2024-09-25 22:37:59,611 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=846566.0, ans=0.2
2024-09-25 22:38:11,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=846612.6666666666, ans=0.2
2024-09-25 22:38:12,284 INFO [train.py:1198] (1/4) Epoch 47, batch 2200, loss[loss=0.1839, ctc_loss=0.1178, cr_loss=0.3302, over 17306.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1199, cr_loss=0.3373, over 3342436.26 frames. ], batch size: 49, lr: 2.52e-03, grad_scale: 8.0
2024-09-25 22:38:23,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=846612.6666666666, ans=0.1
2024-09-25 22:38:46,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=846706.0, ans=0.04949747468305833
2024-09-25 22:39:16,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=846799.3333333334, ans=0.0
2024-09-25 22:39:32,144 INFO [train.py:1198] (1/4) Epoch 47, batch 2250, loss[loss=0.2172, ctc_loss=0.1408, cr_loss=0.382, over 17134.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1189, cr_loss=0.3351, over 3350971.29 frames. ], batch size: 48, lr: 2.52e-03, grad_scale: 8.0
2024-09-25 22:39:58,070 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.312e+02 1.382e+02 1.535e+02 2.267e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-25 22:40:31,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=846986.0, ans=0.125
2024-09-25 22:40:55,216 INFO [train.py:1198] (1/4) Epoch 47, batch 2300, loss[loss=0.1784, ctc_loss=0.115, cr_loss=0.3169, over 17318.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1193, cr_loss=0.336, over 3346942.21 frames. ], batch size: 51, lr: 2.52e-03, grad_scale: 8.0
2024-09-25 22:40:58,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=847079.3333333334, ans=0.2
2024-09-25 22:41:03,750 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=847079.3333333334, ans=0.07
2024-09-25 22:42:05,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=847266.0, ans=0.125
2024-09-25 22:42:12,072 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.50 vs. limit=15.0
2024-09-25 22:42:20,987 INFO [train.py:1198] (1/4) Epoch 47, batch 2350, loss[loss=0.1632, ctc_loss=0.1035, cr_loss=0.2987, over 17257.00 frames. ], tot_loss[loss=0.187, ctc_loss=0.1196, cr_loss=0.3368, over 3344907.17 frames. ], batch size: 44, lr: 2.52e-03, grad_scale: 8.0
2024-09-25 22:42:38,012 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.65 vs. limit=15.0
2024-09-25 22:42:46,752 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.178e+02 1.289e+02 1.363e+02 1.471e+02 2.156e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-25 22:42:47,020 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=847359.3333333334, ans=0.125
2024-09-25 22:42:49,084 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.12 vs. limit=22.5
2024-09-25 22:43:09,464 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.35 vs. limit=15.0
2024-09-25 22:43:43,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=847546.0, ans=0.125
2024-09-25 22:43:43,183 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:43:43,193 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=847546.0, ans=0.125
2024-09-25 22:43:44,391 INFO [train.py:1198] (1/4) Epoch 47, batch 2400, loss[loss=0.1767, ctc_loss=0.1124, cr_loss=0.3214, over 17287.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.12, cr_loss=0.3369, over 3330890.54 frames. ], batch size: 51, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:44:02,231 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=847592.6666666666, ans=0.0
2024-09-25 22:44:34,377 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=847686.0, ans=0.0
2024-09-25 22:45:07,169 INFO [train.py:1198] (1/4) Epoch 47, batch 2450, loss[loss=0.1703, ctc_loss=0.1065, cr_loss=0.3191, over 17152.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1198, cr_loss=0.3363, over 3323173.13 frames. ], batch size: 45, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:45:18,875 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=847779.3333333334, ans=0.0
2024-09-25 22:45:25,091 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=847826.0, ans=0.0
2024-09-25 22:45:26,653 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=847826.0, ans=0.125
2024-09-25 22:45:32,659 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.309e+02 1.414e+02 1.505e+02 2.738e+02, threshold=2.828e+02, percent-clipped=1.0
2024-09-25 22:46:18,621 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=11.66 vs. limit=15.0
2024-09-25 22:46:27,292 INFO [train.py:1198] (1/4) Epoch 47, batch 2500, loss[loss=0.18, ctc_loss=0.1121, cr_loss=0.3391, over 17229.00 frames. ], tot_loss[loss=0.1877, ctc_loss=0.1201, cr_loss=0.3378, over 3325502.34 frames. ], batch size: 50, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:46:38,728 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.94 vs. limit=22.5
2024-09-25 22:46:41,725 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.25 vs. limit=15.0
2024-09-25 22:46:44,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=848059.3333333334, ans=0.04949747468305833
2024-09-25 22:46:49,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=848059.3333333334, ans=0.035
2024-09-25 22:47:14,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.attention_skip_rate, batch_count=848106.0, ans=0.0
2024-09-25 22:47:19,295 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer2.prob, batch_count=848152.6666666666, ans=0.125
2024-09-25 22:47:55,380 INFO [train.py:1198] (1/4) Epoch 47, batch 2550, loss[loss=0.1941, ctc_loss=0.1235, cr_loss=0.353, over 17103.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1199, cr_loss=0.337, over 3332405.58 frames. ], batch size: 49, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:48:11,672 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=848292.6666666666, ans=0.025
2024-09-25 22:48:20,933 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.321e+02 1.406e+02 1.538e+02 2.240e+02, threshold=2.812e+02, percent-clipped=0.0
2024-09-25 22:48:22,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=848292.6666666666, ans=0.1
2024-09-25 22:48:53,559 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=848386.0, ans=0.125
2024-09-25 22:49:15,635 INFO [train.py:1198] (1/4) Epoch 47, batch 2600, loss[loss=0.2155, ctc_loss=0.1399, cr_loss=0.3781, over 16561.00 frames. ], tot_loss[loss=0.1874, ctc_loss=0.1199, cr_loss=0.3374, over 3331676.98 frames. ], batch size: 66, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:49:25,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=848479.3333333334, ans=0.1
2024-09-25 22:49:57,373 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.59 vs. limit=15.0
2024-09-25 22:50:05,302 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.23 vs. limit=22.5
2024-09-25 22:50:22,537 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=848666.0, ans=0.0
2024-09-25 22:50:38,208 INFO [train.py:1198] (1/4) Epoch 47, batch 2650, loss[loss=0.2056, ctc_loss=0.1367, cr_loss=0.3446, over 17003.00 frames. ], tot_loss[loss=0.1875, ctc_loss=0.12, cr_loss=0.3372, over 3333167.95 frames. ], batch size: 51, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:50:49,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=848712.6666666666, ans=0.125
2024-09-25 22:50:59,483 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=848759.3333333334, ans=0.2
2024-09-25 22:51:03,872 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.306e+02 1.404e+02 1.475e+02 2.254e+02, threshold=2.808e+02, percent-clipped=0.0
2024-09-25 22:51:26,926 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=9.14 vs. limit=12.0
2024-09-25 22:51:48,809 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.44 vs. limit=15.0
2024-09-25 22:51:59,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=848899.3333333334, ans=0.125
2024-09-25 22:52:03,698 INFO [train.py:1198] (1/4) Epoch 47, batch 2700, loss[loss=0.1634, ctc_loss=0.103, cr_loss=0.3019, over 17141.00 frames. ], tot_loss[loss=0.1871, ctc_loss=0.1197, cr_loss=0.3371, over 3350250.46 frames. ], batch size: 48, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:52:15,334 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:52:30,974 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=848992.6666666666, ans=0.125
2024-09-25 22:52:45,309 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.79 vs. limit=15.0
2024-09-25 22:53:16,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=849132.6666666666, ans=0.125
2024-09-25 22:53:25,684 INFO [train.py:1198] (1/4) Epoch 47, batch 2750, loss[loss=0.1658, ctc_loss=0.1038, cr_loss=0.3099, over 17264.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1196, cr_loss=0.3367, over 3350011.88 frames. ], batch size: 42, lr: 2.52e-03, grad_scale: 16.0
2024-09-25 22:53:27,560 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=849179.3333333334, ans=0.125
2024-09-25 22:53:51,198 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.348e+02 1.413e+02 1.514e+02 2.410e+02, threshold=2.827e+02, percent-clipped=0.0
2024-09-25 22:53:53,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=849226.0, ans=0.125
2024-09-25 22:53:58,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=849272.6666666666, ans=0.0
2024-09-25 22:54:01,063 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=849272.6666666666, ans=0.125
2024-09-25 22:54:45,976 INFO [train.py:1198] (1/4) Epoch 47, batch 2800, loss[loss=0.1693, ctc_loss=0.1083, cr_loss=0.3049, over 16566.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1194, cr_loss=0.3369, over 3351979.27 frames. ], batch size: 66, lr: 2.51e-03, grad_scale: 32.0
2024-09-25 22:54:47,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=849412.6666666666, ans=0.1
2024-09-25 22:55:01,867 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.02 vs. limit=22.5
2024-09-25 22:55:06,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=849459.3333333334, ans=0.125
2024-09-25 22:55:59,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=849599.3333333334, ans=0.125
2024-09-25 22:56:01,636 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=7.28 vs. limit=15.0
2024-09-25 22:56:08,515 INFO [train.py:1198] (1/4) Epoch 47, batch 2850, loss[loss=0.1553, ctc_loss=0.09653, cr_loss=0.294, over 17264.00 frames. ], tot_loss[loss=0.1872, ctc_loss=0.1198, cr_loss=0.3372, over 3349646.10 frames. ], batch size: 44, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 22:56:13,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=849646.0, ans=0.125
2024-09-25 22:56:38,076 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.322e+02 1.395e+02 1.477e+02 2.122e+02, threshold=2.790e+02, percent-clipped=0.0
2024-09-25 22:57:17,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=849832.6666666666, ans=0.125
2024-09-25 22:57:33,015 INFO [train.py:1198] (1/4) Epoch 47, batch 2900, loss[loss=0.1784, ctc_loss=0.1126, cr_loss=0.329, over 17139.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1194, cr_loss=0.3367, over 3352935.67 frames. ], batch size: 48, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 22:57:41,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=849879.3333333334, ans=0.0
2024-09-25 22:57:47,156 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 22:58:09,700 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.08 vs. limit=10.0
2024-09-25 22:58:15,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=849972.6666666666, ans=0.125
2024-09-25 22:58:18,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=849972.6666666666, ans=10.0
2024-09-25 22:58:24,430 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.74 vs. limit=12.0
2024-09-25 22:58:25,610 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=850019.3333333334, ans=0.125
2024-09-25 22:58:48,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=850066.0, ans=0.0
2024-09-25 22:58:56,180 INFO [train.py:1198] (1/4) Epoch 47, batch 2950, loss[loss=0.1664, ctc_loss=0.1057, cr_loss=0.3035, over 17243.00 frames. ], tot_loss[loss=0.1869, ctc_loss=0.1196, cr_loss=0.3369, over 3355642.56 frames. ], batch size: 44, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 22:58:59,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=850112.6666666666, ans=0.1
2024-09-25 22:59:22,000 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=850159.3333333334, ans=0.0
2024-09-25 22:59:24,952 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.302e+02 1.410e+02 1.499e+02 2.139e+02, threshold=2.820e+02, percent-clipped=0.0
2024-09-25 23:00:04,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=850299.3333333334, ans=0.125
2024-09-25 23:00:17,881 INFO [train.py:1198] (1/4) Epoch 47, batch 3000, loss[loss=0.1699, ctc_loss=0.106, cr_loss=0.3196, over 16914.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.119, cr_loss=0.336, over 3359677.33 frames. ], batch size: 42, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:00:17,882 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 23:00:33,703 INFO [train.py:1230] (1/4) Epoch 47, validation: loss=0.0348, ctc_loss=0.0348, cr_loss=1.036e-14, over 944034.00 frames. 
2024-09-25 23:00:33,704 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 23:00:57,879 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=850392.6666666666, ans=0.125
2024-09-25 23:01:01,011 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=850392.6666666666, ans=0.125
2024-09-25 23:01:01,324 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.69 vs. limit=10.0
2024-09-25 23:01:52,183 INFO [train.py:1198] (1/4) Epoch 47, batch 3050, loss[loss=0.1696, ctc_loss=0.1052, cr_loss=0.3216, over 17224.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1192, cr_loss=0.3364, over 3358468.14 frames. ], batch size: 47, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:02:11,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=850626.0, ans=0.1
2024-09-25 23:02:20,305 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.305e+02 1.413e+02 1.506e+02 4.087e+02, threshold=2.825e+02, percent-clipped=1.0
2024-09-25 23:02:22,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=850672.6666666666, ans=0.0
2024-09-25 23:02:22,192 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=850672.6666666666, ans=0.0
2024-09-25 23:02:36,514 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.max_abs, batch_count=850672.6666666666, ans=10.0
2024-09-25 23:02:39,570 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=850719.3333333334, ans=0.1
2024-09-25 23:02:44,383 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=850719.3333333334, ans=0.125
2024-09-25 23:03:13,093 INFO [train.py:1198] (1/4) Epoch 47, batch 3100, loss[loss=0.1986, ctc_loss=0.1282, cr_loss=0.3518, over 17304.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.119, cr_loss=0.3361, over 3359973.33 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:03:13,317 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=850812.6666666666, ans=0.2
2024-09-25 23:03:16,943 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=5.91 vs. limit=15.0
2024-09-25 23:03:20,047 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.41 vs. limit=15.0
2024-09-25 23:03:20,449 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.22 vs. limit=15.0
2024-09-25 23:03:24,249 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=850812.6666666666, ans=0.125
2024-09-25 23:03:32,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=850859.3333333334, ans=0.125
2024-09-25 23:03:32,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=850859.3333333334, ans=0.125
2024-09-25 23:03:37,681 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:03:44,668 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=192, metric=5.91 vs. limit=15.0
2024-09-25 23:03:45,320 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=850906.0, ans=0.1
2024-09-25 23:04:18,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=850999.3333333334, ans=0.0
2024-09-25 23:04:25,793 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=850999.3333333334, ans=0.125
2024-09-25 23:04:33,493 INFO [train.py:1198] (1/4) Epoch 47, batch 3150, loss[loss=0.1849, ctc_loss=0.1186, cr_loss=0.3314, over 17015.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1189, cr_loss=0.3352, over 3353278.69 frames. ], batch size: 53, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:04:46,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=851046.0, ans=0.2
2024-09-25 23:05:01,409 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.375e+02 1.455e+02 1.638e+02 2.123e+02, threshold=2.910e+02, percent-clipped=0.0
2024-09-25 23:05:16,365 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.51 vs. limit=12.0
2024-09-25 23:05:21,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=851186.0, ans=0.125
2024-09-25 23:05:25,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=851186.0, ans=0.0
2024-09-25 23:05:37,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=851232.6666666666, ans=0.125
2024-09-25 23:05:54,052 INFO [train.py:1198] (1/4) Epoch 47, batch 3200, loss[loss=0.2116, ctc_loss=0.1362, cr_loss=0.3768, over 17089.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1183, cr_loss=0.3345, over 3364002.12 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:06:17,731 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=851326.0, ans=0.125
2024-09-25 23:07:02,956 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=851466.0, ans=0.125
2024-09-25 23:07:06,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=851466.0, ans=0.0
2024-09-25 23:07:12,127 INFO [train.py:1198] (1/4) Epoch 47, batch 3250, loss[loss=0.1596, ctc_loss=0.1018, cr_loss=0.2893, over 17171.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1185, cr_loss=0.335, over 3362135.57 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:07:12,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=851512.6666666666, ans=0.2
2024-09-25 23:07:26,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=851559.3333333334, ans=0.2
2024-09-25 23:07:32,800 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=851559.3333333334, ans=0.2
2024-09-25 23:07:40,356 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.291e+02 1.374e+02 1.460e+02 1.837e+02, threshold=2.749e+02, percent-clipped=0.0
2024-09-25 23:08:06,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=851652.6666666666, ans=0.0
2024-09-25 23:08:30,208 INFO [train.py:1198] (1/4) Epoch 47, batch 3300, loss[loss=0.1669, ctc_loss=0.1054, cr_loss=0.3073, over 16302.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1186, cr_loss=0.3345, over 3354202.23 frames. ], batch size: 36, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:09:15,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=851886.0, ans=0.5
2024-09-25 23:09:48,392 INFO [train.py:1198] (1/4) Epoch 47, batch 3350, loss[loss=0.1664, ctc_loss=0.1037, cr_loss=0.3131, over 16956.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1191, cr_loss=0.3355, over 3335803.03 frames. ], batch size: 42, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:09:55,073 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:09:57,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=851979.3333333334, ans=0.125
2024-09-25 23:10:04,303 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:10:15,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=852026.0, ans=0.1
2024-09-25 23:10:16,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.65 vs. limit=15.0
2024-09-25 23:10:20,045 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.314e+02 1.397e+02 1.497e+02 2.232e+02, threshold=2.793e+02, percent-clipped=0.0
2024-09-25 23:10:45,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=852119.3333333334, ans=0.125
2024-09-25 23:10:53,663 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.85 vs. limit=22.5
2024-09-25 23:11:02,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=852166.0, ans=0.0
2024-09-25 23:11:08,898 INFO [train.py:1198] (1/4) Epoch 47, batch 3400, loss[loss=0.2008, ctc_loss=0.1265, cr_loss=0.3711, over 17103.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1193, cr_loss=0.3364, over 3348526.31 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:11:32,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=852259.3333333334, ans=0.5
2024-09-25 23:11:39,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=852306.0, ans=0.2
2024-09-25 23:11:58,516 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.41 vs. limit=10.0
2024-09-25 23:12:10,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=852399.3333333334, ans=0.0
2024-09-25 23:12:18,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=852399.3333333334, ans=0.125
2024-09-25 23:12:25,949 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=852446.0, ans=0.125
2024-09-25 23:12:27,290 INFO [train.py:1198] (1/4) Epoch 47, batch 3450, loss[loss=0.1884, ctc_loss=0.1203, cr_loss=0.3405, over 17300.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1183, cr_loss=0.3344, over 3349214.77 frames. ], batch size: 51, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:12:46,144 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.min_positive, batch_count=852492.6666666666, ans=0.05
2024-09-25 23:12:56,966 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.319e+02 1.407e+02 1.479e+02 2.069e+02, threshold=2.813e+02, percent-clipped=0.0
2024-09-25 23:12:58,732 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=852539.3333333334, ans=0.125
2024-09-25 23:13:06,627 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=852539.3333333334, ans=0.0
2024-09-25 23:13:12,549 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=852586.0, ans=0.125
2024-09-25 23:13:19,369 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=852586.0, ans=0.125
2024-09-25 23:13:49,279 INFO [train.py:1198] (1/4) Epoch 47, batch 3500, loss[loss=0.1669, ctc_loss=0.1052, cr_loss=0.3089, over 17025.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1185, cr_loss=0.3348, over 3355684.51 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:14:03,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=852726.0, ans=0.1
2024-09-25 23:14:08,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=852726.0, ans=0.125
2024-09-25 23:14:47,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=852819.3333333334, ans=0.125
2024-09-25 23:14:47,221 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=852819.3333333334, ans=0.125
2024-09-25 23:14:58,169 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:15:07,125 INFO [train.py:1198] (1/4) Epoch 47, batch 3550, loss[loss=0.1872, ctc_loss=0.1198, cr_loss=0.3367, over 17112.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1189, cr_loss=0.335, over 3354312.50 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0
2024-09-25 23:15:10,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=852912.6666666666, ans=0.125
2024-09-25 23:15:18,964 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.15 vs. limit=15.0
2024-09-25 23:15:19,697 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=852912.6666666666, ans=0.125
2024-09-25 23:15:32,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=852959.3333333334, ans=0.125
2024-09-25 23:15:34,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=852959.3333333334, ans=0.0
2024-09-25 23:15:36,744 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.whiten_keys.whitening_limit, batch_count=852959.3333333334, ans=6.0
2024-09-25 23:15:38,866 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.284e+02 1.338e+02 1.473e+02 3.313e+02, threshold=2.677e+02, percent-clipped=1.0
2024-09-25 23:16:01,308 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=853052.6666666666, ans=0.0
2024-09-25 23:16:12,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=853099.3333333334, ans=0.1
2024-09-25 23:16:15,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=853099.3333333334, ans=0.1
2024-09-25 23:16:27,425 INFO [train.py:1198] (1/4) Epoch 47, batch 3600, loss[loss=0.2086, ctc_loss=0.1358, cr_loss=0.3639, over 15124.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1184, cr_loss=0.3341, over 3361821.88 frames. ], batch size: 89, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:16:40,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=853146.0, ans=0.125
2024-09-25 23:16:52,184 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=853192.6666666666, ans=0.0
2024-09-25 23:16:56,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=853239.3333333334, ans=0.0
2024-09-25 23:17:12,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=853286.0, ans=0.0
2024-09-25 23:17:22,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=853286.0, ans=0.125
2024-09-25 23:17:45,126 INFO [train.py:1198] (1/4) Epoch 47, batch 3650, loss[loss=0.1775, ctc_loss=0.1118, cr_loss=0.3287, over 17138.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.119, cr_loss=0.3356, over 3356022.55 frames. ], batch size: 48, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:17:56,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=853379.3333333334, ans=0.0
2024-09-25 23:18:01,722 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.26 vs. limit=15.0
2024-09-25 23:18:14,971 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.311e+02 1.426e+02 1.502e+02 2.686e+02, threshold=2.852e+02, percent-clipped=1.0
2024-09-25 23:18:55,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=853566.0, ans=0.0
2024-09-25 23:19:04,231 INFO [train.py:1198] (1/4) Epoch 47, batch 3700, loss[loss=0.1887, ctc_loss=0.1213, cr_loss=0.3369, over 16906.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1185, cr_loss=0.3337, over 3342839.13 frames. ], batch size: 58, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:19:33,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=853659.3333333334, ans=0.2
2024-09-25 23:19:57,428 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.41 vs. limit=15.0
2024-09-25 23:20:11,328 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=853799.3333333334, ans=0.125
2024-09-25 23:20:11,870 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.27 vs. limit=12.0
2024-09-25 23:20:23,764 INFO [train.py:1198] (1/4) Epoch 47, batch 3750, loss[loss=0.1787, ctc_loss=0.1126, cr_loss=0.3303, over 16960.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1188, cr_loss=0.3341, over 3340373.82 frames. ], batch size: 42, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:20:38,418 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.62 vs. limit=22.5
2024-09-25 23:20:53,580 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.332e+02 1.393e+02 1.534e+02 1.821e+02, threshold=2.786e+02, percent-clipped=0.0
2024-09-25 23:21:43,156 INFO [train.py:1198] (1/4) Epoch 47, batch 3800, loss[loss=0.1995, ctc_loss=0.1279, cr_loss=0.3583, over 17317.00 frames. ], tot_loss[loss=0.1853, ctc_loss=0.1185, cr_loss=0.3338, over 3343923.91 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:21:45,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=854079.3333333334, ans=0.0
2024-09-25 23:21:50,469 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=854079.3333333334, ans=22.5
2024-09-25 23:21:56,312 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=854079.3333333334, ans=0.2
2024-09-25 23:21:59,416 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=854126.0, ans=0.0
2024-09-25 23:22:10,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=854126.0, ans=0.1
2024-09-25 23:22:53,455 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=17.97 vs. limit=22.5
2024-09-25 23:23:00,632 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=854266.0, ans=0.2
2024-09-25 23:23:03,554 INFO [train.py:1198] (1/4) Epoch 47, batch 3850, loss[loss=0.206, ctc_loss=0.1374, cr_loss=0.3434, over 11421.00 frames. ], tot_loss[loss=0.1838, ctc_loss=0.1175, cr_loss=0.3312, over 3323817.55 frames. ], batch size: 123, lr: 2.51e-03, grad_scale: 16.0
2024-09-25 23:23:10,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=854312.6666666666, ans=0.125
2024-09-25 23:23:17,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=854359.3333333334, ans=0.125
2024-09-25 23:23:32,701 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.338e+02 1.449e+02 1.615e+02 3.869e+02, threshold=2.899e+02, percent-clipped=2.0
2024-09-25 23:23:39,454 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=9.67 vs. limit=12.0
2024-09-25 23:23:52,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=854452.6666666666, ans=0.125
2024-09-25 23:25:01,209 INFO [train.py:1198] (1/4) Epoch 48, batch 0, loss[loss=0.2009, ctc_loss=0.1297, cr_loss=0.3557, over 16040.00 frames. ], tot_loss[loss=0.2009, ctc_loss=0.1297, cr_loss=0.3557, over 16040.00 frames. ], batch size: 74, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:25:01,209 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-25 23:25:16,468 INFO [train.py:1230] (1/4) Epoch 48, validation: loss=0.0347, ctc_loss=0.0347, cr_loss=1.045e-14, over 944034.00 frames. 
2024-09-25 23:25:16,468 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-25 23:25:29,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=854527.3333333334, ans=0.125
2024-09-25 23:25:44,475 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:25:49,595 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.09 vs. limit=22.5
2024-09-25 23:25:57,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=854620.6666666666, ans=0.125
2024-09-25 23:26:05,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.46 vs. limit=15.0
2024-09-25 23:26:23,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=854714.0, ans=0.125
2024-09-25 23:26:38,951 INFO [train.py:1198] (1/4) Epoch 48, batch 50, loss[loss=0.1617, ctc_loss=0.1013, cr_loss=0.3023, over 17269.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1172, cr_loss=0.3346, over 761924.81 frames. ], batch size: 42, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:27:12,518 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=854854.0, ans=0.125
2024-09-25 23:27:16,936 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.326e+02 1.422e+02 1.657e+02 2.406e+02, threshold=2.844e+02, percent-clipped=0.0
2024-09-25 23:27:22,004 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=854854.0, ans=0.125
2024-09-25 23:27:51,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=854947.3333333334, ans=0.125
2024-09-25 23:27:53,582 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=5.05 vs. limit=15.0
2024-09-25 23:27:57,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=854947.3333333334, ans=0.125
2024-09-25 23:28:00,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=854994.0, ans=0.5
2024-09-25 23:28:02,160 INFO [train.py:1198] (1/4) Epoch 48, batch 100, loss[loss=0.1896, ctc_loss=0.1208, cr_loss=0.3441, over 17294.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1188, cr_loss=0.3369, over 1342212.19 frames. ], batch size: 49, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:28:15,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=854994.0, ans=0.0
2024-09-25 23:28:23,245 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=855040.6666666666, ans=0.125
2024-09-25 23:28:37,922 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=855087.3333333334, ans=0.125
2024-09-25 23:28:47,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=855087.3333333334, ans=0.125
2024-09-25 23:29:06,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=855134.0, ans=0.125
2024-09-25 23:29:09,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.min_positive, batch_count=855180.6666666666, ans=0.05
2024-09-25 23:29:14,627 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.02 vs. limit=22.5
2024-09-25 23:29:25,170 INFO [train.py:1198] (1/4) Epoch 48, batch 150, loss[loss=0.2213, ctc_loss=0.1431, cr_loss=0.3907, over 15144.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1188, cr_loss=0.3361, over 1784611.50 frames. ], batch size: 89, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:29:50,499 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=855274.0, ans=0.125
2024-09-25 23:30:05,999 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.318e+02 1.406e+02 1.496e+02 3.141e+02, threshold=2.812e+02, percent-clipped=1.0
2024-09-25 23:30:38,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=855414.0, ans=0.1
2024-09-25 23:30:47,887 INFO [train.py:1198] (1/4) Epoch 48, batch 200, loss[loss=0.1626, ctc_loss=0.1021, cr_loss=0.3026, over 17027.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1178, cr_loss=0.3334, over 2135229.75 frames. ], batch size: 39, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:31:08,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=855507.3333333334, ans=0.0
2024-09-25 23:31:13,292 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=855507.3333333334, ans=0.125
2024-09-25 23:31:24,444 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=855554.0, ans=0.1
2024-09-25 23:31:25,097 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1.whitening_limit, batch_count=855554.0, ans=10.0
2024-09-25 23:32:01,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=855647.3333333334, ans=0.07
2024-09-25 23:32:02,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=855647.3333333334, ans=0.0
2024-09-25 23:32:10,689 INFO [train.py:1198] (1/4) Epoch 48, batch 250, loss[loss=0.1457, ctc_loss=0.09054, cr_loss=0.2756, over 16962.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1186, cr_loss=0.3344, over 2394519.20 frames. ], batch size: 42, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:32:20,887 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=19.55 vs. limit=22.5
2024-09-25 23:32:29,340 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.78 vs. limit=15.0
2024-09-25 23:32:49,017 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.307e+02 1.374e+02 1.472e+02 2.103e+02, threshold=2.747e+02, percent-clipped=0.0
2024-09-25 23:33:21,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=855880.6666666666, ans=0.07
2024-09-25 23:33:26,878 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=5.57 vs. limit=12.0
2024-09-25 23:33:31,140 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:33:33,797 INFO [train.py:1198] (1/4) Epoch 48, batch 300, loss[loss=0.191, ctc_loss=0.119, cr_loss=0.3598, over 17206.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1193, cr_loss=0.3367, over 2610509.58 frames. ], batch size: 47, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:33:38,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=855927.3333333334, ans=0.0
2024-09-25 23:33:41,796 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:33:55,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=855974.0, ans=0.125
2024-09-25 23:34:26,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=856067.3333333334, ans=0.125
2024-09-25 23:34:37,400 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=856067.3333333334, ans=0.1
2024-09-25 23:34:37,540 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=856067.3333333334, ans=0.0
2024-09-25 23:34:54,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=856114.0, ans=0.1
2024-09-25 23:34:59,419 INFO [train.py:1198] (1/4) Epoch 48, batch 350, loss[loss=0.1827, ctc_loss=0.1134, cr_loss=0.3461, over 17025.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1189, cr_loss=0.3364, over 2783136.68 frames. ], batch size: 44, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:35:09,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=856160.6666666666, ans=0.125
2024-09-25 23:35:14,978 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.whiten, num_groups=1, num_channels=192, metric=5.16 vs. limit=12.0
2024-09-25 23:35:23,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=856207.3333333334, ans=0.125
2024-09-25 23:35:26,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=856207.3333333334, ans=0.125
2024-09-25 23:35:37,611 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.298e+02 1.363e+02 1.438e+02 2.006e+02, threshold=2.725e+02, percent-clipped=0.0
2024-09-25 23:35:44,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=856254.0, ans=0.125
2024-09-25 23:35:53,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=856300.6666666666, ans=0.0
2024-09-25 23:36:17,602 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer1.prob, batch_count=856347.3333333334, ans=0.125
2024-09-25 23:36:20,907 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=856394.0, ans=0.025
2024-09-25 23:36:22,265 INFO [train.py:1198] (1/4) Epoch 48, batch 400, loss[loss=0.1849, ctc_loss=0.1183, cr_loss=0.333, over 17031.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1192, cr_loss=0.3362, over 2912604.24 frames. ], batch size: 51, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:36:45,729 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.16 vs. limit=22.5
2024-09-25 23:36:51,545 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=856440.6666666666, ans=0.0
2024-09-25 23:36:57,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=856487.3333333334, ans=0.125
2024-09-25 23:37:10,664 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=856534.0, ans=0.125
2024-09-25 23:37:24,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=856580.6666666666, ans=0.125
2024-09-25 23:37:30,583 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.39 vs. limit=15.0
2024-09-25 23:37:37,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=856580.6666666666, ans=0.2
2024-09-25 23:37:41,590 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=856627.3333333334, ans=15.0
2024-09-25 23:37:42,216 INFO [train.py:1198] (1/4) Epoch 48, batch 450, loss[loss=0.1775, ctc_loss=0.1137, cr_loss=0.3191, over 17326.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1188, cr_loss=0.3357, over 3005650.87 frames. ], batch size: 48, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:37:45,680 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=856627.3333333334, ans=0.0
2024-09-25 23:37:53,573 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=856627.3333333334, ans=0.125
2024-09-25 23:37:56,801 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=856674.0, ans=0.0
2024-09-25 23:38:01,183 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:38:18,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=856720.6666666666, ans=0.2
2024-09-25 23:38:23,492 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.314e+02 1.437e+02 1.529e+02 1.990e+02, threshold=2.873e+02, percent-clipped=0.0
2024-09-25 23:38:23,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=856720.6666666666, ans=0.2
2024-09-25 23:39:07,875 INFO [train.py:1198] (1/4) Epoch 48, batch 500, loss[loss=0.2064, ctc_loss=0.1325, cr_loss=0.3697, over 15082.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3352, over 3090079.51 frames. ], batch size: 89, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:39:11,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=856860.6666666666, ans=0.025
2024-09-25 23:39:29,194 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.39 vs. limit=10.0
2024-09-25 23:39:35,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=856907.3333333334, ans=0.0
2024-09-25 23:39:38,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=856954.0, ans=0.0
2024-09-25 23:40:30,797 INFO [train.py:1198] (1/4) Epoch 48, batch 550, loss[loss=0.2113, ctc_loss=0.1388, cr_loss=0.3625, over 16938.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1185, cr_loss=0.3347, over 3151026.94 frames. ], batch size: 58, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:40:34,714 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=4.79 vs. limit=15.0
2024-09-25 23:41:11,883 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.319e+02 1.428e+02 1.533e+02 1.901e+02, threshold=2.856e+02, percent-clipped=0.0
2024-09-25 23:41:53,734 INFO [train.py:1198] (1/4) Epoch 48, batch 600, loss[loss=0.1798, ctc_loss=0.1136, cr_loss=0.3311, over 17303.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1187, cr_loss=0.3353, over 3197611.32 frames. ], batch size: 49, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:41:54,591 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=4.95 vs. limit=6.0
2024-09-25 23:41:58,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=857327.3333333334, ans=0.125
2024-09-25 23:42:00,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=857327.3333333334, ans=0.2
2024-09-25 23:42:13,220 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=857374.0, ans=0.1
2024-09-25 23:42:24,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=857420.6666666666, ans=0.1
2024-09-25 23:42:34,206 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=857420.6666666666, ans=0.125
2024-09-25 23:42:35,675 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=857420.6666666666, ans=0.125
2024-09-25 23:42:39,030 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=857420.6666666666, ans=0.125
2024-09-25 23:43:16,491 INFO [train.py:1198] (1/4) Epoch 48, batch 650, loss[loss=0.2161, ctc_loss=0.1403, cr_loss=0.379, over 17234.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1179, cr_loss=0.3344, over 3235096.19 frames. ], batch size: 55, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:43:26,264 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=857560.6666666666, ans=0.2
2024-09-25 23:43:31,129 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.attention_skip_rate, batch_count=857607.3333333334, ans=0.0
2024-09-25 23:43:44,013 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=857607.3333333334, ans=0.07
2024-09-25 23:43:48,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=857654.0, ans=0.1
2024-09-25 23:43:49,139 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=4.84 vs. limit=12.0
2024-09-25 23:43:56,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=857654.0, ans=0.2
2024-09-25 23:43:59,395 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.322e+02 1.414e+02 1.494e+02 2.236e+02, threshold=2.829e+02, percent-clipped=0.0
2024-09-25 23:44:02,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.layerdrop_rate, batch_count=857654.0, ans=0.015
2024-09-25 23:44:17,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=857700.6666666666, ans=0.2
2024-09-25 23:44:40,139 INFO [train.py:1198] (1/4) Epoch 48, batch 700, loss[loss=0.1654, ctc_loss=0.1044, cr_loss=0.305, over 17124.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1177, cr_loss=0.3344, over 3270504.11 frames. ], batch size: 40, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:44:55,690 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=857794.0, ans=0.125
2024-09-25 23:45:02,096 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:45:05,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=857840.6666666666, ans=0.125
2024-09-25 23:45:26,651 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.06 vs. limit=22.5
2024-09-25 23:45:50,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=857980.6666666666, ans=0.1
2024-09-25 23:45:54,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=857980.6666666666, ans=0.0
2024-09-25 23:46:05,006 INFO [train.py:1198] (1/4) Epoch 48, batch 750, loss[loss=0.2221, ctc_loss=0.1412, cr_loss=0.4042, over 15918.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1184, cr_loss=0.3357, over 3296110.72 frames. ], batch size: 74, lr: 2.48e-03, grad_scale: 16.0
2024-09-25 23:46:06,865 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=858027.3333333334, ans=0.2
2024-09-25 23:46:08,726 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:46:10,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.03 vs. limit=12.0
2024-09-25 23:46:13,248 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=858027.3333333334, ans=0.125
2024-09-25 23:46:14,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=858027.3333333334, ans=0.1
2024-09-25 23:46:24,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=858074.0, ans=0.025
2024-09-25 23:46:24,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=858074.0, ans=0.025
2024-09-25 23:46:30,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=858074.0, ans=0.0
2024-09-25 23:46:44,963 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.340e+02 1.448e+02 1.538e+02 3.559e+02, threshold=2.895e+02, percent-clipped=1.0
2024-09-25 23:47:02,735 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer1.prob, batch_count=858167.3333333334, ans=0.125
2024-09-25 23:47:12,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=858214.0, ans=0.125
2024-09-25 23:47:25,136 INFO [train.py:1198] (1/4) Epoch 48, batch 800, loss[loss=0.1583, ctc_loss=0.09819, cr_loss=0.3003, over 16979.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.119, cr_loss=0.3365, over 3310038.43 frames. ], batch size: 42, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:47:43,384 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.2.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:47:51,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=858307.3333333334, ans=0.125
2024-09-25 23:47:52,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=858307.3333333334, ans=0.125
2024-09-25 23:48:11,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=858354.0, ans=0.05
2024-09-25 23:48:19,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=4.10 vs. limit=10.0
2024-09-25 23:48:47,865 INFO [train.py:1198] (1/4) Epoch 48, batch 850, loss[loss=0.184, ctc_loss=0.1172, cr_loss=0.3339, over 17206.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1191, cr_loss=0.3366, over 3314509.15 frames. ], batch size: 47, lr: 2.48e-03, grad_scale: 32.0
2024-09-25 23:49:02,873 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=13.80 vs. limit=22.5
2024-09-25 23:49:03,930 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:49:07,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=858540.6666666666, ans=0.125
2024-09-25 23:49:16,829 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=858540.6666666666, ans=0.125
2024-09-25 23:49:30,828 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.305e+02 1.378e+02 1.479e+02 2.667e+02, threshold=2.756e+02, percent-clipped=0.0
2024-09-25 23:50:12,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=858680.6666666666, ans=0.125
2024-09-25 23:50:14,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=858727.3333333334, ans=0.125
2024-09-25 23:50:15,714 INFO [train.py:1198] (1/4) Epoch 48, batch 900, loss[loss=0.2101, ctc_loss=0.136, cr_loss=0.3704, over 16550.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1193, cr_loss=0.3374, over 3335150.92 frames. ], batch size: 66, lr: 2.47e-03, grad_scale: 32.0
2024-09-25 23:50:17,758 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.79 vs. limit=12.0
2024-09-25 23:50:19,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=858727.3333333334, ans=0.95
2024-09-25 23:50:20,706 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=858727.3333333334, ans=0.125
2024-09-25 23:50:22,548 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.73 vs. limit=15.0
2024-09-25 23:50:25,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=858727.3333333334, ans=0.125
2024-09-25 23:50:39,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=858774.0, ans=0.95
2024-09-25 23:50:49,182 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=858820.6666666666, ans=0.125
2024-09-25 23:51:11,046 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=858867.3333333334, ans=0.2
2024-09-25 23:51:33,125 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=858914.0, ans=0.125
2024-09-25 23:51:37,630 INFO [train.py:1198] (1/4) Epoch 48, batch 950, loss[loss=0.1813, ctc_loss=0.1164, cr_loss=0.3243, over 17049.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1192, cr_loss=0.3372, over 3340319.37 frames. ], batch size: 39, lr: 2.47e-03, grad_scale: 16.0
2024-09-25 23:51:44,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer1.prob, batch_count=858960.6666666666, ans=0.125
2024-09-25 23:51:47,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=858960.6666666666, ans=0.0
2024-09-25 23:51:49,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=858960.6666666666, ans=0.1
2024-09-25 23:52:19,388 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.296e+02 1.387e+02 1.508e+02 2.071e+02, threshold=2.775e+02, percent-clipped=0.0
2024-09-25 23:52:25,946 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=859100.6666666666, ans=0.2
2024-09-25 23:53:00,136 INFO [train.py:1198] (1/4) Epoch 48, batch 1000, loss[loss=0.1705, ctc_loss=0.1081, cr_loss=0.312, over 17227.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1191, cr_loss=0.337, over 3345243.17 frames. ], batch size: 50, lr: 2.47e-03, grad_scale: 16.0
2024-09-25 23:53:27,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=859240.6666666666, ans=0.125
2024-09-25 23:54:13,397 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-25 23:54:19,667 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=859380.6666666666, ans=0.125
2024-09-25 23:54:22,550 INFO [train.py:1198] (1/4) Epoch 48, batch 1050, loss[loss=0.1829, ctc_loss=0.1146, cr_loss=0.3415, over 17062.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1193, cr_loss=0.3372, over 3347814.91 frames. ], batch size: 46, lr: 2.47e-03, grad_scale: 16.0
2024-09-25 23:54:24,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=859427.3333333334, ans=0.0
2024-09-25 23:55:06,820 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.301e+02 1.364e+02 1.474e+02 5.159e+02, threshold=2.728e+02, percent-clipped=1.0
2024-09-25 23:55:11,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=859567.3333333334, ans=0.125
2024-09-25 23:55:13,698 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=859567.3333333334, ans=0.0
2024-09-25 23:55:39,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=859614.0, ans=0.125
2024-09-25 23:55:45,300 INFO [train.py:1198] (1/4) Epoch 48, batch 1100, loss[loss=0.2007, ctc_loss=0.1302, cr_loss=0.3524, over 17015.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1185, cr_loss=0.335, over 3355538.62 frames. ], batch size: 56, lr: 2.47e-03, grad_scale: 16.0
2024-09-25 23:55:52,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=859660.6666666666, ans=0.125
2024-09-25 23:56:09,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=859707.3333333334, ans=0.125
2024-09-25 23:56:12,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=859707.3333333334, ans=0.125
2024-09-25 23:57:07,815 INFO [train.py:1198] (1/4) Epoch 48, batch 1150, loss[loss=0.24, ctc_loss=0.1585, cr_loss=0.4072, over 16598.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.119, cr_loss=0.3357, over 3348931.65 frames. ], batch size: 66, lr: 2.47e-03, grad_scale: 16.0
2024-09-25 23:57:16,139 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=859894.0, ans=0.125
2024-09-25 23:57:34,046 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=12.73 vs. limit=15.0
2024-09-25 23:57:50,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.prob, batch_count=859987.3333333334, ans=0.125
2024-09-25 23:57:52,001 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.298e+02 1.394e+02 1.477e+02 1.999e+02, threshold=2.788e+02, percent-clipped=0.0
2024-09-25 23:57:52,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=859987.3333333334, ans=0.04949747468305833
2024-09-25 23:57:58,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=860034.0, ans=0.1
2024-09-25 23:58:00,273 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=860034.0, ans=0.0
2024-09-25 23:58:08,283 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=860034.0, ans=0.0
2024-09-25 23:58:19,323 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=860080.6666666666, ans=0.125
2024-09-25 23:58:30,254 INFO [train.py:1198] (1/4) Epoch 48, batch 1200, loss[loss=0.2208, ctc_loss=0.1455, cr_loss=0.3766, over 17318.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1193, cr_loss=0.3361, over 3354129.51 frames. ], batch size: 51, lr: 2.47e-03, grad_scale: 32.0
2024-09-25 23:58:43,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=860127.3333333334, ans=0.0
2024-09-25 23:58:43,222 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=860127.3333333334, ans=0.125
2024-09-25 23:59:11,739 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=860220.6666666666, ans=0.125
2024-09-25 23:59:21,993 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.49 vs. limit=12.0
2024-09-25 23:59:32,824 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=860267.3333333334, ans=0.125
2024-09-25 23:59:47,103 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.49 vs. limit=15.0
2024-09-25 23:59:56,069 INFO [train.py:1198] (1/4) Epoch 48, batch 1250, loss[loss=0.1808, ctc_loss=0.1166, cr_loss=0.3212, over 17045.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1195, cr_loss=0.3368, over 3357673.57 frames. ], batch size: 46, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:00:04,226 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=860360.6666666666, ans=0.0
2024-09-26 00:00:28,465 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=860454.0, ans=0.2
2024-09-26 00:00:39,299 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.302e+02 1.385e+02 1.504e+02 2.588e+02, threshold=2.770e+02, percent-clipped=0.0
2024-09-26 00:00:50,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=860500.6666666666, ans=0.125
2024-09-26 00:01:02,044 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=860547.3333333334, ans=0.025
2024-09-26 00:01:13,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=860547.3333333334, ans=0.1
2024-09-26 00:01:19,889 INFO [train.py:1198] (1/4) Epoch 48, batch 1300, loss[loss=0.1979, ctc_loss=0.1272, cr_loss=0.3533, over 17349.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1191, cr_loss=0.3362, over 3358492.78 frames. ], batch size: 48, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:01:20,149 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=860594.0, ans=0.125
2024-09-26 00:01:20,374 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 00:01:31,567 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 00:01:33,293 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=860594.0, ans=0.0
2024-09-26 00:01:39,702 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=860640.6666666666, ans=0.025
2024-09-26 00:01:41,389 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=860640.6666666666, ans=0.0
2024-09-26 00:01:41,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=860640.6666666666, ans=0.125
2024-09-26 00:01:46,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=860640.6666666666, ans=0.0
2024-09-26 00:01:57,520 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=860687.3333333334, ans=0.125
2024-09-26 00:02:10,733 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=860734.0, ans=0.1
2024-09-26 00:02:12,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten.whitening_limit, batch_count=860734.0, ans=22.5
2024-09-26 00:02:23,471 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=860780.6666666666, ans=0.125
2024-09-26 00:02:38,015 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=860780.6666666666, ans=0.0
2024-09-26 00:02:40,988 INFO [train.py:1198] (1/4) Epoch 48, batch 1350, loss[loss=0.2063, ctc_loss=0.1341, cr_loss=0.3606, over 17227.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.119, cr_loss=0.3354, over 3357357.80 frames. ], batch size: 55, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:02:42,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=860827.3333333334, ans=0.125
2024-09-26 00:02:48,661 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.28 vs. limit=15.0
2024-09-26 00:02:50,142 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.whiten, num_groups=1, num_channels=192, metric=4.08 vs. limit=12.0
2024-09-26 00:02:56,613 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=860827.3333333334, ans=0.025
2024-09-26 00:02:59,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=860874.0, ans=0.125
2024-09-26 00:03:12,707 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=860874.0, ans=0.125
2024-09-26 00:03:26,899 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.284e+02 1.347e+02 1.442e+02 2.015e+02, threshold=2.695e+02, percent-clipped=0.0
2024-09-26 00:03:35,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=860967.3333333334, ans=0.035
2024-09-26 00:03:45,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer1.prob, batch_count=860967.3333333334, ans=0.125
2024-09-26 00:04:07,153 INFO [train.py:1198] (1/4) Epoch 48, batch 1400, loss[loss=0.1963, ctc_loss=0.1275, cr_loss=0.3437, over 17153.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1188, cr_loss=0.3352, over 3354807.06 frames. ], batch size: 48, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:04:12,251 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=861060.6666666666, ans=0.125
2024-09-26 00:05:05,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.20 vs. limit=15.0
2024-09-26 00:05:28,970 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer2.prob, batch_count=861294.0, ans=0.125
2024-09-26 00:05:30,162 INFO [train.py:1198] (1/4) Epoch 48, batch 1450, loss[loss=0.2041, ctc_loss=0.1298, cr_loss=0.3714, over 16996.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.1189, cr_loss=0.3357, over 3358772.03 frames. ], batch size: 53, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:05:32,056 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=861294.0, ans=0.2
2024-09-26 00:05:35,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=861294.0, ans=0.125
2024-09-26 00:05:38,306 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=861294.0, ans=0.125
2024-09-26 00:05:54,654 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=861340.6666666666, ans=0.0
2024-09-26 00:06:15,972 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.346e+02 1.409e+02 1.541e+02 2.437e+02, threshold=2.817e+02, percent-clipped=0.0
2024-09-26 00:06:19,929 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten.whitening_limit, batch_count=861434.0, ans=15.0
2024-09-26 00:06:22,809 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=861434.0, ans=0.1
2024-09-26 00:06:29,275 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=861434.0, ans=0.125
2024-09-26 00:06:41,075 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=9.64 vs. limit=22.5
2024-09-26 00:06:52,970 INFO [train.py:1198] (1/4) Epoch 48, batch 1500, loss[loss=0.1962, ctc_loss=0.1255, cr_loss=0.3537, over 17256.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.1188, cr_loss=0.3363, over 3361493.73 frames. ], batch size: 44, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:07:00,070 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.27 vs. limit=15.0
2024-09-26 00:07:08,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.88 vs. limit=15.0
2024-09-26 00:07:36,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=861620.6666666666, ans=0.0
2024-09-26 00:07:38,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=861620.6666666666, ans=0.1
2024-09-26 00:08:12,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=861714.0, ans=0.2
2024-09-26 00:08:15,404 INFO [train.py:1198] (1/4) Epoch 48, batch 1550, loss[loss=0.1799, ctc_loss=0.1141, cr_loss=0.3289, over 17161.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.118, cr_loss=0.3344, over 3373316.68 frames. ], batch size: 45, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:08:42,539 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=861807.3333333334, ans=0.1
2024-09-26 00:08:44,712 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.38 vs. limit=10.0
2024-09-26 00:09:00,538 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.301e+02 1.350e+02 1.480e+02 2.232e+02, threshold=2.701e+02, percent-clipped=0.0
2024-09-26 00:09:02,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=861854.0, ans=0.125
2024-09-26 00:09:22,335 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.94 vs. limit=15.0
2024-09-26 00:09:37,074 INFO [train.py:1198] (1/4) Epoch 48, batch 1600, loss[loss=0.2008, ctc_loss=0.1302, cr_loss=0.3527, over 17042.00 frames. ], tot_loss[loss=0.1843, ctc_loss=0.1176, cr_loss=0.3334, over 3376860.50 frames. ], batch size: 52, lr: 2.47e-03, grad_scale: 32.0
2024-09-26 00:09:42,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass_mid.scale_min, batch_count=861994.0, ans=0.2
2024-09-26 00:10:22,178 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.01 vs. limit=15.0
2024-09-26 00:10:48,826 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=862180.6666666666, ans=0.125
2024-09-26 00:10:49,260 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=5.10 vs. limit=15.0
2024-09-26 00:11:01,993 INFO [train.py:1198] (1/4) Epoch 48, batch 1650, loss[loss=0.2066, ctc_loss=0.1345, cr_loss=0.3606, over 17068.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1179, cr_loss=0.3335, over 3379369.01 frames. ], batch size: 46, lr: 2.47e-03, grad_scale: 32.0
2024-09-26 00:11:07,244 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass_mid.scale_min, batch_count=862227.3333333334, ans=0.2
2024-09-26 00:11:22,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=862274.0, ans=0.125
2024-09-26 00:11:29,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=862274.0, ans=0.0
2024-09-26 00:11:47,277 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.306e+02 1.368e+02 1.453e+02 1.769e+02, threshold=2.736e+02, percent-clipped=0.0
2024-09-26 00:11:49,791 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.76 vs. limit=15.0
2024-09-26 00:12:22,677 INFO [train.py:1198] (1/4) Epoch 48, batch 1700, loss[loss=0.2028, ctc_loss=0.1284, cr_loss=0.3718, over 16570.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.1181, cr_loss=0.3345, over 3378240.49 frames. ], batch size: 66, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:12:24,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=862460.6666666666, ans=0.1
2024-09-26 00:12:43,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=862507.3333333334, ans=0.125
2024-09-26 00:12:57,397 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.hidden_balancer.prob, batch_count=862554.0, ans=0.125
2024-09-26 00:13:05,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=862554.0, ans=0.0
2024-09-26 00:13:07,199 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=862554.0, ans=0.125
2024-09-26 00:13:28,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=862647.3333333334, ans=0.0
2024-09-26 00:13:31,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=862647.3333333334, ans=0.125
2024-09-26 00:13:41,534 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.17 vs. limit=15.0
2024-09-26 00:13:45,463 INFO [train.py:1198] (1/4) Epoch 48, batch 1750, loss[loss=0.1831, ctc_loss=0.1153, cr_loss=0.3393, over 17150.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.119, cr_loss=0.3361, over 3369482.21 frames. ], batch size: 48, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:14:04,692 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.03 vs. limit=15.0
2024-09-26 00:14:31,806 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.57 vs. limit=15.0
2024-09-26 00:14:32,512 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.295e+02 1.382e+02 1.510e+02 2.646e+02, threshold=2.763e+02, percent-clipped=0.0
2024-09-26 00:14:55,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=862880.6666666666, ans=0.125
2024-09-26 00:15:09,852 INFO [train.py:1198] (1/4) Epoch 48, batch 1800, loss[loss=0.1971, ctc_loss=0.1272, cr_loss=0.3493, over 16855.00 frames. ], tot_loss[loss=0.1873, ctc_loss=0.1197, cr_loss=0.338, over 3365756.95 frames. ], batch size: 58, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:15:11,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=862927.3333333334, ans=0.1
2024-09-26 00:15:23,117 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=862927.3333333334, ans=0.2
2024-09-26 00:15:26,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=862974.0, ans=0.2
2024-09-26 00:15:29,883 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=7.18 vs. limit=15.0
2024-09-26 00:15:31,326 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=9.69 vs. limit=15.0
2024-09-26 00:16:02,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=863067.3333333334, ans=0.0
2024-09-26 00:16:05,447 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=863067.3333333334, ans=0.1
2024-09-26 00:16:14,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.skip_rate, batch_count=863114.0, ans=0.09899494936611666
2024-09-26 00:16:31,983 INFO [train.py:1198] (1/4) Epoch 48, batch 1850, loss[loss=0.1806, ctc_loss=0.114, cr_loss=0.3329, over 17006.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1201, cr_loss=0.3388, over 3362944.51 frames. ], batch size: 51, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:16:34,423 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=7.89 vs. limit=15.0
2024-09-26 00:16:35,553 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=863160.6666666666, ans=0.125
2024-09-26 00:16:46,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=863207.3333333334, ans=0.125
2024-09-26 00:16:48,754 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=512, metric=11.15 vs. limit=22.5
2024-09-26 00:17:03,331 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.70 vs. limit=6.0
2024-09-26 00:17:09,319 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.skip_rate, batch_count=863254.0, ans=0.04949747468305833
2024-09-26 00:17:11,244 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.78 vs. limit=15.0
2024-09-26 00:17:14,130 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=256, metric=18.43 vs. limit=22.5
2024-09-26 00:17:16,786 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.322e+02 1.421e+02 1.543e+02 2.119e+02, threshold=2.843e+02, percent-clipped=0.0
2024-09-26 00:17:54,932 INFO [train.py:1198] (1/4) Epoch 48, batch 1900, loss[loss=0.1633, ctc_loss=0.1026, cr_loss=0.3037, over 17220.00 frames. ], tot_loss[loss=0.1876, ctc_loss=0.12, cr_loss=0.3381, over 3346475.16 frames. ], batch size: 41, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:17:55,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=863394.0, ans=0.0
2024-09-26 00:18:02,317 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=5.40 vs. limit=15.0
2024-09-26 00:18:16,243 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=863440.6666666666, ans=0.0
2024-09-26 00:18:24,176 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=863440.6666666666, ans=0.125
2024-09-26 00:18:47,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=863534.0, ans=0.125
2024-09-26 00:18:53,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=863534.0, ans=0.2
2024-09-26 00:18:55,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=863534.0, ans=0.2
2024-09-26 00:19:00,797 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=3.51 vs. limit=6.0
2024-09-26 00:19:17,622 INFO [train.py:1198] (1/4) Epoch 48, batch 1950, loss[loss=0.2294, ctc_loss=0.1523, cr_loss=0.3857, over 15155.00 frames. ], tot_loss[loss=0.1878, ctc_loss=0.1201, cr_loss=0.3386, over 3342505.57 frames. ], batch size: 89, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:19:47,992 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=863674.0, ans=0.125
2024-09-26 00:20:00,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=863720.6666666666, ans=0.0
2024-09-26 00:20:05,198 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.335e+02 1.400e+02 1.522e+02 2.815e+02, threshold=2.800e+02, percent-clipped=0.0
2024-09-26 00:20:07,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward3.hidden_balancer.prob, batch_count=863767.3333333334, ans=0.125
2024-09-26 00:20:24,572 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer1.min_positive, batch_count=863814.0, ans=0.025
2024-09-26 00:20:37,784 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.whiten, num_groups=1, num_channels=512, metric=5.42 vs. limit=12.0
2024-09-26 00:20:40,168 INFO [train.py:1198] (1/4) Epoch 48, batch 2000, loss[loss=0.1514, ctc_loss=0.09135, cr_loss=0.3001, over 17124.00 frames. ], tot_loss[loss=0.1885, ctc_loss=0.1206, cr_loss=0.3395, over 3331708.08 frames. ], batch size: 40, lr: 2.47e-03, grad_scale: 32.0
2024-09-26 00:21:28,698 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.90 vs. limit=6.0
2024-09-26 00:21:54,624 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=864047.3333333334, ans=0.0
2024-09-26 00:22:02,510 INFO [train.py:1198] (1/4) Epoch 48, batch 2050, loss[loss=0.1558, ctc_loss=0.09869, cr_loss=0.2853, over 17262.00 frames. ], tot_loss[loss=0.1888, ctc_loss=0.1208, cr_loss=0.3399, over 3344957.55 frames. ], batch size: 42, lr: 2.47e-03, grad_scale: 32.0
2024-09-26 00:22:51,183 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.300e+02 1.355e+02 1.459e+02 2.186e+02, threshold=2.709e+02, percent-clipped=0.0
2024-09-26 00:23:25,053 INFO [train.py:1198] (1/4) Epoch 48, batch 2100, loss[loss=0.2217, ctc_loss=0.1419, cr_loss=0.3989, over 17041.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1207, cr_loss=0.3398, over 3347282.60 frames. ], batch size: 52, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:23:51,629 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.41 vs. limit=15.0
2024-09-26 00:23:57,277 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=864374.0, ans=0.1
2024-09-26 00:24:11,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=864420.6666666666, ans=0.0
2024-09-26 00:24:15,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=864467.3333333334, ans=0.125
2024-09-26 00:24:16,734 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.98 vs. limit=12.0
2024-09-26 00:24:30,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.skip_rate, batch_count=864467.3333333334, ans=0.09899494936611666
2024-09-26 00:24:50,876 INFO [train.py:1198] (1/4) Epoch 48, batch 2150, loss[loss=0.2002, ctc_loss=0.1311, cr_loss=0.3453, over 16499.00 frames. ], tot_loss[loss=0.187, ctc_loss=0.1195, cr_loss=0.3377, over 3353171.39 frames. ], batch size: 66, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:24:56,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=864560.6666666666, ans=0.1
2024-09-26 00:24:57,554 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.const_attention_rate, batch_count=864560.6666666666, ans=0.025
2024-09-26 00:25:07,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer1.prob, batch_count=864607.3333333334, ans=0.125
2024-09-26 00:25:37,464 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.295e+02 1.362e+02 1.467e+02 1.863e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-26 00:25:48,367 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=864700.6666666666, ans=0.125
2024-09-26 00:25:48,811 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=13.31 vs. limit=22.5
2024-09-26 00:26:13,770 INFO [train.py:1198] (1/4) Epoch 48, batch 2200, loss[loss=0.1652, ctc_loss=0.103, cr_loss=0.311, over 16949.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1192, cr_loss=0.3374, over 3353423.13 frames. ], batch size: 42, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:26:19,638 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=6.82 vs. limit=15.0
2024-09-26 00:26:48,591 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.74 vs. limit=22.5
2024-09-26 00:27:14,932 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=864934.0, ans=0.125
2024-09-26 00:27:30,786 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=864980.6666666666, ans=0.125
2024-09-26 00:27:36,711 INFO [train.py:1198] (1/4) Epoch 48, batch 2250, loss[loss=0.2211, ctc_loss=0.1422, cr_loss=0.3943, over 17207.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.119, cr_loss=0.3365, over 3363140.61 frames. ], batch size: 55, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:27:52,964 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=865074.0, ans=0.2
2024-09-26 00:27:55,388 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.51 vs. limit=15.0
2024-09-26 00:28:01,663 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=7.40 vs. limit=15.0
2024-09-26 00:28:14,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=865120.6666666666, ans=0.0
2024-09-26 00:28:23,488 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.181e+02 1.300e+02 1.376e+02 1.452e+02 2.028e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-26 00:28:26,982 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=865167.3333333334, ans=0.125
2024-09-26 00:28:39,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=865167.3333333334, ans=0.1
2024-09-26 00:28:56,339 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.05 vs. limit=22.5
2024-09-26 00:29:00,162 INFO [train.py:1198] (1/4) Epoch 48, batch 2300, loss[loss=0.229, ctc_loss=0.1528, cr_loss=0.3811, over 16961.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1193, cr_loss=0.3369, over 3367677.85 frames. ], batch size: 58, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:29:40,038 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=865354.0, ans=0.125
2024-09-26 00:29:45,239 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.11 vs. limit=22.5
2024-09-26 00:30:13,708 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=865447.3333333334, ans=0.025
2024-09-26 00:30:14,213 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.34 vs. limit=15.0
2024-09-26 00:30:20,233 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 00:30:22,865 INFO [train.py:1198] (1/4) Epoch 48, batch 2350, loss[loss=0.151, ctc_loss=0.09638, cr_loss=0.2731, over 17304.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1193, cr_loss=0.3362, over 3359239.06 frames. ], batch size: 46, lr: 2.47e-03, grad_scale: 16.0
2024-09-26 00:30:56,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=865587.3333333334, ans=0.125
2024-09-26 00:31:10,842 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.52 vs. limit=10.0
2024-09-26 00:31:11,656 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.301e+02 1.386e+02 1.481e+02 2.528e+02, threshold=2.772e+02, percent-clipped=0.0
2024-09-26 00:31:15,089 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=865634.0, ans=0.125
2024-09-26 00:31:26,726 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.79 vs. limit=12.0
2024-09-26 00:31:29,466 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=865680.6666666666, ans=0.025
2024-09-26 00:31:44,986 INFO [train.py:1198] (1/4) Epoch 48, batch 2400, loss[loss=0.1765, ctc_loss=0.1113, cr_loss=0.3261, over 17355.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.119, cr_loss=0.3357, over 3364488.12 frames. ], batch size: 48, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:31:56,738 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=865727.3333333334, ans=0.2
2024-09-26 00:32:10,297 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.02 vs. limit=22.5
2024-09-26 00:32:38,005 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.11 vs. limit=15.0
2024-09-26 00:33:07,653 INFO [train.py:1198] (1/4) Epoch 48, batch 2450, loss[loss=0.2327, ctc_loss=0.1536, cr_loss=0.3954, over 15046.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.118, cr_loss=0.3333, over 3371954.99 frames. ], batch size: 89, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:33:11,190 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=865960.6666666666, ans=0.0
2024-09-26 00:33:45,149 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=5.25 vs. limit=15.0
2024-09-26 00:33:58,228 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.328e+02 1.396e+02 1.507e+02 2.681e+02, threshold=2.792e+02, percent-clipped=0.0
2024-09-26 00:34:32,129 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.81 vs. limit=15.0
2024-09-26 00:34:32,988 INFO [train.py:1198] (1/4) Epoch 48, batch 2500, loss[loss=0.2319, ctc_loss=0.1554, cr_loss=0.3829, over 11343.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1179, cr_loss=0.3333, over 3365094.55 frames. ], batch size: 123, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:34:42,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=866194.0, ans=0.0
2024-09-26 00:34:42,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=866194.0, ans=0.025
2024-09-26 00:34:46,677 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.22 vs. limit=12.0
2024-09-26 00:35:36,572 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=13.20 vs. limit=15.0
2024-09-26 00:35:51,517 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=866380.6666666666, ans=0.0
2024-09-26 00:35:56,124 INFO [train.py:1198] (1/4) Epoch 48, batch 2550, loss[loss=0.1971, ctc_loss=0.125, cr_loss=0.3607, over 16712.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1185, cr_loss=0.3345, over 3353071.50 frames. ], batch size: 61, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:36:43,905 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.316e+02 1.412e+02 1.521e+02 2.388e+02, threshold=2.824e+02, percent-clipped=0.0
2024-09-26 00:36:52,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=866567.3333333334, ans=0.125
2024-09-26 00:36:54,148 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=866567.3333333334, ans=0.0
2024-09-26 00:36:54,259 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=866567.3333333334, ans=0.125
2024-09-26 00:37:15,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=866660.6666666666, ans=0.125
2024-09-26 00:37:16,412 INFO [train.py:1198] (1/4) Epoch 48, batch 2600, loss[loss=0.1938, ctc_loss=0.1231, cr_loss=0.3532, over 17155.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.119, cr_loss=0.3355, over 3345499.92 frames. ], batch size: 45, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:37:18,370 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff3_skip_rate, batch_count=866660.6666666666, ans=0.0
2024-09-26 00:37:38,502 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=866707.3333333334, ans=0.2
2024-09-26 00:37:38,935 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.94 vs. limit=6.0
2024-09-26 00:38:41,785 INFO [train.py:1198] (1/4) Epoch 48, batch 2650, loss[loss=0.2109, ctc_loss=0.137, cr_loss=0.3698, over 17237.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.1189, cr_loss=0.3359, over 3347571.23 frames. ], batch size: 55, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:38:45,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=866894.0, ans=0.0
2024-09-26 00:38:46,715 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=866894.0, ans=0.125
2024-09-26 00:38:56,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer2.prob, batch_count=866940.6666666666, ans=0.125
2024-09-26 00:38:56,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=866940.6666666666, ans=0.125
2024-09-26 00:39:07,619 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=866940.6666666666, ans=0.125
2024-09-26 00:39:12,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer2.prob, batch_count=866987.3333333334, ans=0.125
2024-09-26 00:39:29,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=866987.3333333334, ans=0.125
2024-09-26 00:39:32,299 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.306e+02 1.381e+02 1.471e+02 2.187e+02, threshold=2.762e+02, percent-clipped=0.0
2024-09-26 00:39:32,622 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=867034.0, ans=0.0
2024-09-26 00:39:56,856 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=867080.6666666666, ans=0.1
2024-09-26 00:40:04,502 INFO [train.py:1198] (1/4) Epoch 48, batch 2700, loss[loss=0.1604, ctc_loss=0.1019, cr_loss=0.2922, over 16964.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1179, cr_loss=0.3336, over 3351023.73 frames. ], batch size: 42, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:40:19,888 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.93 vs. limit=15.0
2024-09-26 00:40:42,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=867220.6666666666, ans=0.0
2024-09-26 00:41:17,221 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.81 vs. limit=15.0
2024-09-26 00:41:27,333 INFO [train.py:1198] (1/4) Epoch 48, batch 2750, loss[loss=0.1912, ctc_loss=0.123, cr_loss=0.3413, over 17033.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.1181, cr_loss=0.3343, over 3352382.71 frames. ], batch size: 52, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:42:05,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=867454.0, ans=0.125
2024-09-26 00:42:12,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=867454.0, ans=0.0
2024-09-26 00:42:15,064 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.302e+02 1.400e+02 1.492e+02 2.005e+02, threshold=2.800e+02, percent-clipped=0.0
2024-09-26 00:42:34,749 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.11 vs. limit=15.0
2024-09-26 00:42:49,904 INFO [train.py:1198] (1/4) Epoch 48, batch 2800, loss[loss=0.1588, ctc_loss=0.09822, cr_loss=0.3027, over 17111.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.1181, cr_loss=0.3346, over 3355451.55 frames. ], batch size: 40, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:43:15,748 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=867640.6666666666, ans=0.125
2024-09-26 00:43:22,580 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.21 vs. limit=12.0
2024-09-26 00:44:04,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=867780.6666666666, ans=0.1
2024-09-26 00:44:10,192 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=6.11 vs. limit=15.0
2024-09-26 00:44:12,428 INFO [train.py:1198] (1/4) Epoch 48, batch 2850, loss[loss=0.1626, ctc_loss=0.1014, cr_loss=0.3063, over 17162.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1178, cr_loss=0.3338, over 3355259.97 frames. ], batch size: 41, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:44:17,505 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=867827.3333333334, ans=0.0
2024-09-26 00:44:28,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=867827.3333333334, ans=0.025
2024-09-26 00:44:45,623 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=867920.6666666666, ans=0.2
2024-09-26 00:45:02,969 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.286e+02 1.342e+02 1.424e+02 1.734e+02, threshold=2.685e+02, percent-clipped=0.0
2024-09-26 00:45:08,100 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=867967.3333333334, ans=0.0
2024-09-26 00:45:11,669 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.91 vs. limit=6.0
2024-09-26 00:45:29,222 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.91 vs. limit=15.0
2024-09-26 00:45:37,911 INFO [train.py:1198] (1/4) Epoch 48, batch 2900, loss[loss=0.204, ctc_loss=0.1302, cr_loss=0.3688, over 17302.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1175, cr_loss=0.3335, over 3360010.22 frames. ], batch size: 46, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:45:46,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=868060.6666666666, ans=0.0
2024-09-26 00:45:47,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=868060.6666666666, ans=10.0
2024-09-26 00:46:00,857 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=2.93 vs. limit=15.0
2024-09-26 00:46:11,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=868154.0, ans=0.0
2024-09-26 00:46:12,326 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=9.53 vs. limit=15.0
2024-09-26 00:46:12,682 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.18 vs. limit=15.0
2024-09-26 00:46:19,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=868154.0, ans=0.125
2024-09-26 00:46:48,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=868247.3333333334, ans=10.0
2024-09-26 00:46:53,440 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=868247.3333333334, ans=0.125
2024-09-26 00:46:57,808 INFO [train.py:1198] (1/4) Epoch 48, batch 2950, loss[loss=0.2432, ctc_loss=0.1598, cr_loss=0.417, over 15992.00 frames. ], tot_loss[loss=0.1844, ctc_loss=0.1177, cr_loss=0.3334, over 3359067.85 frames. ], batch size: 74, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:47:34,504 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=868387.3333333334, ans=0.2
2024-09-26 00:47:50,008 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.284e+02 1.365e+02 1.462e+02 2.440e+02, threshold=2.731e+02, percent-clipped=0.0
2024-09-26 00:48:18,123 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=192, metric=6.01 vs. limit=15.0
2024-09-26 00:48:20,378 INFO [train.py:1198] (1/4) Epoch 48, batch 3000, loss[loss=0.1782, ctc_loss=0.1138, cr_loss=0.3219, over 17074.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1186, cr_loss=0.3347, over 3356763.74 frames. ], batch size: 46, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:48:20,378 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-26 00:48:38,783 INFO [train.py:1230] (1/4) Epoch 48, validation: loss=0.03527, ctc_loss=0.03527, cr_loss=1.067e-14, over 944034.00 frames. 
2024-09-26 00:48:38,784 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-26 00:48:59,859 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=384, metric=4.73 vs. limit=15.0
2024-09-26 00:49:15,473 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.33 vs. limit=10.0
2024-09-26 00:49:25,330 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=10.68 vs. limit=10.0
2024-09-26 00:49:25,990 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=868667.3333333334, ans=0.0
2024-09-26 00:49:36,913 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=868667.3333333334, ans=0.125
2024-09-26 00:49:57,146 INFO [train.py:1198] (1/4) Epoch 48, batch 3050, loss[loss=0.199, ctc_loss=0.1265, cr_loss=0.3625, over 16932.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1187, cr_loss=0.3347, over 3363070.42 frames. ], batch size: 58, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:50:13,019 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=868807.3333333334, ans=0.1
2024-09-26 00:50:31,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=868854.0, ans=0.2
2024-09-26 00:50:43,849 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=868854.0, ans=0.0
2024-09-26 00:50:47,123 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=868900.6666666666, ans=0.0
2024-09-26 00:50:48,353 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.292e+02 1.394e+02 1.474e+02 1.856e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-26 00:51:07,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=868947.3333333334, ans=0.0
2024-09-26 00:51:17,165 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=868994.0, ans=0.05
2024-09-26 00:51:18,421 INFO [train.py:1198] (1/4) Epoch 48, batch 3100, loss[loss=0.1981, ctc_loss=0.1307, cr_loss=0.3371, over 17118.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1183, cr_loss=0.3344, over 3360515.48 frames. ], batch size: 48, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:51:41,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=869040.6666666666, ans=0.125
2024-09-26 00:51:43,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=869040.6666666666, ans=0.125
2024-09-26 00:51:52,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=869087.3333333334, ans=0.0
2024-09-26 00:51:55,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer1.prob, batch_count=869087.3333333334, ans=0.125
2024-09-26 00:52:12,776 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=869134.0, ans=0.1
2024-09-26 00:52:25,318 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=869180.6666666666, ans=0.125
2024-09-26 00:52:30,095 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=2.539e-03
2024-09-26 00:52:35,870 INFO [train.py:1198] (1/4) Epoch 48, batch 3150, loss[loss=0.1932, ctc_loss=0.1253, cr_loss=0.3398, over 16989.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1187, cr_loss=0.3349, over 3360148.01 frames. ], batch size: 56, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 00:52:52,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.attention_skip_rate, batch_count=869274.0, ans=0.0
2024-09-26 00:53:18,762 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 00:53:26,182 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.286e+02 1.377e+02 1.489e+02 2.573e+02, threshold=2.753e+02, percent-clipped=0.0
2024-09-26 00:53:48,372 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=869414.0, ans=0.125
2024-09-26 00:53:48,816 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten.whitening_limit, batch_count=869414.0, ans=22.5
2024-09-26 00:53:55,916 INFO [train.py:1198] (1/4) Epoch 48, batch 3200, loss[loss=0.1976, ctc_loss=0.1321, cr_loss=0.3278, over 16014.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1179, cr_loss=0.3331, over 3367941.74 frames. ], batch size: 74, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:53:56,130 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=869460.6666666666, ans=0.125
2024-09-26 00:54:02,472 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=869460.6666666666, ans=0.1
2024-09-26 00:54:19,639 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=869507.3333333334, ans=0.125
2024-09-26 00:54:35,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=869554.0, ans=0.2
2024-09-26 00:54:43,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=869600.6666666666, ans=0.125
2024-09-26 00:54:51,398 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=869600.6666666666, ans=0.125
2024-09-26 00:55:14,463 INFO [train.py:1198] (1/4) Epoch 48, batch 3250, loss[loss=0.1891, ctc_loss=0.1178, cr_loss=0.3563, over 17069.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1176, cr_loss=0.3329, over 3359885.72 frames. ], batch size: 46, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:55:18,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.64 vs. limit=8.0
2024-09-26 00:55:33,983 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=11.92 vs. limit=15.0
2024-09-26 00:55:57,164 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=869787.3333333334, ans=0.2
2024-09-26 00:56:03,196 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.301e+02 1.439e+02 1.548e+02 2.033e+02, threshold=2.879e+02, percent-clipped=0.0
2024-09-26 00:56:33,153 INFO [train.py:1198] (1/4) Epoch 48, batch 3300, loss[loss=0.1946, ctc_loss=0.1247, cr_loss=0.3493, over 17203.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1179, cr_loss=0.3334, over 3352917.58 frames. ], batch size: 47, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:57:34,107 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=870067.3333333334, ans=10.0
2024-09-26 00:57:54,015 INFO [train.py:1198] (1/4) Epoch 48, batch 3350, loss[loss=0.1974, ctc_loss=0.1268, cr_loss=0.3529, over 17150.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1179, cr_loss=0.3334, over 3353609.58 frames. ], batch size: 48, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:58:06,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass_mid.scale_min, batch_count=870160.6666666666, ans=0.2
2024-09-26 00:58:08,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=870207.3333333334, ans=0.2
2024-09-26 00:58:42,523 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.313e+02 1.389e+02 1.485e+02 2.642e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-26 00:59:09,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=870347.3333333334, ans=0.125
2024-09-26 00:59:12,293 INFO [train.py:1198] (1/4) Epoch 48, batch 3400, loss[loss=0.1532, ctc_loss=0.0929, cr_loss=0.3017, over 17184.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1179, cr_loss=0.3337, over 3357130.09 frames. ], batch size: 41, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 00:59:28,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=870440.6666666666, ans=0.0
2024-09-26 00:59:55,450 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=870487.3333333334, ans=0.2
2024-09-26 00:59:58,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=870487.3333333334, ans=0.025
2024-09-26 00:59:59,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=870534.0, ans=0.125
2024-09-26 01:00:03,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=870534.0, ans=0.0
2024-09-26 01:00:09,474 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=870534.0, ans=0.5
2024-09-26 01:00:32,484 INFO [train.py:1198] (1/4) Epoch 48, batch 3450, loss[loss=0.1618, ctc_loss=0.1034, cr_loss=0.292, over 17164.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1176, cr_loss=0.333, over 3352529.72 frames. ], batch size: 45, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 01:01:19,247 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.79 vs. limit=15.0
2024-09-26 01:01:22,138 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.51 vs. limit=15.0
2024-09-26 01:01:24,344 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.301e+02 1.395e+02 1.467e+02 2.192e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-26 01:01:30,967 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=870767.3333333334, ans=0.1
2024-09-26 01:01:34,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=870767.3333333334, ans=0.125
2024-09-26 01:01:52,535 INFO [train.py:1198] (1/4) Epoch 48, batch 3500, loss[loss=0.2289, ctc_loss=0.1485, cr_loss=0.4021, over 14949.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.119, cr_loss=0.3359, over 3339633.06 frames. ], batch size: 89, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 01:02:22,943 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=870954.0, ans=0.125
2024-09-26 01:02:27,760 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.ff2_skip_rate, batch_count=870954.0, ans=0.0
2024-09-26 01:02:39,147 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=870954.0, ans=0.125
2024-09-26 01:02:42,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=871000.6666666666, ans=0.125
2024-09-26 01:03:02,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=871047.3333333334, ans=0.1
2024-09-26 01:03:03,830 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=871047.3333333334, ans=0.0
2024-09-26 01:03:12,762 INFO [train.py:1198] (1/4) Epoch 48, batch 3550, loss[loss=0.1691, ctc_loss=0.1038, cr_loss=0.3265, over 16956.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1186, cr_loss=0.335, over 3339793.53 frames. ], batch size: 42, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 01:03:14,655 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=871094.0, ans=0.125
2024-09-26 01:03:25,759 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=871094.0, ans=0.1
2024-09-26 01:03:28,822 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=871140.6666666666, ans=0.0
2024-09-26 01:03:33,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=871140.6666666666, ans=0.1
2024-09-26 01:03:50,371 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=871187.3333333334, ans=0.0
2024-09-26 01:04:02,352 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.323e+02 1.392e+02 1.491e+02 3.535e+02, threshold=2.785e+02, percent-clipped=1.0
2024-09-26 01:04:21,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=871280.6666666666, ans=0.1
2024-09-26 01:04:22,851 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=871280.6666666666, ans=0.2
2024-09-26 01:04:29,331 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.48 vs. limit=15.0
2024-09-26 01:04:30,423 INFO [train.py:1198] (1/4) Epoch 48, batch 3600, loss[loss=0.1486, ctc_loss=0.09222, cr_loss=0.2816, over 16325.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1188, cr_loss=0.3351, over 3343891.17 frames. ], batch size: 36, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 01:04:46,353 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=871374.0, ans=0.0
2024-09-26 01:04:57,239 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=871374.0, ans=0.1
2024-09-26 01:05:00,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=871420.6666666666, ans=0.125
2024-09-26 01:05:02,094 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=871420.6666666666, ans=0.125
2024-09-26 01:05:16,034 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.min_positive, batch_count=871467.3333333334, ans=0.05
2024-09-26 01:05:17,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=871467.3333333334, ans=0.0
2024-09-26 01:05:17,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=871467.3333333334, ans=0.1
2024-09-26 01:05:19,087 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=871467.3333333334, ans=0.125
2024-09-26 01:05:22,092 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=871467.3333333334, ans=0.0
2024-09-26 01:05:32,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=871514.0, ans=0.0
2024-09-26 01:05:48,461 INFO [train.py:1198] (1/4) Epoch 48, batch 3650, loss[loss=0.2013, ctc_loss=0.1305, cr_loss=0.354, over 17219.00 frames. ], tot_loss[loss=0.1853, ctc_loss=0.1184, cr_loss=0.3344, over 3351933.14 frames. ], batch size: 55, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 01:05:54,942 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.skip_rate, batch_count=871560.6666666666, ans=0.07
2024-09-26 01:06:08,390 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=512, metric=15.29 vs. limit=22.5
2024-09-26 01:06:09,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=871607.3333333334, ans=10.0
2024-09-26 01:06:40,145 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.328e+02 1.397e+02 1.530e+02 1.851e+02, threshold=2.795e+02, percent-clipped=0.0
2024-09-26 01:06:49,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=871700.6666666666, ans=0.125
2024-09-26 01:07:02,292 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.22 vs. limit=15.0
2024-09-26 01:07:09,263 INFO [train.py:1198] (1/4) Epoch 48, batch 3700, loss[loss=0.174, ctc_loss=0.1102, cr_loss=0.319, over 17285.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.118, cr_loss=0.3337, over 3346517.91 frames. ], batch size: 46, lr: 2.46e-03, grad_scale: 32.0
2024-09-26 01:07:12,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer1.prob, batch_count=871794.0, ans=0.125
2024-09-26 01:07:14,354 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.balancer.min_positive, batch_count=871794.0, ans=0.05
2024-09-26 01:07:17,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=871794.0, ans=0.1
2024-09-26 01:07:25,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=871840.6666666666, ans=0.1
2024-09-26 01:07:54,502 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=192, metric=13.37 vs. limit=15.0
2024-09-26 01:07:58,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=871934.0, ans=0.125
2024-09-26 01:07:58,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=871934.0, ans=0.2
2024-09-26 01:08:16,574 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.87 vs. limit=12.0
2024-09-26 01:08:21,332 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.70 vs. limit=15.0
2024-09-26 01:08:28,604 INFO [train.py:1198] (1/4) Epoch 48, batch 3750, loss[loss=0.2009, ctc_loss=0.1338, cr_loss=0.3358, over 17160.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1185, cr_loss=0.3343, over 3334174.00 frames. ], batch size: 48, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 01:08:28,860 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=872027.3333333334, ans=0.125
2024-09-26 01:08:49,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=872074.0, ans=0.125
2024-09-26 01:08:55,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=872074.0, ans=0.0
2024-09-26 01:09:10,604 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=13.54 vs. limit=15.0
2024-09-26 01:09:15,024 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=5.24 vs. limit=10.0
2024-09-26 01:09:20,719 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.319e+02 1.400e+02 1.543e+02 5.735e+02, threshold=2.801e+02, percent-clipped=1.0
2024-09-26 01:09:29,633 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=872167.3333333334, ans=0.025
2024-09-26 01:09:45,044 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=3.88 vs. limit=12.0
2024-09-26 01:09:48,870 INFO [train.py:1198] (1/4) Epoch 48, batch 3800, loss[loss=0.2124, ctc_loss=0.1401, cr_loss=0.3614, over 15086.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1187, cr_loss=0.3342, over 3316058.71 frames. ], batch size: 89, lr: 2.46e-03, grad_scale: 16.0
2024-09-26 01:09:51,608 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.39 vs. limit=15.0
2024-09-26 01:09:53,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=872260.6666666666, ans=0.125
2024-09-26 01:09:58,634 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=872260.6666666666, ans=0.0
2024-09-26 01:10:57,832 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 01:11:05,545 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 01:11:06,798 INFO [train.py:1198] (1/4) Epoch 48, batch 3850, loss[loss=0.1995, ctc_loss=0.13, cr_loss=0.3475, over 16678.00 frames. ], tot_loss[loss=0.1887, ctc_loss=0.1212, cr_loss=0.3375, over 3275785.55 frames. ], batch size: 66, lr: 2.46e-03, grad_scale: 8.0
2024-09-26 01:11:08,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=872494.0, ans=0.125
2024-09-26 01:11:12,446 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=16.97 vs. limit=22.5
2024-09-26 01:11:59,142 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.219e+02 1.389e+02 1.529e+02 1.705e+02 2.274e+02, threshold=3.058e+02, percent-clipped=0.0
2024-09-26 01:12:13,896 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=9.36 vs. limit=10.0
2024-09-26 01:13:04,120 INFO [train.py:1198] (1/4) Epoch 49, batch 0, loss[loss=0.2067, ctc_loss=0.1299, cr_loss=0.3838, over 17011.00 frames. ], tot_loss[loss=0.2067, ctc_loss=0.1299, cr_loss=0.3838, over 17011.00 frames. ], batch size: 51, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:13:04,121 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-26 01:13:19,513 INFO [train.py:1230] (1/4) Epoch 49, validation: loss=0.03487, ctc_loss=0.03487, cr_loss=1.087e-14, over 944034.00 frames. 
2024-09-26 01:13:19,513 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-26 01:13:28,872 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 01:13:31,866 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=872708.6666666666, ans=0.025
2024-09-26 01:13:34,894 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=872708.6666666666, ans=0.2
2024-09-26 01:13:38,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=872755.3333333334, ans=0.125
2024-09-26 01:14:44,163 INFO [train.py:1198] (1/4) Epoch 49, batch 50, loss[loss=0.1945, ctc_loss=0.1229, cr_loss=0.3575, over 17217.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.1186, cr_loss=0.3374, over 757693.81 frames. ], batch size: 47, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:14:55,098 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=872942.0, ans=0.2
2024-09-26 01:14:56,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=872942.0, ans=0.125
2024-09-26 01:15:20,843 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.48 vs. limit=22.5
2024-09-26 01:15:47,637 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.321e+02 1.394e+02 1.552e+02 2.223e+02, threshold=2.788e+02, percent-clipped=0.0
2024-09-26 01:15:52,646 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=873128.6666666666, ans=0.125
2024-09-26 01:15:59,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=873128.6666666666, ans=0.125
2024-09-26 01:16:07,026 INFO [train.py:1198] (1/4) Epoch 49, batch 100, loss[loss=0.1977, ctc_loss=0.1274, cr_loss=0.3513, over 17025.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1179, cr_loss=0.3361, over 1334488.99 frames. ], batch size: 52, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:16:09,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=873175.3333333334, ans=0.1
2024-09-26 01:16:10,979 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.66 vs. limit=15.0
2024-09-26 01:16:20,844 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.70 vs. limit=15.0
2024-09-26 01:16:28,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=873222.0, ans=0.025
2024-09-26 01:17:11,348 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=873362.0, ans=0.1
2024-09-26 01:17:29,483 INFO [train.py:1198] (1/4) Epoch 49, batch 150, loss[loss=0.1855, ctc_loss=0.118, cr_loss=0.3378, over 17287.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1184, cr_loss=0.3367, over 1792638.86 frames. ], batch size: 49, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:18:08,409 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.66 vs. limit=15.0
2024-09-26 01:18:31,433 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=873548.6666666666, ans=0.125
2024-09-26 01:18:32,853 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.298e+02 1.390e+02 1.504e+02 2.457e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-26 01:18:52,325 INFO [train.py:1198] (1/4) Epoch 49, batch 200, loss[loss=0.1948, ctc_loss=0.1237, cr_loss=0.3553, over 17200.00 frames. ], tot_loss[loss=0.184, ctc_loss=0.1174, cr_loss=0.333, over 2146555.51 frames. ], batch size: 55, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:18:57,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=873642.0, ans=0.125
2024-09-26 01:19:08,815 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.76 vs. limit=22.5
2024-09-26 01:19:16,810 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=873688.6666666666, ans=0.125
2024-09-26 01:19:40,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer2.prob, batch_count=873782.0, ans=0.125
2024-09-26 01:20:05,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=873828.6666666666, ans=0.025
2024-09-26 01:20:17,864 INFO [train.py:1198] (1/4) Epoch 49, batch 250, loss[loss=0.1875, ctc_loss=0.1185, cr_loss=0.3449, over 17247.00 frames. ], tot_loss[loss=0.1844, ctc_loss=0.1176, cr_loss=0.3339, over 2417883.66 frames. ], batch size: 44, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:20:45,619 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.94 vs. limit=6.0
2024-09-26 01:21:18,234 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.257e+02 1.339e+02 1.417e+02 1.603e+02, threshold=2.678e+02, percent-clipped=0.0
2024-09-26 01:21:33,747 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=512, metric=7.20 vs. limit=15.0
2024-09-26 01:21:37,797 INFO [train.py:1198] (1/4) Epoch 49, batch 300, loss[loss=0.1761, ctc_loss=0.111, cr_loss=0.3257, over 17267.00 frames. ], tot_loss[loss=0.1839, ctc_loss=0.1173, cr_loss=0.3332, over 2628243.77 frames. ], batch size: 44, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:22:53,746 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=17.89 vs. limit=22.5
2024-09-26 01:22:55,451 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.whiten_keys.whitening_limit, batch_count=874295.3333333334, ans=6.0
2024-09-26 01:22:56,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=874295.3333333334, ans=0.0
2024-09-26 01:23:00,851 INFO [train.py:1198] (1/4) Epoch 49, batch 350, loss[loss=0.1997, ctc_loss=0.1288, cr_loss=0.3543, over 17298.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1174, cr_loss=0.3338, over 2794204.66 frames. ], batch size: 51, lr: 2.43e-03, grad_scale: 16.0
2024-09-26 01:23:45,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=874435.3333333334, ans=0.1
2024-09-26 01:24:04,152 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.315e+02 1.403e+02 1.517e+02 7.992e+02, threshold=2.807e+02, percent-clipped=1.0
2024-09-26 01:24:04,771 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.30 vs. limit=15.0
2024-09-26 01:24:23,274 INFO [train.py:1198] (1/4) Epoch 49, batch 400, loss[loss=0.2017, ctc_loss=0.1249, cr_loss=0.3841, over 17097.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1181, cr_loss=0.3348, over 2904061.77 frames. ], batch size: 49, lr: 2.43e-03, grad_scale: 32.0
2024-09-26 01:24:25,236 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=874575.3333333334, ans=0.0
2024-09-26 01:24:31,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=874575.3333333334, ans=10.0
2024-09-26 01:25:31,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=874762.0, ans=0.125
2024-09-26 01:25:34,313 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=874762.0, ans=0.0
2024-09-26 01:25:48,234 INFO [train.py:1198] (1/4) Epoch 49, batch 450, loss[loss=0.148, ctc_loss=0.09125, cr_loss=0.2836, over 17107.00 frames. ], tot_loss[loss=0.1844, ctc_loss=0.1176, cr_loss=0.3341, over 3012112.51 frames. ], batch size: 40, lr: 2.43e-03, grad_scale: 32.0
2024-09-26 01:25:53,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=874808.6666666666, ans=0.1
2024-09-26 01:26:14,069 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=874855.3333333334, ans=0.2
2024-09-26 01:26:19,070 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff3_skip_rate, batch_count=874902.0, ans=0.0
2024-09-26 01:26:25,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=874902.0, ans=0.04949747468305833
2024-09-26 01:26:38,914 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.23 vs. limit=12.0
2024-09-26 01:26:40,551 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=15.72 vs. limit=22.5
2024-09-26 01:26:49,065 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.319e+02 1.417e+02 1.516e+02 2.635e+02, threshold=2.833e+02, percent-clipped=0.0
2024-09-26 01:27:08,162 INFO [train.py:1198] (1/4) Epoch 49, batch 500, loss[loss=0.1842, ctc_loss=0.1155, cr_loss=0.3435, over 17152.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1174, cr_loss=0.3334, over 3092130.62 frames. ], batch size: 45, lr: 2.43e-03, grad_scale: 32.0
2024-09-26 01:27:09,957 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.min_positive, batch_count=875042.0, ans=0.05
2024-09-26 01:27:41,406 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=875135.3333333334, ans=0.125
2024-09-26 01:27:46,785 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.61 vs. limit=15.0
2024-09-26 01:27:47,730 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=875135.3333333334, ans=0.0
2024-09-26 01:27:51,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=875135.3333333334, ans=0.125
2024-09-26 01:28:32,675 INFO [train.py:1198] (1/4) Epoch 49, batch 550, loss[loss=0.1806, ctc_loss=0.1161, cr_loss=0.3225, over 17232.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1175, cr_loss=0.3328, over 3157194.23 frames. ], batch size: 50, lr: 2.43e-03, grad_scale: 32.0
2024-09-26 01:28:50,431 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=875322.0, ans=0.0
2024-09-26 01:28:58,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=875322.0, ans=0.125
2024-09-26 01:29:03,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=875368.6666666666, ans=0.125
2024-09-26 01:29:16,315 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 01:29:19,325 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=875415.3333333334, ans=0.125
2024-09-26 01:29:20,993 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.prob, batch_count=875415.3333333334, ans=0.125
2024-09-26 01:29:33,607 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.295e+02 1.361e+02 1.476e+02 2.484e+02, threshold=2.722e+02, percent-clipped=0.0
2024-09-26 01:29:51,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer2.prob, batch_count=875462.0, ans=0.125
2024-09-26 01:29:58,449 INFO [train.py:1198] (1/4) Epoch 49, batch 600, loss[loss=0.1938, ctc_loss=0.1269, cr_loss=0.3343, over 17232.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3354, over 3203740.61 frames. ], batch size: 50, lr: 2.43e-03, grad_scale: 32.0
2024-09-26 01:30:35,228 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=875602.0, ans=0.0
2024-09-26 01:30:48,402 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.94 vs. limit=15.0
2024-09-26 01:30:48,823 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=4.95 vs. limit=15.0
2024-09-26 01:31:01,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=875695.3333333334, ans=0.1
2024-09-26 01:31:18,110 INFO [train.py:1198] (1/4) Epoch 49, batch 650, loss[loss=0.1979, ctc_loss=0.131, cr_loss=0.3345, over 16026.00 frames. ], tot_loss[loss=0.1838, ctc_loss=0.1172, cr_loss=0.3331, over 3243910.28 frames. ], batch size: 74, lr: 2.43e-03, grad_scale: 32.0
2024-09-26 01:31:21,674 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.min_abs, batch_count=875742.0, ans=0.5
2024-09-26 01:31:38,375 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=5.00 vs. limit=6.0
2024-09-26 01:31:49,071 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.skip_rate, batch_count=875835.3333333334, ans=0.07
2024-09-26 01:32:09,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=875882.0, ans=0.125
2024-09-26 01:32:19,316 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=8.16 vs. limit=15.0
2024-09-26 01:32:21,520 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.285e+02 1.382e+02 1.503e+02 2.355e+02, threshold=2.765e+02, percent-clipped=0.0
2024-09-26 01:32:40,441 INFO [train.py:1198] (1/4) Epoch 49, batch 700, loss[loss=0.1984, ctc_loss=0.1256, cr_loss=0.3643, over 17015.00 frames. ], tot_loss[loss=0.1837, ctc_loss=0.1171, cr_loss=0.3328, over 3271552.24 frames. ], batch size: 56, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:32:48,789 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=875975.3333333334, ans=0.125
2024-09-26 01:33:11,502 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.13 vs. limit=15.0
2024-09-26 01:33:42,011 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=12.06 vs. limit=22.5
2024-09-26 01:33:46,168 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=876162.0, ans=0.2
2024-09-26 01:33:57,288 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=876162.0, ans=0.1
2024-09-26 01:34:03,487 INFO [train.py:1198] (1/4) Epoch 49, batch 750, loss[loss=0.191, ctc_loss=0.121, cr_loss=0.3502, over 17070.00 frames. ], tot_loss[loss=0.1838, ctc_loss=0.1172, cr_loss=0.333, over 3294413.56 frames. ], batch size: 46, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:34:12,223 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=9.90 vs. limit=12.0
2024-09-26 01:34:30,954 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=876255.3333333334, ans=0.125
2024-09-26 01:34:37,523 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.prob, batch_count=876302.0, ans=0.125
2024-09-26 01:34:37,625 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=876302.0, ans=0.0
2024-09-26 01:34:46,662 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=876302.0, ans=0.1
2024-09-26 01:34:53,048 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=876348.6666666666, ans=0.125
2024-09-26 01:35:09,766 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.321e+02 1.400e+02 1.492e+02 1.805e+02, threshold=2.801e+02, percent-clipped=0.0
2024-09-26 01:35:19,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.prob, batch_count=876395.3333333334, ans=0.125
2024-09-26 01:35:28,993 INFO [train.py:1198] (1/4) Epoch 49, batch 800, loss[loss=0.2136, ctc_loss=0.1388, cr_loss=0.3742, over 17027.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.118, cr_loss=0.3347, over 3298601.86 frames. ], batch size: 52, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:35:50,254 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=876488.6666666666, ans=0.125
2024-09-26 01:36:09,102 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=876535.3333333334, ans=0.035
2024-09-26 01:36:10,749 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=876535.3333333334, ans=0.2
2024-09-26 01:36:18,669 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=876582.0, ans=0.1
2024-09-26 01:36:49,021 INFO [train.py:1198] (1/4) Epoch 49, batch 850, loss[loss=0.1947, ctc_loss=0.1254, cr_loss=0.3464, over 16036.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3353, over 3313066.93 frames. ], batch size: 74, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:36:54,601 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=9.94 vs. limit=22.5
2024-09-26 01:37:48,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=876815.3333333334, ans=0.0
2024-09-26 01:37:51,966 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.324e+02 1.412e+02 1.511e+02 2.737e+02, threshold=2.823e+02, percent-clipped=0.0
2024-09-26 01:38:05,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=876862.0, ans=0.0
2024-09-26 01:38:09,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=876908.6666666666, ans=0.1
2024-09-26 01:38:11,392 INFO [train.py:1198] (1/4) Epoch 49, batch 900, loss[loss=0.2224, ctc_loss=0.1516, cr_loss=0.3539, over 11657.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1188, cr_loss=0.3356, over 3313867.71 frames. ], batch size: 123, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:38:14,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=876908.6666666666, ans=0.1
2024-09-26 01:38:18,062 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=876908.6666666666, ans=0.0
2024-09-26 01:38:26,833 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=876908.6666666666, ans=0.0
2024-09-26 01:38:40,036 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.54 vs. limit=22.5
2024-09-26 01:38:51,377 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=6.75 vs. limit=12.0
2024-09-26 01:38:51,482 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=5.04 vs. limit=12.0
2024-09-26 01:39:07,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=877048.6666666666, ans=0.2
2024-09-26 01:39:09,248 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.self_attn1.whiten, num_groups=1, num_channels=512, metric=12.90 vs. limit=22.5
2024-09-26 01:39:18,363 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=877095.3333333334, ans=0.125
2024-09-26 01:39:33,924 INFO [train.py:1198] (1/4) Epoch 49, batch 950, loss[loss=0.1593, ctc_loss=0.09739, cr_loss=0.3096, over 17096.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1184, cr_loss=0.3353, over 3330401.02 frames. ], batch size: 40, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:39:51,606 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=12.75 vs. limit=15.0
2024-09-26 01:40:08,280 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=877188.6666666666, ans=0.0
2024-09-26 01:40:14,493 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=877235.3333333334, ans=0.0
2024-09-26 01:40:31,410 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=12.87 vs. limit=22.5
2024-09-26 01:40:41,474 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.300e+02 1.395e+02 1.483e+02 3.362e+02, threshold=2.790e+02, percent-clipped=1.0
2024-09-26 01:41:00,336 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=877375.3333333334, ans=0.0
2024-09-26 01:41:01,590 INFO [train.py:1198] (1/4) Epoch 49, batch 1000, loss[loss=0.1938, ctc_loss=0.1243, cr_loss=0.3474, over 17319.00 frames. ], tot_loss[loss=0.1868, ctc_loss=0.1193, cr_loss=0.3373, over 3327186.44 frames. ], batch size: 51, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:41:09,773 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=877375.3333333334, ans=0.5
2024-09-26 01:41:14,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=877375.3333333334, ans=0.1
2024-09-26 01:41:30,548 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=877422.0, ans=0.0
2024-09-26 01:41:57,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=877515.3333333334, ans=0.2
2024-09-26 01:42:00,996 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=877515.3333333334, ans=0.125
2024-09-26 01:42:03,089 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.39 vs. limit=15.0
2024-09-26 01:42:21,068 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=877562.0, ans=0.0
2024-09-26 01:42:23,942 INFO [train.py:1198] (1/4) Epoch 49, batch 1050, loss[loss=0.1933, ctc_loss=0.1246, cr_loss=0.3435, over 17350.00 frames. ], tot_loss[loss=0.1865, ctc_loss=0.1192, cr_loss=0.3365, over 3331801.76 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:42:27,565 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=877608.6666666666, ans=0.1
2024-09-26 01:42:30,599 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=877608.6666666666, ans=0.125
2024-09-26 01:42:37,027 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=877608.6666666666, ans=0.2
2024-09-26 01:43:18,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=877748.6666666666, ans=0.125
2024-09-26 01:43:19,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=877748.6666666666, ans=0.125
2024-09-26 01:43:23,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=877748.6666666666, ans=0.125
2024-09-26 01:43:28,469 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.276e+02 1.380e+02 1.490e+02 2.320e+02, threshold=2.759e+02, percent-clipped=0.0
2024-09-26 01:43:30,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.min_positive, batch_count=877795.3333333334, ans=0.025
2024-09-26 01:43:41,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=877795.3333333334, ans=0.125
2024-09-26 01:43:46,129 INFO [train.py:1198] (1/4) Epoch 49, batch 1100, loss[loss=0.2085, ctc_loss=0.1339, cr_loss=0.3729, over 17209.00 frames. ], tot_loss[loss=0.1866, ctc_loss=0.1192, cr_loss=0.3366, over 3323257.61 frames. ], batch size: 55, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:44:10,233 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=877888.6666666666, ans=0.0
2024-09-26 01:44:10,269 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=877888.6666666666, ans=0.0
2024-09-26 01:44:30,876 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=877935.3333333334, ans=0.025
2024-09-26 01:44:40,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer1.prob, batch_count=877982.0, ans=0.125
2024-09-26 01:45:06,085 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.max_positive, batch_count=878028.6666666666, ans=0.95
2024-09-26 01:45:10,651 INFO [train.py:1198] (1/4) Epoch 49, batch 1150, loss[loss=0.2002, ctc_loss=0.1336, cr_loss=0.3331, over 11877.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1187, cr_loss=0.3356, over 3333684.57 frames. ], batch size: 123, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:45:14,050 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=878075.3333333334, ans=0.0
2024-09-26 01:45:28,409 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=878122.0, ans=0.125
2024-09-26 01:45:48,681 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.33 vs. limit=6.0
2024-09-26 01:45:54,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=878168.6666666666, ans=0.125
2024-09-26 01:46:01,015 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.56 vs. limit=22.5
2024-09-26 01:46:08,742 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=878215.3333333334, ans=0.0
2024-09-26 01:46:13,008 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.291e+02 1.352e+02 1.450e+02 2.079e+02, threshold=2.704e+02, percent-clipped=0.0
2024-09-26 01:46:16,666 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=878262.0, ans=0.0
2024-09-26 01:46:23,158 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=878262.0, ans=0.035
2024-09-26 01:46:30,812 INFO [train.py:1198] (1/4) Epoch 49, batch 1200, loss[loss=0.1838, ctc_loss=0.1138, cr_loss=0.3501, over 17103.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1182, cr_loss=0.3348, over 3337057.56 frames. ], batch size: 49, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:46:35,768 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=878308.6666666666, ans=0.125
2024-09-26 01:46:51,844 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn1.whiten, num_groups=1, num_channels=512, metric=10.56 vs. limit=22.5
2024-09-26 01:47:36,818 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=878495.3333333334, ans=0.125
2024-09-26 01:47:52,403 INFO [train.py:1198] (1/4) Epoch 49, batch 1250, loss[loss=0.1715, ctc_loss=0.1101, cr_loss=0.3068, over 17160.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.1189, cr_loss=0.3361, over 3344323.83 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:48:03,885 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=878542.0, ans=0.125
2024-09-26 01:48:44,596 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.46 vs. limit=10.0
2024-09-26 01:48:52,946 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module2.whiten, num_groups=1, num_channels=192, metric=3.94 vs. limit=15.0
2024-09-26 01:48:56,555 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.287e+02 1.357e+02 1.444e+02 1.832e+02, threshold=2.714e+02, percent-clipped=0.0
2024-09-26 01:49:09,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=878728.6666666666, ans=0.025
2024-09-26 01:49:14,199 INFO [train.py:1198] (1/4) Epoch 49, batch 1300, loss[loss=0.1881, ctc_loss=0.1193, cr_loss=0.344, over 17279.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1179, cr_loss=0.3342, over 3351093.61 frames. ], batch size: 51, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:49:30,386 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=878822.0, ans=0.125
2024-09-26 01:49:30,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.conv_module2.whiten, num_groups=1, num_channels=512, metric=8.01 vs. limit=15.0
2024-09-26 01:49:32,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.33 vs. limit=15.0
2024-09-26 01:49:35,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=878822.0, ans=0.125
2024-09-26 01:49:36,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=878822.0, ans=0.1
2024-09-26 01:49:53,057 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=878868.6666666666, ans=0.2
2024-09-26 01:50:01,077 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=878868.6666666666, ans=0.125
2024-09-26 01:50:18,736 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer1.prob, batch_count=878915.3333333334, ans=0.125
2024-09-26 01:50:39,184 INFO [train.py:1198] (1/4) Epoch 49, batch 1350, loss[loss=0.1869, ctc_loss=0.1173, cr_loss=0.348, over 17152.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1185, cr_loss=0.3352, over 3333655.19 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:51:03,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=879055.3333333334, ans=0.1
2024-09-26 01:51:09,705 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=879102.0, ans=0.0
2024-09-26 01:51:27,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=879148.6666666666, ans=0.1
2024-09-26 01:51:40,490 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=879148.6666666666, ans=0.1
2024-09-26 01:51:41,925 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.323e+02 1.398e+02 1.529e+02 2.878e+02, threshold=2.797e+02, percent-clipped=1.0
2024-09-26 01:51:55,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_skip_rate, batch_count=879195.3333333334, ans=0.0
2024-09-26 01:51:55,558 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.06 vs. limit=15.0
2024-09-26 01:51:59,737 INFO [train.py:1198] (1/4) Epoch 49, batch 1400, loss[loss=0.1554, ctc_loss=0.09916, cr_loss=0.281, over 17277.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1179, cr_loss=0.3346, over 3341716.66 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:52:12,047 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.const_attention_rate, batch_count=879242.0, ans=0.025
2024-09-26 01:52:37,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=879335.3333333334, ans=0.1
2024-09-26 01:52:37,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer1.prob, batch_count=879335.3333333334, ans=0.125
2024-09-26 01:52:37,393 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=879335.3333333334, ans=0.125
2024-09-26 01:52:57,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=879382.0, ans=0.125
2024-09-26 01:53:24,001 INFO [train.py:1198] (1/4) Epoch 49, batch 1450, loss[loss=0.1619, ctc_loss=0.1013, cr_loss=0.3032, over 17136.00 frames. ], tot_loss[loss=0.184, ctc_loss=0.1172, cr_loss=0.3337, over 3347821.12 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:53:34,066 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=879475.3333333334, ans=0.125
2024-09-26 01:53:34,167 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 01:53:43,605 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=879522.0, ans=0.0
2024-09-26 01:54:27,290 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=879662.0, ans=0.125
2024-09-26 01:54:28,634 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.317e+02 1.389e+02 1.499e+02 2.448e+02, threshold=2.778e+02, percent-clipped=0.0
2024-09-26 01:54:34,325 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.35 vs. limit=6.0
2024-09-26 01:54:36,746 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.out_balancer.prob, batch_count=879662.0, ans=0.125
2024-09-26 01:54:38,532 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.min_abs, batch_count=879662.0, ans=0.5
2024-09-26 01:54:47,225 INFO [train.py:1198] (1/4) Epoch 49, batch 1500, loss[loss=0.2031, ctc_loss=0.1324, cr_loss=0.3534, over 16047.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1174, cr_loss=0.3337, over 3353642.86 frames. ], batch size: 74, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:55:35,535 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=879848.6666666666, ans=0.125
2024-09-26 01:56:07,498 INFO [train.py:1198] (1/4) Epoch 49, batch 1550, loss[loss=0.173, ctc_loss=0.1112, cr_loss=0.3086, over 17221.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.118, cr_loss=0.3343, over 3360741.05 frames. ], batch size: 50, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 01:56:12,595 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=879942.0, ans=0.09899494936611666
2024-09-26 01:56:22,358 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=879988.6666666666, ans=0.125
2024-09-26 01:56:43,126 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.balancer1.prob, batch_count=880035.3333333334, ans=0.125
2024-09-26 01:56:49,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=880035.3333333334, ans=0.0
2024-09-26 01:57:13,894 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.314e+02 1.386e+02 1.460e+02 1.935e+02, threshold=2.771e+02, percent-clipped=0.0
2024-09-26 01:57:17,790 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_skip_rate, batch_count=880128.6666666666, ans=0.0
2024-09-26 01:57:29,245 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.29 vs. limit=15.0
2024-09-26 01:57:30,178 INFO [train.py:1198] (1/4) Epoch 49, batch 1600, loss[loss=0.1664, ctc_loss=0.1041, cr_loss=0.3118, over 17302.00 frames. ], tot_loss[loss=0.184, ctc_loss=0.1174, cr_loss=0.3328, over 3353931.21 frames. ], batch size: 46, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:57:40,018 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=880175.3333333334, ans=0.0
2024-09-26 01:57:55,955 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=880222.0, ans=0.125
2024-09-26 01:58:16,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=880268.6666666666, ans=0.1
2024-09-26 01:58:35,405 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=880362.0, ans=0.2
2024-09-26 01:58:36,114 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.34 vs. limit=15.0
2024-09-26 01:58:51,925 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=6.28 vs. limit=10.0
2024-09-26 01:58:52,672 INFO [train.py:1198] (1/4) Epoch 49, batch 1650, loss[loss=0.1992, ctc_loss=0.1274, cr_loss=0.3588, over 17140.00 frames. ], tot_loss[loss=0.1839, ctc_loss=0.1174, cr_loss=0.3328, over 3350376.74 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 01:59:10,571 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass_mid.scale_min, batch_count=880455.3333333334, ans=0.2
2024-09-26 01:59:14,128 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.30 vs. limit=10.0
2024-09-26 01:59:15,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.const_attention_rate, batch_count=880455.3333333334, ans=0.025
2024-09-26 01:59:20,430 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=9.27 vs. limit=15.0
2024-09-26 01:59:24,923 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=880502.0, ans=0.125
2024-09-26 01:59:31,645 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.74 vs. limit=15.0
2024-09-26 01:59:36,847 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.51 vs. limit=15.0
2024-09-26 01:59:49,305 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=880548.6666666666, ans=0.125
2024-09-26 02:00:01,887 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.295e+02 1.382e+02 1.560e+02 2.405e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-26 02:00:16,506 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=880642.0, ans=0.125
2024-09-26 02:00:17,670 INFO [train.py:1198] (1/4) Epoch 49, batch 1700, loss[loss=0.1904, ctc_loss=0.124, cr_loss=0.332, over 17143.00 frames. ], tot_loss[loss=0.1839, ctc_loss=0.1174, cr_loss=0.3326, over 3352264.73 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 02:00:25,877 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.ff2_skip_rate, batch_count=880642.0, ans=0.0
2024-09-26 02:00:53,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer_ff3.min_abs, batch_count=880735.3333333334, ans=0.2
2024-09-26 02:00:58,024 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.const_attention_rate, batch_count=880735.3333333334, ans=0.025
2024-09-26 02:01:01,208 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=880735.3333333334, ans=0.0
2024-09-26 02:01:30,031 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=880828.6666666666, ans=0.0
2024-09-26 02:01:37,517 INFO [train.py:1198] (1/4) Epoch 49, batch 1750, loss[loss=0.1718, ctc_loss=0.1084, cr_loss=0.3172, over 17211.00 frames. ], tot_loss[loss=0.1835, ctc_loss=0.117, cr_loss=0.3324, over 3351830.12 frames. ], batch size: 47, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:01:44,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=880875.3333333334, ans=0.025
2024-09-26 02:01:48,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=880875.3333333334, ans=0.1
2024-09-26 02:01:58,637 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=880922.0, ans=0.125
2024-09-26 02:02:26,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=881015.3333333334, ans=10.0
2024-09-26 02:02:31,418 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=881015.3333333334, ans=0.125
2024-09-26 02:02:39,209 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=881015.3333333334, ans=0.95
2024-09-26 02:02:45,570 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.296e+02 1.405e+02 1.536e+02 1.900e+02, threshold=2.809e+02, percent-clipped=0.0
2024-09-26 02:02:54,008 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.26 vs. limit=10.0
2024-09-26 02:02:59,569 INFO [train.py:1198] (1/4) Epoch 49, batch 1800, loss[loss=0.2069, ctc_loss=0.1336, cr_loss=0.3663, over 17038.00 frames. ], tot_loss[loss=0.1843, ctc_loss=0.1177, cr_loss=0.3331, over 3349451.41 frames. ], batch size: 52, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:03:05,366 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.whiten, num_groups=1, num_channels=384, metric=4.17 vs. limit=12.0
2024-09-26 02:03:09,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=881108.6666666666, ans=0.1
2024-09-26 02:04:03,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=881248.6666666666, ans=0.125
2024-09-26 02:04:03,713 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.74 vs. limit=10.0
2024-09-26 02:04:21,862 INFO [train.py:1198] (1/4) Epoch 49, batch 1850, loss[loss=0.1761, ctc_loss=0.1112, cr_loss=0.3245, over 17220.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1181, cr_loss=0.3336, over 3353709.56 frames. ], batch size: 47, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:04:23,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=881342.0, ans=0.1
2024-09-26 02:04:59,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=881435.3333333334, ans=0.09899494936611666
2024-09-26 02:05:01,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=881435.3333333334, ans=0.1
2024-09-26 02:05:07,566 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff3_skip_rate, batch_count=881435.3333333334, ans=0.0
2024-09-26 02:05:16,255 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.20 vs. limit=15.0
2024-09-26 02:05:18,843 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer2.prob, batch_count=881482.0, ans=0.125
2024-09-26 02:05:20,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.min_positive, batch_count=881482.0, ans=0.05
2024-09-26 02:05:26,656 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=881482.0, ans=0.125
2024-09-26 02:05:26,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=881482.0, ans=0.0
2024-09-26 02:05:33,164 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.348e+02 1.416e+02 1.489e+02 2.449e+02, threshold=2.831e+02, percent-clipped=0.0
2024-09-26 02:05:38,262 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=881528.6666666666, ans=0.035
2024-09-26 02:05:41,488 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=881528.6666666666, ans=0.0
2024-09-26 02:05:47,592 INFO [train.py:1198] (1/4) Epoch 49, batch 1900, loss[loss=0.1763, ctc_loss=0.1146, cr_loss=0.3086, over 17364.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.1182, cr_loss=0.3335, over 3357965.72 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:05:56,536 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.whiten_keys.whitening_limit, batch_count=881575.3333333334, ans=6.0
2024-09-26 02:05:57,763 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.86 vs. limit=15.0
2024-09-26 02:06:04,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.scale_min, batch_count=881622.0, ans=0.2
2024-09-26 02:06:40,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=881715.3333333334, ans=0.0
2024-09-26 02:07:10,385 INFO [train.py:1198] (1/4) Epoch 49, batch 1950, loss[loss=0.1947, ctc_loss=0.125, cr_loss=0.3483, over 17127.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1181, cr_loss=0.3336, over 3355742.92 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:07:10,649 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=881808.6666666666, ans=0.1
2024-09-26 02:07:18,718 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=881808.6666666666, ans=0.025
2024-09-26 02:07:45,717 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=881902.0, ans=0.0
2024-09-26 02:08:18,104 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.309e+02 1.411e+02 1.494e+02 2.442e+02, threshold=2.822e+02, percent-clipped=0.0
2024-09-26 02:08:32,518 INFO [train.py:1198] (1/4) Epoch 49, batch 2000, loss[loss=0.2092, ctc_loss=0.1345, cr_loss=0.3733, over 16651.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1178, cr_loss=0.3338, over 3359872.05 frames. ], batch size: 66, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 02:08:53,862 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.77 vs. limit=22.5
2024-09-26 02:08:58,287 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=882088.6666666666, ans=0.125
2024-09-26 02:09:20,589 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.skip_rate, batch_count=882182.0, ans=0.04949747468305833
2024-09-26 02:09:25,218 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.out_combiner.scale_min, batch_count=882182.0, ans=0.2
2024-09-26 02:09:57,472 INFO [train.py:1198] (1/4) Epoch 49, batch 2050, loss[loss=0.2548, ctc_loss=0.1645, cr_loss=0.4516, over 17014.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1184, cr_loss=0.3349, over 3347706.76 frames. ], batch size: 53, lr: 2.42e-03, grad_scale: 32.0
2024-09-26 02:10:24,951 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=882322.0, ans=0.125
2024-09-26 02:10:47,454 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=882415.3333333334, ans=0.125
2024-09-26 02:11:04,655 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.334e+02 1.409e+02 1.527e+02 2.352e+02, threshold=2.819e+02, percent-clipped=0.0
2024-09-26 02:11:17,568 INFO [train.py:1198] (1/4) Epoch 49, batch 2100, loss[loss=0.1897, ctc_loss=0.1224, cr_loss=0.3365, over 17018.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1188, cr_loss=0.3355, over 3348933.79 frames. ], batch size: 53, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:11:17,794 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=882508.6666666666, ans=0.125
2024-09-26 02:12:05,494 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.bypass.scale_min, batch_count=882602.0, ans=0.2
2024-09-26 02:12:16,486 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=882648.6666666666, ans=0.1
2024-09-26 02:12:24,309 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:12:35,278 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=882695.3333333334, ans=0.1
2024-09-26 02:12:39,852 INFO [train.py:1198] (1/4) Epoch 49, batch 2150, loss[loss=0.2132, ctc_loss=0.1397, cr_loss=0.3676, over 16981.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1187, cr_loss=0.3358, over 3351510.11 frames. ], batch size: 58, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:12:40,238 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=882742.0, ans=0.0
2024-09-26 02:12:52,854 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=882742.0, ans=0.1
2024-09-26 02:13:00,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=882788.6666666666, ans=0.125
2024-09-26 02:13:17,892 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.skip_rate, batch_count=882835.3333333334, ans=0.07
2024-09-26 02:13:21,051 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.skip_rate, batch_count=882835.3333333334, ans=0.04949747468305833
2024-09-26 02:13:22,795 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass.scale_min, batch_count=882835.3333333334, ans=0.2
2024-09-26 02:13:24,880 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.80 vs. limit=15.0
2024-09-26 02:13:29,025 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass_mid.scale_min, batch_count=882882.0, ans=0.2
2024-09-26 02:13:49,689 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.355e+02 1.416e+02 1.495e+02 3.199e+02, threshold=2.832e+02, percent-clipped=1.0
2024-09-26 02:14:02,463 INFO [train.py:1198] (1/4) Epoch 49, batch 2200, loss[loss=0.1977, ctc_loss=0.1256, cr_loss=0.3606, over 17027.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1185, cr_loss=0.3349, over 3354251.83 frames. ], batch size: 51, lr: 2.42e-03, grad_scale: 16.0
2024-09-26 02:14:05,872 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=882975.3333333334, ans=0.0
2024-09-26 02:14:12,315 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=882975.3333333334, ans=0.0
2024-09-26 02:14:15,606 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=882975.3333333334, ans=0.125
2024-09-26 02:14:50,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=883068.6666666666, ans=0.0
2024-09-26 02:14:59,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.min_positive, batch_count=883115.3333333334, ans=0.025
2024-09-26 02:15:13,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=883162.0, ans=0.125
2024-09-26 02:15:16,745 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer2.min_abs, batch_count=883162.0, ans=0.5
2024-09-26 02:15:27,836 INFO [train.py:1198] (1/4) Epoch 49, batch 2250, loss[loss=0.1872, ctc_loss=0.1178, cr_loss=0.3473, over 17006.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1179, cr_loss=0.3344, over 3350797.17 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:15:39,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=883208.6666666666, ans=0.2
2024-09-26 02:15:52,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.whiten.whitening_limit, batch_count=883255.3333333334, ans=15.0
2024-09-26 02:15:57,360 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.98 vs. limit=15.0
2024-09-26 02:16:08,508 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.37 vs. limit=15.0
2024-09-26 02:16:22,600 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.18 vs. limit=15.0
2024-09-26 02:16:33,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=883395.3333333334, ans=0.0
2024-09-26 02:16:34,923 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.298e+02 1.376e+02 1.462e+02 1.948e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-26 02:16:38,826 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.14 vs. limit=15.0
2024-09-26 02:16:47,759 INFO [train.py:1198] (1/4) Epoch 49, batch 2300, loss[loss=0.1546, ctc_loss=0.09491, cr_loss=0.2984, over 16974.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1179, cr_loss=0.334, over 3351171.24 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:16:54,786 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.26 vs. limit=10.0
2024-09-26 02:17:14,451 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.conv_module1.whiten, num_groups=1, num_channels=192, metric=9.02 vs. limit=15.0
2024-09-26 02:17:27,059 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.49 vs. limit=15.0
2024-09-26 02:18:08,408 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=883628.6666666666, ans=0.0
2024-09-26 02:18:13,002 INFO [train.py:1198] (1/4) Epoch 49, batch 2350, loss[loss=0.2152, ctc_loss=0.1397, cr_loss=0.3774, over 16098.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1178, cr_loss=0.3343, over 3350601.14 frames. ], batch size: 74, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:18:31,324 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=883722.0, ans=0.125
2024-09-26 02:18:31,499 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=8.94 vs. limit=15.0
2024-09-26 02:18:34,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=883722.0, ans=0.0
2024-09-26 02:18:42,631 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:18:47,284 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.min_positive, batch_count=883768.6666666666, ans=0.025
2024-09-26 02:18:54,981 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer2.prob, batch_count=883768.6666666666, ans=0.125
2024-09-26 02:19:12,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=883815.3333333334, ans=0.1
2024-09-26 02:19:20,302 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.282e+02 1.377e+02 1.471e+02 1.818e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-26 02:19:35,628 INFO [train.py:1198] (1/4) Epoch 49, batch 2400, loss[loss=0.1777, ctc_loss=0.1119, cr_loss=0.3289, over 17304.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3358, over 3360026.11 frames. ], batch size: 46, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:20:37,880 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=10.01 vs. limit=15.0
2024-09-26 02:20:53,681 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=884095.3333333334, ans=0.125
2024-09-26 02:20:58,123 INFO [train.py:1198] (1/4) Epoch 49, batch 2450, loss[loss=0.2066, ctc_loss=0.1345, cr_loss=0.3602, over 15301.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1183, cr_loss=0.3347, over 3362298.82 frames. ], batch size: 89, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:21:17,844 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer2.prob, batch_count=884188.6666666666, ans=0.125
2024-09-26 02:21:32,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=884235.3333333334, ans=0.125
2024-09-26 02:21:58,823 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=884282.0, ans=0.125
2024-09-26 02:22:09,462 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.309e+02 1.374e+02 1.463e+02 2.759e+02, threshold=2.748e+02, percent-clipped=1.0
2024-09-26 02:22:18,054 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer2.prob, batch_count=884328.6666666666, ans=0.125
2024-09-26 02:22:19,030 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.60 vs. limit=5.0
2024-09-26 02:22:20,948 INFO [train.py:1198] (1/4) Epoch 49, batch 2500, loss[loss=0.2356, ctc_loss=0.1555, cr_loss=0.4008, over 14904.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.119, cr_loss=0.3364, over 3359103.05 frames. ], batch size: 88, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:22:46,143 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=7.52 vs. limit=15.0
2024-09-26 02:22:47,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.86 vs. limit=15.0
2024-09-26 02:23:29,971 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=884562.0, ans=0.2
2024-09-26 02:23:31,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=884562.0, ans=0.0
2024-09-26 02:23:44,123 INFO [train.py:1198] (1/4) Epoch 49, batch 2550, loss[loss=0.1487, ctc_loss=0.09261, cr_loss=0.2804, over 17201.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1186, cr_loss=0.3358, over 3356685.58 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:24:05,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=884655.3333333334, ans=0.125
2024-09-26 02:24:12,131 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.35 vs. limit=10.0
2024-09-26 02:24:20,238 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.42 vs. limit=15.0
2024-09-26 02:24:30,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=884702.0, ans=0.0
2024-09-26 02:24:33,361 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=884748.6666666666, ans=0.2
2024-09-26 02:24:40,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=884748.6666666666, ans=0.1
2024-09-26 02:24:48,687 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=884748.6666666666, ans=0.1
2024-09-26 02:24:58,344 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.313e+02 1.367e+02 1.472e+02 2.257e+02, threshold=2.734e+02, percent-clipped=0.0
2024-09-26 02:25:09,543 INFO [train.py:1198] (1/4) Epoch 49, batch 2600, loss[loss=0.1742, ctc_loss=0.1107, cr_loss=0.3175, over 17355.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.118, cr_loss=0.3348, over 3361655.48 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:25:13,101 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=884842.0, ans=0.0
2024-09-26 02:25:16,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=884842.0, ans=0.09899494936611666
2024-09-26 02:25:21,072 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=884842.0, ans=0.1
2024-09-26 02:25:27,888 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.08 vs. limit=15.0
2024-09-26 02:25:46,722 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=884935.3333333334, ans=0.2
2024-09-26 02:26:01,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=884982.0, ans=0.2
2024-09-26 02:26:29,481 INFO [train.py:1198] (1/4) Epoch 49, batch 2650, loss[loss=0.1488, ctc_loss=0.09229, cr_loss=0.2827, over 17098.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1179, cr_loss=0.3342, over 3356945.34 frames. ], batch size: 40, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:26:47,314 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=885122.0, ans=0.125
2024-09-26 02:26:47,511 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=885122.0, ans=0.125
2024-09-26 02:27:18,813 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer1.prob, batch_count=885215.3333333334, ans=0.125
2024-09-26 02:27:23,659 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:27:35,085 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=12.16 vs. limit=22.5
2024-09-26 02:27:40,642 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.322e+02 1.427e+02 1.515e+02 2.814e+02, threshold=2.853e+02, percent-clipped=1.0
2024-09-26 02:27:44,007 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.8.prob, batch_count=885262.0, ans=0.125
2024-09-26 02:27:51,040 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=10.00 vs. limit=22.5
2024-09-26 02:27:51,958 INFO [train.py:1198] (1/4) Epoch 49, batch 2700, loss[loss=0.1846, ctc_loss=0.118, cr_loss=0.3332, over 17029.00 frames. ], tot_loss[loss=0.1843, ctc_loss=0.1175, cr_loss=0.334, over 3364841.97 frames. ], batch size: 51, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:27:53,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=885308.6666666666, ans=0.125
2024-09-26 02:28:45,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_skip_rate, batch_count=885448.6666666666, ans=0.0
2024-09-26 02:29:05,237 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=885495.3333333334, ans=0.025
2024-09-26 02:29:14,280 INFO [train.py:1198] (1/4) Epoch 49, batch 2750, loss[loss=0.1866, ctc_loss=0.1174, cr_loss=0.346, over 17257.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1184, cr_loss=0.3359, over 3368416.12 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:30:07,385 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=885682.0, ans=0.125
2024-09-26 02:30:12,202 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=885682.0, ans=0.125
2024-09-26 02:30:23,442 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=885728.6666666666, ans=0.0
2024-09-26 02:30:28,233 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.341e+02 1.441e+02 1.583e+02 2.930e+02, threshold=2.882e+02, percent-clipped=1.0
2024-09-26 02:30:33,197 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=885728.6666666666, ans=0.125
2024-09-26 02:30:39,197 INFO [train.py:1198] (1/4) Epoch 49, batch 2800, loss[loss=0.2216, ctc_loss=0.1462, cr_loss=0.377, over 17207.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1182, cr_loss=0.3348, over 3372491.77 frames. ], batch size: 55, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:30:50,767 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=885775.3333333334, ans=0.0
2024-09-26 02:31:54,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=885962.0, ans=0.125
2024-09-26 02:32:01,787 INFO [train.py:1198] (1/4) Epoch 49, batch 2850, loss[loss=0.2219, ctc_loss=0.1462, cr_loss=0.3781, over 12143.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1181, cr_loss=0.3347, over 3369081.65 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:32:26,331 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=886055.3333333334, ans=0.0
2024-09-26 02:32:41,240 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.32 vs. limit=15.0
2024-09-26 02:32:53,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=886148.6666666666, ans=0.1
2024-09-26 02:33:11,162 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:33:12,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=886195.3333333334, ans=0.125
2024-09-26 02:33:13,930 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.301e+02 1.414e+02 1.534e+02 2.528e+02, threshold=2.827e+02, percent-clipped=0.0
2024-09-26 02:33:25,144 INFO [train.py:1198] (1/4) Epoch 49, batch 2900, loss[loss=0.2245, ctc_loss=0.1457, cr_loss=0.3943, over 15047.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1177, cr_loss=0.3338, over 3358954.22 frames. ], batch size: 89, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:33:25,489 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=886242.0, ans=0.125
2024-09-26 02:33:51,251 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.96 vs. limit=15.0
2024-09-26 02:33:51,468 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=886288.6666666666, ans=10.0
2024-09-26 02:33:55,552 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=886335.3333333334, ans=0.125
2024-09-26 02:33:55,805 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=11.69 vs. limit=15.0
2024-09-26 02:34:03,793 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=11.43 vs. limit=12.0
2024-09-26 02:34:43,740 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=886428.6666666666, ans=0.125
2024-09-26 02:34:49,794 INFO [train.py:1198] (1/4) Epoch 49, batch 2950, loss[loss=0.2092, ctc_loss=0.1337, cr_loss=0.3779, over 17139.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.1181, cr_loss=0.3345, over 3351734.26 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:35:01,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer1.prob, batch_count=886475.3333333334, ans=0.125
2024-09-26 02:35:09,414 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.balancer.prob, batch_count=886522.0, ans=0.125
2024-09-26 02:35:18,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=886522.0, ans=0.0
2024-09-26 02:35:26,840 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff3_skip_rate, batch_count=886568.6666666666, ans=0.0
2024-09-26 02:35:39,661 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=886615.3333333334, ans=0.125
2024-09-26 02:35:49,201 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=886615.3333333334, ans=0.0
2024-09-26 02:35:58,783 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.303e+02 1.409e+02 1.487e+02 2.405e+02, threshold=2.817e+02, percent-clipped=0.0
2024-09-26 02:35:59,134 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.min_abs, batch_count=886662.0, ans=0.5
2024-09-26 02:35:59,213 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=886662.0, ans=0.125
2024-09-26 02:36:09,977 INFO [train.py:1198] (1/4) Epoch 49, batch 3000, loss[loss=0.1589, ctc_loss=0.1008, cr_loss=0.2909, over 17114.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1189, cr_loss=0.3351, over 3328267.41 frames. ], batch size: 40, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:36:09,978 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-26 02:36:25,651 INFO [train.py:1230] (1/4) Epoch 49, validation: loss=0.03501, ctc_loss=0.03501, cr_loss=1.043e-14, over 944034.00 frames. 
2024-09-26 02:36:25,652 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-26 02:36:38,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=886708.6666666666, ans=0.025
2024-09-26 02:36:43,216 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=886755.3333333334, ans=0.0
2024-09-26 02:37:47,487 INFO [train.py:1198] (1/4) Epoch 49, batch 3050, loss[loss=0.1512, ctc_loss=0.09252, cr_loss=0.2932, over 17033.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1187, cr_loss=0.3349, over 3337010.73 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:37:52,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=886942.0, ans=0.1
2024-09-26 02:38:04,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=886988.6666666666, ans=0.1
2024-09-26 02:38:54,743 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.193e+02 1.323e+02 1.385e+02 1.464e+02 2.781e+02, threshold=2.770e+02, percent-clipped=0.0
2024-09-26 02:38:57,531 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=11.91 vs. limit=15.0
2024-09-26 02:38:58,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=887128.6666666666, ans=0.0
2024-09-26 02:39:05,809 INFO [train.py:1198] (1/4) Epoch 49, batch 3100, loss[loss=0.1931, ctc_loss=0.1254, cr_loss=0.3389, over 17230.00 frames. ], tot_loss[loss=0.1852, ctc_loss=0.1184, cr_loss=0.3342, over 3341574.07 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:40:01,585 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=6.95 vs. limit=15.0
2024-09-26 02:40:05,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=887315.3333333334, ans=0.125
2024-09-26 02:40:10,300 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:40:22,995 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.92 vs. limit=6.0
2024-09-26 02:40:26,988 INFO [train.py:1198] (1/4) Epoch 49, batch 3150, loss[loss=0.1907, ctc_loss=0.1211, cr_loss=0.3477, over 15841.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.1181, cr_loss=0.3344, over 3352472.45 frames. ], batch size: 74, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:40:27,316 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer1.prob, batch_count=887408.6666666666, ans=0.125
2024-09-26 02:41:06,204 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=887502.0, ans=0.1
2024-09-26 02:41:08,065 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=13.56 vs. limit=22.5
2024-09-26 02:41:14,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=887548.6666666666, ans=0.125
2024-09-26 02:41:33,137 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff2_skip_rate, batch_count=887595.3333333334, ans=0.0
2024-09-26 02:41:35,930 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.299e+02 1.369e+02 1.479e+02 2.318e+02, threshold=2.737e+02, percent-clipped=0.0
2024-09-26 02:41:45,297 INFO [train.py:1198] (1/4) Epoch 49, batch 3200, loss[loss=0.1878, ctc_loss=0.122, cr_loss=0.3289, over 16862.00 frames. ], tot_loss[loss=0.184, ctc_loss=0.1174, cr_loss=0.333, over 3361087.76 frames. ], batch size: 58, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:41:45,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module1.balancer2.prob, batch_count=887642.0, ans=0.125
2024-09-26 02:42:01,224 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=887688.6666666666, ans=0.025
2024-09-26 02:42:27,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=887735.3333333334, ans=0.125
2024-09-26 02:42:32,349 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=887782.0, ans=0.0
2024-09-26 02:42:39,427 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=4.72 vs. limit=15.0
2024-09-26 02:42:51,802 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=887828.6666666666, ans=0.09899494936611666
2024-09-26 02:43:05,707 INFO [train.py:1198] (1/4) Epoch 49, batch 3250, loss[loss=0.1827, ctc_loss=0.1144, cr_loss=0.3417, over 17263.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1178, cr_loss=0.3337, over 3364762.34 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 32.0
2024-09-26 02:43:13,870 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=887875.3333333334, ans=0.0
2024-09-26 02:43:36,625 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:43:45,547 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=887968.6666666666, ans=0.1
2024-09-26 02:44:18,280 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.305e+02 1.378e+02 1.462e+02 1.990e+02, threshold=2.757e+02, percent-clipped=0.0
2024-09-26 02:44:26,131 INFO [train.py:1198] (1/4) Epoch 49, batch 3300, loss[loss=0.2042, ctc_loss=0.1318, cr_loss=0.362, over 17094.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1175, cr_loss=0.3335, over 3368136.58 frames. ], batch size: 49, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:44:42,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer2.prob, batch_count=888155.3333333334, ans=0.125
2024-09-26 02:45:44,479 INFO [train.py:1198] (1/4) Epoch 49, batch 3350, loss[loss=0.1788, ctc_loss=0.1141, cr_loss=0.3235, over 17011.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1177, cr_loss=0.3339, over 3369676.24 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:45:48,601 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=4.76 vs. limit=15.0
2024-09-26 02:45:49,452 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=888342.0, ans=0.0
2024-09-26 02:45:50,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=888342.0, ans=0.1
2024-09-26 02:46:06,564 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=888388.6666666666, ans=0.125
2024-09-26 02:46:08,265 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_skip_rate, batch_count=888388.6666666666, ans=0.0
2024-09-26 02:46:09,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=888388.6666666666, ans=0.0
2024-09-26 02:46:10,256 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=7.84 vs. limit=15.0
2024-09-26 02:46:16,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=888435.3333333334, ans=0.125
2024-09-26 02:46:22,597 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=7.48 vs. limit=15.0
2024-09-26 02:46:51,173 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=5.88 vs. limit=15.0
2024-09-26 02:46:54,935 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.181e+02 1.320e+02 1.389e+02 1.476e+02 1.760e+02, threshold=2.779e+02, percent-clipped=0.0
2024-09-26 02:47:02,746 INFO [train.py:1198] (1/4) Epoch 49, batch 3400, loss[loss=0.1561, ctc_loss=0.09976, cr_loss=0.2816, over 17068.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1175, cr_loss=0.3331, over 3369782.75 frames. ], batch size: 46, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:47:04,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=888575.3333333334, ans=0.125
2024-09-26 02:47:15,480 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=888575.3333333334, ans=0.1
2024-09-26 02:47:19,541 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=13.21 vs. limit=22.5
2024-09-26 02:47:24,921 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=888622.0, ans=0.1
2024-09-26 02:47:28,026 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=888622.0, ans=0.0
2024-09-26 02:48:06,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.max_abs, batch_count=888762.0, ans=10.0
2024-09-26 02:48:14,434 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module2.balancer1.prob, batch_count=888762.0, ans=0.125
2024-09-26 02:48:23,485 INFO [train.py:1198] (1/4) Epoch 49, batch 3450, loss[loss=0.2158, ctc_loss=0.1427, cr_loss=0.3657, over 16013.00 frames. ], tot_loss[loss=0.1838, ctc_loss=0.1173, cr_loss=0.3323, over 3352173.23 frames. ], batch size: 74, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:49:05,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=888902.0, ans=0.2
2024-09-26 02:49:05,352 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=888902.0, ans=0.1
2024-09-26 02:49:27,437 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.71 vs. limit=15.0
2024-09-26 02:49:27,525 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=11.73 vs. limit=22.5
2024-09-26 02:49:34,581 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.304e+02 1.377e+02 1.518e+02 2.125e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-26 02:49:36,427 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer1.max_abs, batch_count=888995.3333333334, ans=10.0
2024-09-26 02:49:40,814 INFO [train.py:1198] (1/4) Epoch 49, batch 3500, loss[loss=0.1904, ctc_loss=0.1231, cr_loss=0.3363, over 17324.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.1181, cr_loss=0.3341, over 3348438.03 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 8.0
2024-09-26 02:49:47,753 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=889042.0, ans=0.025
2024-09-26 02:50:08,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=889088.6666666666, ans=0.2
2024-09-26 02:50:11,260 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=889088.6666666666, ans=0.0
2024-09-26 02:50:17,679 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=889135.3333333334, ans=0.2
2024-09-26 02:50:19,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff3_skip_rate, batch_count=889135.3333333334, ans=0.0
2024-09-26 02:50:29,491 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=19.86 vs. limit=22.5
2024-09-26 02:50:32,195 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.11 vs. limit=22.5
2024-09-26 02:50:38,311 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff2_skip_rate, batch_count=889182.0, ans=0.0
2024-09-26 02:50:43,108 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=1.98 vs. limit=6.0
2024-09-26 02:50:44,458 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=889228.6666666666, ans=0.125
2024-09-26 02:50:58,413 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=889228.6666666666, ans=0.125
2024-09-26 02:51:01,305 INFO [train.py:1198] (1/4) Epoch 49, batch 3550, loss[loss=0.2196, ctc_loss=0.1404, cr_loss=0.396, over 17011.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1184, cr_loss=0.335, over 3345395.56 frames. ], batch size: 56, lr: 2.41e-03, grad_scale: 8.0
2024-09-26 02:51:56,272 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=889415.3333333334, ans=0.125
2024-09-26 02:52:13,579 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.292e+02 1.382e+02 1.482e+02 3.513e+02, threshold=2.765e+02, percent-clipped=1.0
2024-09-26 02:52:15,529 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=889462.0, ans=0.0
2024-09-26 02:52:19,985 INFO [train.py:1198] (1/4) Epoch 49, batch 3600, loss[loss=0.2014, ctc_loss=0.1273, cr_loss=0.3704, over 16921.00 frames. ], tot_loss[loss=0.1855, ctc_loss=0.1185, cr_loss=0.3351, over 3337006.87 frames. ], batch size: 58, lr: 2.41e-03, grad_scale: 16.0
2024-09-26 02:52:43,000 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=7.33 vs. limit=15.0
2024-09-26 02:52:51,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=889602.0, ans=0.1
2024-09-26 02:53:38,183 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass_mid.scale_min, batch_count=889695.3333333334, ans=0.2
2024-09-26 02:53:42,723 INFO [train.py:1198] (1/4) Epoch 49, batch 3650, loss[loss=0.1812, ctc_loss=0.1182, cr_loss=0.3147, over 17345.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1179, cr_loss=0.3339, over 3351360.51 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0
2024-09-26 02:53:43,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_skip_rate, batch_count=889742.0, ans=0.0
2024-09-26 02:54:08,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.hidden_balancer.prob, batch_count=889788.6666666666, ans=0.125
2024-09-26 02:54:26,780 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=889835.3333333334, ans=0.1
2024-09-26 02:54:32,237 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.08 vs. limit=15.0
2024-09-26 02:54:45,737 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.const_attention_rate, batch_count=889928.6666666666, ans=0.025
2024-09-26 02:54:55,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=889928.6666666666, ans=0.125
2024-09-26 02:54:56,317 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.314e+02 1.382e+02 1.463e+02 2.217e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-26 02:54:57,527 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.34 vs. limit=8.0
2024-09-26 02:55:01,083 INFO [train.py:1198] (1/4) Epoch 49, batch 3700, loss[loss=0.2121, ctc_loss=0.1398, cr_loss=0.3614, over 15289.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1186, cr_loss=0.3352, over 3347193.68 frames. ], batch size: 89, lr: 2.41e-03, grad_scale: 8.0
2024-09-26 02:55:26,817 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=890022.0, ans=0.2
2024-09-26 02:55:36,298 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=890068.6666666666, ans=0.1
2024-09-26 02:55:44,008 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=890068.6666666666, ans=0.2
2024-09-26 02:55:54,020 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:56:01,920 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer2.prob, batch_count=890115.3333333334, ans=0.125
2024-09-26 02:56:14,267 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=890162.0, ans=0.0
2024-09-26 02:56:18,191 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.54 vs. limit=12.0
2024-09-26 02:56:19,330 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass_mid.scale_min, batch_count=890208.6666666666, ans=0.2
2024-09-26 02:56:20,583 INFO [train.py:1198] (1/4) Epoch 49, batch 3750, loss[loss=0.2132, ctc_loss=0.1396, cr_loss=0.368, over 17014.00 frames. ], tot_loss[loss=0.1863, ctc_loss=0.1191, cr_loss=0.336, over 3353315.18 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0
2024-09-26 02:56:45,145 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=11.38 vs. limit=15.0
2024-09-26 02:56:46,286 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 02:56:57,411 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=890302.0, ans=0.125
2024-09-26 02:57:35,557 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.349e+02 1.442e+02 1.558e+02 6.975e+02, threshold=2.884e+02, percent-clipped=1.0
2024-09-26 02:57:40,293 INFO [train.py:1198] (1/4) Epoch 49, batch 3800, loss[loss=0.1698, ctc_loss=0.1061, cr_loss=0.3183, over 17309.00 frames. ], tot_loss[loss=0.1879, ctc_loss=0.1203, cr_loss=0.3378, over 3335596.00 frames. ], batch size: 51, lr: 2.40e-03, grad_scale: 8.0
2024-09-26 02:58:07,124 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.scale_min, batch_count=890488.6666666666, ans=0.2
2024-09-26 02:58:54,748 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn2.whiten, num_groups=1, num_channels=384, metric=14.85 vs. limit=22.5
2024-09-26 02:58:58,407 INFO [train.py:1198] (1/4) Epoch 49, batch 3850, loss[loss=0.2292, ctc_loss=0.1504, cr_loss=0.3941, over 14908.00 frames. ], tot_loss[loss=0.1904, ctc_loss=0.1223, cr_loss=0.3402, over 3287175.60 frames. ], batch size: 89, lr: 2.40e-03, grad_scale: 8.0
2024-09-26 02:58:58,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff2_skip_rate, batch_count=890675.3333333334, ans=0.0
2024-09-26 02:59:00,803 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=890675.3333333334, ans=0.035
2024-09-26 02:59:12,925 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=890722.0, ans=0.1
2024-09-26 02:59:12,945 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_skip_rate, batch_count=890722.0, ans=0.0
2024-09-26 02:59:16,022 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.ff2_skip_rate, batch_count=890722.0, ans=0.0
2024-09-26 02:59:22,084 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer2.prob, batch_count=890722.0, ans=0.125
2024-09-26 02:59:23,542 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=890722.0, ans=0.125
2024-09-26 02:59:31,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.ff2_skip_rate, batch_count=890768.6666666666, ans=0.0
2024-09-26 02:59:40,169 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=890768.6666666666, ans=0.125
2024-09-26 02:59:41,721 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=890768.6666666666, ans=0.04949747468305833
2024-09-26 02:59:53,819 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff3_skip_rate, batch_count=890815.3333333334, ans=0.0
2024-09-26 03:00:02,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module1.balancer1.max_abs, batch_count=890862.0, ans=10.0
2024-09-26 03:00:56,111 INFO [train.py:1198] (1/4) Epoch 50, batch 0, loss[loss=0.1747, ctc_loss=0.1097, cr_loss=0.3254, over 17313.00 frames. ], tot_loss[loss=0.1747, ctc_loss=0.1097, cr_loss=0.3254, over 17313.00 frames. ], batch size: 51, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:00:56,112 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-26 03:01:04,529 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.1.self_attn_weights, attn_weights_entropy = tensor([4.6020, 4.1050, 4.0388, 4.2310], device='cuda:1')
2024-09-26 03:01:07,052 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.2.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([4.6630, 4.5258, 4.3202, 4.2708], device='cuda:1')
2024-09-26 03:01:12,094 INFO [train.py:1230] (1/4) Epoch 50, validation: loss=0.03452, ctc_loss=0.03452, cr_loss=1.145e-14, over 944034.00 frames. 
2024-09-26 03:01:12,094 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-26 03:01:13,602 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.424e+02 1.584e+02 1.731e+02 2.410e+02, threshold=3.169e+02, percent-clipped=0.0
2024-09-26 03:01:20,032 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.dropout.p, batch_count=890890.0, ans=0.1
2024-09-26 03:01:32,319 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=11.92 vs. limit=22.5
2024-09-26 03:01:35,050 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=6.66 vs. limit=15.0
2024-09-26 03:01:58,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=890983.3333333334, ans=0.0
2024-09-26 03:02:01,455 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.50 vs. limit=15.0
2024-09-26 03:02:07,576 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=891030.0, ans=0.1
2024-09-26 03:02:20,368 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=891076.6666666666, ans=0.125
2024-09-26 03:02:34,238 INFO [train.py:1198] (1/4) Epoch 50, batch 50, loss[loss=0.2287, ctc_loss=0.1496, cr_loss=0.3956, over 11637.00 frames. ], tot_loss[loss=0.1813, ctc_loss=0.1151, cr_loss=0.3312, over 765673.19 frames. ], batch size: 123, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:02:36,454 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=12.43 vs. limit=15.0
2024-09-26 03:03:03,556 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=891170.0, ans=0.0
2024-09-26 03:03:17,895 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.32 vs. limit=10.0
2024-09-26 03:03:24,197 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=13.30 vs. limit=15.0
2024-09-26 03:03:30,194 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=891263.3333333334, ans=0.125
2024-09-26 03:03:36,533 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=891263.3333333334, ans=0.125
2024-09-26 03:03:57,321 INFO [train.py:1198] (1/4) Epoch 50, batch 100, loss[loss=0.1966, ctc_loss=0.1284, cr_loss=0.3412, over 16422.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1173, cr_loss=0.3343, over 1345412.42 frames. ], batch size: 66, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:03:58,964 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.291e+02 1.362e+02 1.431e+02 2.417e+02, threshold=2.724e+02, percent-clipped=0.0
2024-09-26 03:03:59,258 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer1.prob, batch_count=891356.6666666666, ans=0.125
2024-09-26 03:04:00,825 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=891356.6666666666, ans=0.125
2024-09-26 03:04:00,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=891356.6666666666, ans=0.125
2024-09-26 03:04:24,782 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=891403.3333333334, ans=0.2
2024-09-26 03:04:27,995 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.skip_rate, batch_count=891450.0, ans=0.09899494936611666
2024-09-26 03:04:35,770 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=7.19 vs. limit=15.0
2024-09-26 03:04:48,133 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=891496.6666666666, ans=0.125
2024-09-26 03:04:51,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=891496.6666666666, ans=0.035
2024-09-26 03:04:54,575 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.conv_module1.whiten, num_groups=1, num_channels=256, metric=7.36 vs. limit=15.0
2024-09-26 03:05:22,561 INFO [train.py:1198] (1/4) Epoch 50, batch 150, loss[loss=0.1877, ctc_loss=0.1233, cr_loss=0.3217, over 16027.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1175, cr_loss=0.3346, over 1793734.70 frames. ], batch size: 74, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:05:57,570 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=512, metric=9.37 vs. limit=15.0
2024-09-26 03:06:17,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.const_attention_rate, batch_count=891730.0, ans=0.025
2024-09-26 03:06:20,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass_mid.scale_min, batch_count=891730.0, ans=0.2
2024-09-26 03:06:45,775 INFO [train.py:1198] (1/4) Epoch 50, batch 200, loss[loss=0.2454, ctc_loss=0.1587, cr_loss=0.4337, over 15154.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1185, cr_loss=0.3363, over 2137279.75 frames. ], batch size: 89, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:06:47,296 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.325e+02 1.401e+02 1.516e+02 2.050e+02, threshold=2.802e+02, percent-clipped=0.0
2024-09-26 03:06:49,210 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=891823.3333333334, ans=0.0
2024-09-26 03:07:11,256 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward3.hidden_balancer.prob, batch_count=891870.0, ans=0.125
2024-09-26 03:07:12,984 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.skip_rate, batch_count=891870.0, ans=0.07
2024-09-26 03:07:17,935 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer1.prob, batch_count=891916.6666666666, ans=0.125
2024-09-26 03:07:32,818 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=4.02 vs. limit=15.0
2024-09-26 03:07:40,154 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.balancer2.prob, batch_count=891963.3333333334, ans=0.125
2024-09-26 03:07:48,187 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass.skip_rate, batch_count=892010.0, ans=0.07
2024-09-26 03:07:53,170 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=892010.0, ans=0.1
2024-09-26 03:08:05,418 INFO [train.py:1198] (1/4) Epoch 50, batch 250, loss[loss=0.2157, ctc_loss=0.1404, cr_loss=0.3765, over 16748.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.1179, cr_loss=0.3349, over 2416630.81 frames. ], batch size: 61, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:08:16,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.const_attention_rate, batch_count=892056.6666666666, ans=0.025
2024-09-26 03:08:29,362 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=892103.3333333334, ans=0.0
2024-09-26 03:08:37,889 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_whiten.whitening_limit, batch_count=892103.3333333334, ans=15.0
2024-09-26 03:08:51,833 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=3.68 vs. limit=10.0
2024-09-26 03:08:56,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass_mid.scale_min, batch_count=892196.6666666666, ans=0.2
2024-09-26 03:09:22,974 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=7.70 vs. limit=15.0
2024-09-26 03:09:23,861 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=892243.3333333334, ans=0.0
2024-09-26 03:09:28,275 INFO [train.py:1198] (1/4) Epoch 50, batch 300, loss[loss=0.1983, ctc_loss=0.1301, cr_loss=0.3408, over 17139.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1182, cr_loss=0.3349, over 2631184.63 frames. ], batch size: 48, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:09:29,767 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.289e+02 1.359e+02 1.478e+02 2.731e+02, threshold=2.717e+02, percent-clipped=0.0
2024-09-26 03:09:46,040 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=892336.6666666666, ans=0.125
2024-09-26 03:09:48,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.05 vs. limit=15.0
2024-09-26 03:09:52,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=892336.6666666666, ans=0.2
2024-09-26 03:10:05,382 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=892383.3333333334, ans=0.0
2024-09-26 03:10:55,339 INFO [train.py:1198] (1/4) Epoch 50, batch 350, loss[loss=0.1873, ctc_loss=0.1193, cr_loss=0.3395, over 17070.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1176, cr_loss=0.3344, over 2802208.48 frames. ], batch size: 46, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:11:21,563 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 03:11:31,632 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=384, metric=4.41 vs. limit=15.0
2024-09-26 03:11:40,494 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=11.88 vs. limit=15.0
2024-09-26 03:11:45,226 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn2.whiten, num_groups=1, num_channels=512, metric=9.48 vs. limit=22.5
2024-09-26 03:11:57,698 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=14.10 vs. limit=15.0
2024-09-26 03:12:17,699 INFO [train.py:1198] (1/4) Epoch 50, batch 400, loss[loss=0.1849, ctc_loss=0.1148, cr_loss=0.3502, over 17019.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1178, cr_loss=0.3347, over 2930983.56 frames. ], batch size: 39, lr: 2.38e-03, grad_scale: 32.0
2024-09-26 03:12:19,212 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.302e+02 1.364e+02 1.437e+02 1.796e+02, threshold=2.729e+02, percent-clipped=0.0
2024-09-26 03:12:29,294 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.3.self_attn_weights, loss-sum=0.000e+00
2024-09-26 03:13:03,868 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=892896.6666666666, ans=0.0
2024-09-26 03:13:03,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=892896.6666666666, ans=0.1
2024-09-26 03:13:39,998 INFO [train.py:1198] (1/4) Epoch 50, batch 450, loss[loss=0.1374, ctc_loss=0.08419, cr_loss=0.2661, over 17094.00 frames. ], tot_loss[loss=0.1844, ctc_loss=0.1175, cr_loss=0.3343, over 3027335.52 frames. ], batch size: 43, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:13:55,890 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module1.balancer1.prob, batch_count=893036.6666666666, ans=0.125
2024-09-26 03:14:07,003 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=893036.6666666666, ans=0.0
2024-09-26 03:14:14,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=893083.3333333334, ans=10.0
2024-09-26 03:15:02,680 INFO [train.py:1198] (1/4) Epoch 50, batch 500, loss[loss=0.2194, ctc_loss=0.1427, cr_loss=0.3835, over 16461.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1173, cr_loss=0.334, over 3108623.98 frames. ], batch size: 66, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:15:05,799 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.293e+02 1.377e+02 1.474e+02 1.981e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-26 03:15:27,341 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=893270.0, ans=0.025
2024-09-26 03:15:46,741 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=893316.6666666666, ans=0.125
2024-09-26 03:15:48,196 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=893316.6666666666, ans=0.0
2024-09-26 03:15:54,704 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=893363.3333333334, ans=0.2
2024-09-26 03:16:23,769 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=893410.0, ans=0.125
2024-09-26 03:16:26,694 INFO [train.py:1198] (1/4) Epoch 50, batch 550, loss[loss=0.198, ctc_loss=0.1292, cr_loss=0.3437, over 16875.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1173, cr_loss=0.3341, over 3167370.67 frames. ], batch size: 58, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:16:31,858 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=893456.6666666666, ans=0.125
2024-09-26 03:16:36,010 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer2.prob, batch_count=893456.6666666666, ans=0.125
2024-09-26 03:16:38,419 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.84 vs. limit=10.0
2024-09-26 03:17:16,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=893596.6666666666, ans=0.1
2024-09-26 03:17:47,275 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.62 vs. limit=15.0
2024-09-26 03:17:48,890 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=9.52 vs. limit=15.0
2024-09-26 03:17:49,349 INFO [train.py:1198] (1/4) Epoch 50, batch 600, loss[loss=0.2197, ctc_loss=0.1424, cr_loss=0.3864, over 15122.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1174, cr_loss=0.3341, over 3204838.06 frames. ], batch size: 89, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:17:52,578 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.327e+02 1.372e+02 1.501e+02 3.825e+02, threshold=2.745e+02, percent-clipped=1.0
2024-09-26 03:18:23,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten1.whitening_limit, batch_count=893783.3333333334, ans=10.0
2024-09-26 03:18:29,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer2.prob, batch_count=893783.3333333334, ans=0.125
2024-09-26 03:18:30,928 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=893783.3333333334, ans=0.1
2024-09-26 03:18:48,459 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=893830.0, ans=0.125
2024-09-26 03:18:58,343 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.whiten, num_groups=1, num_channels=384, metric=4.97 vs. limit=12.0
2024-09-26 03:19:12,059 INFO [train.py:1198] (1/4) Epoch 50, batch 650, loss[loss=0.1666, ctc_loss=0.1052, cr_loss=0.3068, over 17040.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1174, cr_loss=0.3337, over 3239404.67 frames. ], batch size: 44, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:19:12,404 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=893923.3333333334, ans=0.1
2024-09-26 03:19:56,911 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=894016.6666666666, ans=0.95
2024-09-26 03:20:12,688 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff2_skip_rate, batch_count=894063.3333333334, ans=0.0
2024-09-26 03:20:14,219 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=894063.3333333334, ans=0.125
2024-09-26 03:20:17,448 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=894110.0, ans=0.0
2024-09-26 03:20:37,744 INFO [train.py:1198] (1/4) Epoch 50, batch 700, loss[loss=0.1734, ctc_loss=0.111, cr_loss=0.3118, over 17034.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1175, cr_loss=0.3336, over 3271847.68 frames. ], batch size: 44, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:20:40,921 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.326e+02 1.434e+02 1.550e+02 1.872e+02, threshold=2.869e+02, percent-clipped=0.0
2024-09-26 03:21:04,049 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=10.31 vs. limit=15.0
2024-09-26 03:21:33,486 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.56 vs. limit=15.0
2024-09-26 03:21:37,880 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.bypass.scale_min, batch_count=894296.6666666666, ans=0.2
2024-09-26 03:21:47,657 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=894343.3333333334, ans=0.0
2024-09-26 03:21:54,058 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward2.hidden_balancer.prob, batch_count=894343.3333333334, ans=0.125
2024-09-26 03:21:54,854 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn1.whiten, num_groups=1, num_channels=192, metric=9.64 vs. limit=22.5
2024-09-26 03:21:59,199 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=10.75 vs. limit=12.0
2024-09-26 03:22:00,146 INFO [train.py:1198] (1/4) Epoch 50, batch 750, loss[loss=0.1575, ctc_loss=0.0966, cr_loss=0.3045, over 17111.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1182, cr_loss=0.3347, over 3289471.65 frames. ], batch size: 40, lr: 2.38e-03, grad_scale: 16.0
2024-09-26 03:22:16,745 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=256, metric=12.88 vs. limit=22.5
2024-09-26 03:22:25,888 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.ff2_skip_rate, batch_count=894436.6666666666, ans=0.0
2024-09-26 03:22:54,460 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_skip_rate, batch_count=894530.0, ans=0.0
2024-09-26 03:22:55,897 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.5.prob, batch_count=894530.0, ans=0.125
2024-09-26 03:23:02,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer2.prob, batch_count=894576.6666666666, ans=0.125
2024-09-26 03:23:22,024 INFO [train.py:1198] (1/4) Epoch 50, batch 800, loss[loss=0.1549, ctc_loss=0.09746, cr_loss=0.2874, over 17261.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.118, cr_loss=0.334, over 3311844.84 frames. ], batch size: 42, lr: 2.37e-03, grad_scale: 32.0
2024-09-26 03:23:23,926 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=894623.3333333334, ans=0.125
2024-09-26 03:23:25,275 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.307e+02 1.380e+02 1.476e+02 1.772e+02, threshold=2.760e+02, percent-clipped=0.0
2024-09-26 03:23:36,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.const_attention_rate, batch_count=894670.0, ans=0.025
2024-09-26 03:23:43,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=894670.0, ans=0.125
2024-09-26 03:24:03,931 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=894716.6666666666, ans=0.1
2024-09-26 03:24:07,947 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=4.77 vs. limit=10.0
2024-09-26 03:24:10,421 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=894763.3333333334, ans=0.1
2024-09-26 03:24:21,761 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.balancer2.prob, batch_count=894763.3333333334, ans=0.125
2024-09-26 03:24:37,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=894810.0, ans=0.2
2024-09-26 03:24:37,579 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=894810.0, ans=0.125
2024-09-26 03:24:45,378 INFO [train.py:1198] (1/4) Epoch 50, batch 850, loss[loss=0.1586, ctc_loss=0.1007, cr_loss=0.2895, over 16324.00 frames. ], tot_loss[loss=0.1843, ctc_loss=0.1177, cr_loss=0.333, over 3318309.88 frames. ], batch size: 36, lr: 2.37e-03, grad_scale: 32.0
2024-09-26 03:24:49,208 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.63 vs. limit=15.0
2024-09-26 03:26:01,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=895043.3333333334, ans=0.2
2024-09-26 03:26:04,394 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=895043.3333333334, ans=0.0
2024-09-26 03:26:08,943 INFO [train.py:1198] (1/4) Epoch 50, batch 900, loss[loss=0.1513, ctc_loss=0.0928, cr_loss=0.2923, over 17103.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.119, cr_loss=0.336, over 3320192.32 frames. ], batch size: 40, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:26:13,763 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.310e+02 1.400e+02 1.511e+02 3.836e+02, threshold=2.800e+02, percent-clipped=1.0
2024-09-26 03:26:15,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=895090.0, ans=0.125
2024-09-26 03:26:26,930 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module2.balancer1.prob, batch_count=895136.6666666666, ans=0.125
2024-09-26 03:27:12,695 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff2_skip_rate, batch_count=895230.0, ans=0.0
2024-09-26 03:27:14,399 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.ff3_skip_rate, batch_count=895276.6666666666, ans=0.0
2024-09-26 03:27:24,484 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=6.72 vs. limit=15.0
2024-09-26 03:27:29,549 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.88 vs. limit=22.5
2024-09-26 03:27:31,734 INFO [train.py:1198] (1/4) Epoch 50, batch 950, loss[loss=0.2202, ctc_loss=0.1424, cr_loss=0.389, over 16983.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3354, over 3323153.03 frames. ], batch size: 53, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:27:46,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.balancer2.prob, batch_count=895370.0, ans=0.125
2024-09-26 03:28:09,479 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=895416.6666666666, ans=0.0
2024-09-26 03:28:16,073 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=9.80 vs. limit=15.0
2024-09-26 03:28:27,028 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=895463.3333333334, ans=0.125
2024-09-26 03:28:35,118 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=895463.3333333334, ans=0.125
2024-09-26 03:28:54,018 INFO [train.py:1198] (1/4) Epoch 50, batch 1000, loss[loss=0.1567, ctc_loss=0.1002, cr_loss=0.2827, over 17019.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1187, cr_loss=0.335, over 3332744.52 frames. ], batch size: 44, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:28:57,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=895556.6666666666, ans=0.2
2024-09-26 03:28:58,809 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 1.309e+02 1.391e+02 1.502e+02 1.865e+02, threshold=2.782e+02, percent-clipped=0.0
2024-09-26 03:29:11,340 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.self_attn1.whiten, num_groups=1, num_channels=192, metric=13.14 vs. limit=22.5
2024-09-26 03:29:15,390 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=895603.3333333334, ans=0.025
2024-09-26 03:29:23,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=895603.3333333334, ans=0.125
2024-09-26 03:29:30,359 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=895650.0, ans=0.0
2024-09-26 03:29:36,659 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=895650.0, ans=0.125
2024-09-26 03:29:49,857 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=895696.6666666666, ans=0.125
2024-09-26 03:29:51,212 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=895696.6666666666, ans=0.1
2024-09-26 03:30:02,584 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=895743.3333333334, ans=0.0
2024-09-26 03:30:17,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=895790.0, ans=0.125
2024-09-26 03:30:19,366 INFO [train.py:1198] (1/4) Epoch 50, batch 1050, loss[loss=0.2019, ctc_loss=0.1295, cr_loss=0.3618, over 16495.00 frames. ], tot_loss[loss=0.1843, ctc_loss=0.1177, cr_loss=0.3329, over 3326447.10 frames. ], batch size: 66, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:30:22,743 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward3.hidden_balancer.prob, batch_count=895790.0, ans=0.125
2024-09-26 03:30:27,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=895790.0, ans=0.125
2024-09-26 03:30:34,937 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=5.38 vs. limit=15.0
2024-09-26 03:30:59,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=895883.3333333334, ans=0.125
2024-09-26 03:31:01,482 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer2.prob, batch_count=895883.3333333334, ans=0.125
2024-09-26 03:31:44,417 INFO [train.py:1198] (1/4) Epoch 50, batch 1100, loss[loss=0.1835, ctc_loss=0.1164, cr_loss=0.3352, over 17180.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1179, cr_loss=0.3339, over 3341084.96 frames. ], batch size: 41, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:31:49,230 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.312e+02 1.377e+02 1.461e+02 2.179e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-26 03:31:55,983 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.ff2_skip_rate, batch_count=896023.3333333334, ans=0.0
2024-09-26 03:32:05,516 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.bypass.scale_min, batch_count=896070.0, ans=0.2
2024-09-26 03:32:10,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=896070.0, ans=0.1
2024-09-26 03:32:24,668 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.skip_rate, batch_count=896116.6666666666, ans=0.07
2024-09-26 03:32:43,762 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=896163.3333333334, ans=0.125
2024-09-26 03:32:45,555 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.bypass_mid.scale_min, batch_count=896163.3333333334, ans=0.2
2024-09-26 03:33:04,559 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=5.28 vs. limit=15.0
2024-09-26 03:33:06,912 INFO [train.py:1198] (1/4) Epoch 50, batch 1150, loss[loss=0.1773, ctc_loss=0.112, cr_loss=0.3262, over 17031.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1177, cr_loss=0.3343, over 3351040.69 frames. ], batch size: 44, lr: 2.37e-03, grad_scale: 8.0
2024-09-26 03:33:21,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=896303.3333333334, ans=0.025
2024-09-26 03:34:00,800 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=6.76 vs. limit=10.0
2024-09-26 03:34:16,509 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.balancer1.prob, batch_count=896443.3333333334, ans=0.125
2024-09-26 03:34:22,895 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=896443.3333333334, ans=0.0
2024-09-26 03:34:29,985 INFO [train.py:1198] (1/4) Epoch 50, batch 1200, loss[loss=0.1984, ctc_loss=0.1292, cr_loss=0.3458, over 11688.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1178, cr_loss=0.3345, over 3351928.06 frames. ], batch size: 123, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:34:36,176 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.309e+02 1.371e+02 1.493e+02 2.008e+02, threshold=2.741e+02, percent-clipped=0.0
2024-09-26 03:34:44,575 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=896536.6666666666, ans=0.0
2024-09-26 03:35:16,543 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.balancer1.prob, batch_count=896630.0, ans=0.125
2024-09-26 03:35:35,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=896676.6666666666, ans=0.125
2024-09-26 03:35:36,090 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=14.42 vs. limit=15.0
2024-09-26 03:35:52,907 INFO [train.py:1198] (1/4) Epoch 50, batch 1250, loss[loss=0.2198, ctc_loss=0.1409, cr_loss=0.3945, over 16997.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1179, cr_loss=0.3345, over 3341976.41 frames. ], batch size: 53, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:36:02,310 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.convnext.out_whiten, num_groups=1, num_channels=128, metric=4.81 vs. limit=5.0
2024-09-26 03:36:13,986 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward2.hidden_balancer.prob, batch_count=896770.0, ans=0.125
2024-09-26 03:36:35,666 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=9.41 vs. limit=15.0
2024-09-26 03:37:15,374 INFO [train.py:1198] (1/4) Epoch 50, batch 1300, loss[loss=0.1698, ctc_loss=0.1035, cr_loss=0.3315, over 17235.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1176, cr_loss=0.3332, over 3335376.63 frames. ], batch size: 44, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:37:15,777 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=896956.6666666666, ans=0.125
2024-09-26 03:37:21,700 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.310e+02 1.392e+02 1.499e+02 2.433e+02, threshold=2.784e+02, percent-clipped=0.0
2024-09-26 03:37:29,838 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=897003.3333333334, ans=0.0
2024-09-26 03:38:38,299 INFO [train.py:1198] (1/4) Epoch 50, batch 1350, loss[loss=0.1667, ctc_loss=0.1047, cr_loss=0.3097, over 17091.00 frames. ], tot_loss[loss=0.1836, ctc_loss=0.1171, cr_loss=0.3324, over 3350643.06 frames. ], batch size: 46, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:39:01,109 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=897236.6666666666, ans=0.125
2024-09-26 03:39:40,636 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.max_abs, batch_count=897330.0, ans=10.0
2024-09-26 03:39:46,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=897376.6666666666, ans=0.1
2024-09-26 03:40:01,285 INFO [train.py:1198] (1/4) Epoch 50, batch 1400, loss[loss=0.2067, ctc_loss=0.1316, cr_loss=0.3756, over 16031.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1177, cr_loss=0.3343, over 3358206.05 frames. ], batch size: 74, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:40:07,585 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.312e+02 1.402e+02 1.516e+02 2.766e+02, threshold=2.805e+02, percent-clipped=0.0
2024-09-26 03:40:09,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=897423.3333333334, ans=0.0
2024-09-26 03:40:53,779 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=897563.3333333334, ans=0.1
2024-09-26 03:40:53,839 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff3_skip_rate, batch_count=897563.3333333334, ans=0.0
2024-09-26 03:40:53,927 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.bypass.scale_min, batch_count=897563.3333333334, ans=0.2
2024-09-26 03:41:16,920 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=12.44 vs. limit=15.0
2024-09-26 03:41:21,961 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.77 vs. limit=6.0
2024-09-26 03:41:23,489 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.52 vs. limit=10.0
2024-09-26 03:41:24,326 INFO [train.py:1198] (1/4) Epoch 50, batch 1450, loss[loss=0.1617, ctc_loss=0.1018, cr_loss=0.2996, over 17215.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1177, cr_loss=0.3344, over 3364389.89 frames. ], batch size: 47, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:41:42,987 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=897703.3333333334, ans=0.2
2024-09-26 03:41:59,053 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.ff2_skip_rate, batch_count=897750.0, ans=0.0
2024-09-26 03:42:26,247 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.feed_forward1.out_proj.dropout_p, batch_count=897796.6666666666, ans=0.1
2024-09-26 03:42:34,417 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=897843.3333333334, ans=0.125
2024-09-26 03:42:35,988 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer2.prob, batch_count=897843.3333333334, ans=0.125
2024-09-26 03:42:37,546 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.ff3_skip_rate, batch_count=897843.3333333334, ans=0.0
2024-09-26 03:42:46,929 INFO [train.py:1198] (1/4) Epoch 50, batch 1500, loss[loss=0.1824, ctc_loss=0.1173, cr_loss=0.3254, over 17143.00 frames. ], tot_loss[loss=0.1854, ctc_loss=0.1184, cr_loss=0.335, over 3353141.81 frames. ], batch size: 48, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:42:53,350 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.299e+02 1.380e+02 1.479e+02 2.541e+02, threshold=2.761e+02, percent-clipped=0.0
2024-09-26 03:43:03,113 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.11 vs. limit=15.0
2024-09-26 03:43:05,755 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=2.510e-03
2024-09-26 03:43:17,083 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.nonlin_attention.balancer.max_positive, batch_count=897936.6666666666, ans=0.95
2024-09-26 03:43:26,724 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=897983.3333333334, ans=0.125
2024-09-26 03:43:41,074 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer1.prob, batch_count=898030.0, ans=0.125
2024-09-26 03:44:06,725 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module1.balancer2.prob, batch_count=898076.6666666666, ans=0.125
2024-09-26 03:44:09,794 INFO [train.py:1198] (1/4) Epoch 50, batch 1550, loss[loss=0.2042, ctc_loss=0.1299, cr_loss=0.3714, over 16998.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1181, cr_loss=0.3348, over 3359945.00 frames. ], batch size: 53, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:44:22,812 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.max_positive, batch_count=898123.3333333334, ans=0.95
2024-09-26 03:44:57,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass_mid.scale_min, batch_count=898216.6666666666, ans=0.2
2024-09-26 03:45:06,887 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=898263.3333333334, ans=0.025
2024-09-26 03:45:35,381 INFO [train.py:1198] (1/4) Epoch 50, batch 1600, loss[loss=0.2223, ctc_loss=0.1451, cr_loss=0.386, over 16708.00 frames. ], tot_loss[loss=0.1853, ctc_loss=0.1183, cr_loss=0.335, over 3366229.00 frames. ], batch size: 61, lr: 2.37e-03, grad_scale: 32.0
2024-09-26 03:45:41,792 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.345e+02 1.418e+02 1.518e+02 2.144e+02, threshold=2.837e+02, percent-clipped=0.0
2024-09-26 03:46:01,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_skip_rate, batch_count=898403.3333333334, ans=0.0
2024-09-26 03:46:21,085 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=384, metric=4.13 vs. limit=15.0
2024-09-26 03:46:29,550 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=898496.6666666666, ans=0.1
2024-09-26 03:46:42,055 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=898543.3333333334, ans=0.2
2024-09-26 03:46:57,840 INFO [train.py:1198] (1/4) Epoch 50, batch 1650, loss[loss=0.1726, ctc_loss=0.1093, cr_loss=0.3164, over 17014.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1177, cr_loss=0.3339, over 3369316.87 frames. ], batch size: 39, lr: 2.37e-03, grad_scale: 32.0
2024-09-26 03:47:17,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=898636.6666666666, ans=0.1
2024-09-26 03:47:36,628 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=898683.3333333334, ans=0.125
2024-09-26 03:48:08,200 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.attention_skip_rate, batch_count=898776.6666666666, ans=0.0
2024-09-26 03:48:10,170 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.63 vs. limit=15.0
2024-09-26 03:48:11,303 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.balancer.prob, batch_count=898776.6666666666, ans=0.125
2024-09-26 03:48:20,713 INFO [train.py:1198] (1/4) Epoch 50, batch 1700, loss[loss=0.1942, ctc_loss=0.1254, cr_loss=0.3437, over 15299.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1174, cr_loss=0.334, over 3367689.62 frames. ], batch size: 89, lr: 2.37e-03, grad_scale: 32.0
2024-09-26 03:48:27,051 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.297e+02 1.382e+02 1.469e+02 2.239e+02, threshold=2.764e+02, percent-clipped=0.0
2024-09-26 03:49:02,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=898916.6666666666, ans=0.0
2024-09-26 03:49:10,498 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=898963.3333333334, ans=0.0
2024-09-26 03:49:19,076 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.whiten, num_groups=1, num_channels=512, metric=5.16 vs. limit=12.0
2024-09-26 03:49:42,838 INFO [train.py:1198] (1/4) Epoch 50, batch 1750, loss[loss=0.176, ctc_loss=0.1099, cr_loss=0.3304, over 17091.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1176, cr_loss=0.3344, over 3374953.72 frames. ], batch size: 43, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:50:00,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=899103.3333333334, ans=0.2
2024-09-26 03:50:10,207 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_skip_rate, batch_count=899103.3333333334, ans=0.0
2024-09-26 03:50:35,741 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.self_attn2.whiten, num_groups=1, num_channels=384, metric=11.90 vs. limit=22.5
2024-09-26 03:50:41,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=899196.6666666666, ans=0.0
2024-09-26 03:50:42,450 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.82 vs. limit=15.0
2024-09-26 03:51:05,454 INFO [train.py:1198] (1/4) Epoch 50, batch 1800, loss[loss=0.215, ctc_loss=0.1427, cr_loss=0.3611, over 16565.00 frames. ], tot_loss[loss=0.184, ctc_loss=0.1172, cr_loss=0.3339, over 3373617.21 frames. ], batch size: 66, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:51:13,299 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.324e+02 1.397e+02 1.479e+02 2.541e+02, threshold=2.793e+02, percent-clipped=0.0
2024-09-26 03:51:15,162 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=899290.0, ans=0.125
2024-09-26 03:51:31,746 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=256, metric=4.57 vs. limit=15.0
2024-09-26 03:51:32,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module1.balancer2.prob, batch_count=899336.6666666666, ans=0.125
2024-09-26 03:51:34,195 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.balancer2.prob, batch_count=899336.6666666666, ans=0.125
2024-09-26 03:51:56,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.scale_min, batch_count=899430.0, ans=0.2
2024-09-26 03:51:58,119 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_skip_rate, batch_count=899430.0, ans=0.0
2024-09-26 03:52:12,765 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=899476.6666666666, ans=0.1
2024-09-26 03:52:28,318 INFO [train.py:1198] (1/4) Epoch 50, batch 1850, loss[loss=0.2037, ctc_loss=0.1288, cr_loss=0.3744, over 16904.00 frames. ], tot_loss[loss=0.1827, ctc_loss=0.1163, cr_loss=0.3318, over 3370656.52 frames. ], batch size: 58, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:52:33,953 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=13.68 vs. limit=15.0
2024-09-26 03:52:41,804 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=899523.3333333334, ans=0.125
2024-09-26 03:52:58,060 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.self_attn1.whiten, num_groups=1, num_channels=384, metric=12.26 vs. limit=22.5
2024-09-26 03:53:19,240 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=899663.3333333334, ans=0.1
2024-09-26 03:53:21,002 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn_weights.pos_emb_skip_rate, batch_count=899663.3333333334, ans=0.0
2024-09-26 03:53:51,421 INFO [train.py:1198] (1/4) Epoch 50, batch 1900, loss[loss=0.1961, ctc_loss=0.1309, cr_loss=0.3258, over 16883.00 frames. ], tot_loss[loss=0.1827, ctc_loss=0.1164, cr_loss=0.3315, over 3369891.17 frames. ], batch size: 58, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:53:58,096 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.attention_skip_rate, batch_count=899756.6666666666, ans=0.0
2024-09-26 03:53:59,441 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.288e+02 1.384e+02 1.484e+02 2.274e+02, threshold=2.768e+02, percent-clipped=0.0
2024-09-26 03:54:55,829 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=11.53 vs. limit=15.0
2024-09-26 03:55:01,763 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=899943.3333333334, ans=0.0
2024-09-26 03:55:14,170 INFO [train.py:1198] (1/4) Epoch 50, batch 1950, loss[loss=0.1942, ctc_loss=0.1272, cr_loss=0.3349, over 17172.00 frames. ], tot_loss[loss=0.1833, ctc_loss=0.1169, cr_loss=0.3321, over 3352104.73 frames. ], batch size: 45, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:55:14,446 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=899990.0, ans=0.1
2024-09-26 03:55:39,603 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=900036.6666666666, ans=0.1
2024-09-26 03:55:39,714 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=900036.6666666666, ans=0.125
2024-09-26 03:55:47,525 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=900083.3333333334, ans=0.1
2024-09-26 03:55:52,253 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=900083.3333333334, ans=0.1
2024-09-26 03:55:52,294 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=900083.3333333334, ans=0.0
2024-09-26 03:56:04,059 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=8.52 vs. limit=15.0
2024-09-26 03:56:35,326 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.whiten, num_groups=1, num_channels=512, metric=4.27 vs. limit=12.0
2024-09-26 03:56:39,349 INFO [train.py:1198] (1/4) Epoch 50, batch 2000, loss[loss=0.1803, ctc_loss=0.1131, cr_loss=0.3357, over 17151.00 frames. ], tot_loss[loss=0.1834, ctc_loss=0.117, cr_loss=0.3321, over 3345095.10 frames. ], batch size: 45, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:56:48,773 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.333e+02 1.399e+02 1.506e+02 2.393e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-26 03:56:53,972 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=900270.0, ans=0.1
2024-09-26 03:57:03,384 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.conv.2.prob, batch_count=900270.0, ans=0.125
2024-09-26 03:57:09,915 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module1.balancer1.prob, batch_count=900316.6666666666, ans=0.125
2024-09-26 03:57:14,612 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module2.balancer1.prob, batch_count=900316.6666666666, ans=0.125
2024-09-26 03:57:17,899 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=11.44 vs. limit=12.0
2024-09-26 03:57:27,271 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.attention_skip_rate, batch_count=900363.3333333334, ans=0.0
2024-09-26 03:57:30,342 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=900363.3333333334, ans=0.125
2024-09-26 03:57:35,227 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.skip_rate, batch_count=900363.3333333334, ans=0.04949747468305833
2024-09-26 03:57:51,445 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=10.49 vs. limit=12.0
2024-09-26 03:57:51,860 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.32 vs. limit=6.0
2024-09-26 03:58:01,483 INFO [train.py:1198] (1/4) Epoch 50, batch 2050, loss[loss=0.2158, ctc_loss=0.1363, cr_loss=0.3976, over 17213.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1178, cr_loss=0.3341, over 3350897.35 frames. ], batch size: 47, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:58:19,122 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=900503.3333333334, ans=0.125
2024-09-26 03:58:39,796 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.bypass.scale_min, batch_count=900550.0, ans=0.2
2024-09-26 03:58:41,242 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.hidden_balancer.prob, batch_count=900550.0, ans=0.125
2024-09-26 03:58:44,510 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_skip_rate, batch_count=900550.0, ans=0.0
2024-09-26 03:58:46,110 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=900550.0, ans=0.1
2024-09-26 03:58:59,025 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 03:59:24,076 INFO [train.py:1198] (1/4) Epoch 50, batch 2100, loss[loss=0.1758, ctc_loss=0.1135, cr_loss=0.3113, over 17337.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1179, cr_loss=0.3342, over 3346142.48 frames. ], batch size: 48, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 03:59:27,712 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_skip_rate, batch_count=900690.0, ans=0.0
2024-09-26 03:59:33,714 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.304e+02 1.389e+02 1.531e+02 2.569e+02, threshold=2.777e+02, percent-clipped=0.0
2024-09-26 03:59:33,998 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=900690.0, ans=0.125
2024-09-26 04:00:01,321 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=900783.3333333334, ans=0.2
2024-09-26 04:00:15,959 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff3_skip_rate, batch_count=900830.0, ans=0.0
2024-09-26 04:00:23,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=900830.0, ans=0.125
2024-09-26 04:00:34,432 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=900876.6666666666, ans=0.0
2024-09-26 04:00:46,960 INFO [train.py:1198] (1/4) Epoch 50, batch 2150, loss[loss=0.2193, ctc_loss=0.1429, cr_loss=0.3823, over 15114.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3352, over 3347270.96 frames. ], batch size: 89, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 04:00:53,660 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.conv_module1.balancer2.prob, batch_count=900923.3333333334, ans=0.125
2024-09-26 04:00:56,874 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=900923.3333333334, ans=0.125
2024-09-26 04:01:31,519 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=901016.6666666666, ans=0.1
2024-09-26 04:01:55,916 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module2.balancer2.prob, batch_count=901110.0, ans=0.125
2024-09-26 04:01:57,475 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=901110.0, ans=0.0
2024-09-26 04:01:57,567 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.ff2_skip_rate, batch_count=901110.0, ans=0.0
2024-09-26 04:02:10,064 INFO [train.py:1198] (1/4) Epoch 50, batch 2200, loss[loss=0.2349, ctc_loss=0.1572, cr_loss=0.3886, over 12631.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1192, cr_loss=0.3361, over 3349957.60 frames. ], batch size: 123, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 04:02:13,526 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.bypass.scale_min, batch_count=901156.6666666666, ans=0.2
2024-09-26 04:02:19,445 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.179e+02 1.317e+02 1.377e+02 1.488e+02 2.594e+02, threshold=2.754e+02, percent-clipped=0.0
2024-09-26 04:02:38,527 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.max_abs, batch_count=901203.3333333334, ans=10.0
2024-09-26 04:03:05,166 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.const_attention_rate, batch_count=901296.6666666666, ans=0.025
2024-09-26 04:03:32,032 INFO [train.py:1198] (1/4) Epoch 50, batch 2250, loss[loss=0.2196, ctc_loss=0.148, cr_loss=0.358, over 11528.00 frames. ], tot_loss[loss=0.1864, ctc_loss=0.1191, cr_loss=0.3363, over 3352011.95 frames. ], batch size: 123, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 04:03:34,039 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=901390.0, ans=0.125
2024-09-26 04:03:48,561 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.self_attn2.whiten.whitening_limit, batch_count=901436.6666666666, ans=22.5
2024-09-26 04:03:49,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward2.hidden_balancer.prob, batch_count=901436.6666666666, ans=0.125
2024-09-26 04:04:04,551 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.5.encoder.layers.1.self_attn_weights, loss-sum=0.000e+00
2024-09-26 04:04:29,948 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=384, metric=18.42 vs. limit=22.5
2024-09-26 04:04:35,678 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass_mid.scale_min, batch_count=901530.0, ans=0.2
2024-09-26 04:04:36,410 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=3.25 vs. limit=6.0
2024-09-26 04:04:42,127 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.balancer_ff3.min_abs, batch_count=901576.6666666666, ans=0.2
2024-09-26 04:04:51,965 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.conv_module1.balancer1.prob, batch_count=901576.6666666666, ans=0.125
2024-09-26 04:04:54,690 INFO [train.py:1198] (1/4) Epoch 50, batch 2300, loss[loss=0.1587, ctc_loss=0.0955, cr_loss=0.3159, over 17262.00 frames. ], tot_loss[loss=0.1858, ctc_loss=0.1187, cr_loss=0.3355, over 3344303.07 frames. ], batch size: 42, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 04:05:02,119 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.feed_forward2.out_whiten, num_groups=1, num_channels=256, metric=3.66 vs. limit=15.0
2024-09-26 04:05:04,459 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.310e+02 1.388e+02 1.508e+02 2.945e+02, threshold=2.775e+02, percent-clipped=1.0
2024-09-26 04:05:12,771 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=901670.0, ans=0.2
2024-09-26 04:05:45,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module2.balancer2.prob, batch_count=901763.3333333334, ans=0.125
2024-09-26 04:05:48,775 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer2.prob, batch_count=901763.3333333334, ans=0.125
2024-09-26 04:05:50,645 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.32 vs. limit=15.0
2024-09-26 04:05:56,783 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.attention_skip_rate, batch_count=901763.3333333334, ans=0.0
2024-09-26 04:06:19,856 INFO [train.py:1198] (1/4) Epoch 50, batch 2350, loss[loss=0.1715, ctc_loss=0.1077, cr_loss=0.3188, over 17165.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3353, over 3346611.13 frames. ], batch size: 45, lr: 2.37e-03, grad_scale: 16.0
2024-09-26 04:07:12,841 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.ff2_skip_rate, batch_count=901996.6666666666, ans=0.0
2024-09-26 04:07:19,203 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.balancer1.prob, batch_count=901996.6666666666, ans=0.125
2024-09-26 04:07:39,158 INFO [train.py:1198] (1/4) Epoch 50, batch 2400, loss[loss=0.1994, ctc_loss=0.1267, cr_loss=0.3636, over 16566.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.1181, cr_loss=0.3346, over 3357637.65 frames. ], batch size: 66, lr: 2.37e-03, grad_scale: 32.0
2024-09-26 04:07:46,288 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.nonlin_attention.whiten1, num_groups=1, num_channels=144, metric=10.31 vs. limit=10.0
2024-09-26 04:07:51,372 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.310e+02 1.369e+02 1.448e+02 2.051e+02, threshold=2.738e+02, percent-clipped=0.0
2024-09-26 04:07:55,815 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=13.56 vs. limit=15.0
2024-09-26 04:08:01,441 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.attention_skip_rate, batch_count=902136.6666666666, ans=0.0
2024-09-26 04:08:17,379 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=902183.3333333334, ans=0.125
2024-09-26 04:08:41,979 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer2.prob, batch_count=902230.0, ans=0.125
2024-09-26 04:09:02,438 INFO [train.py:1198] (1/4) Epoch 50, batch 2450, loss[loss=0.1387, ctc_loss=0.08606, cr_loss=0.263, over 17088.00 frames. ], tot_loss[loss=0.184, ctc_loss=0.1173, cr_loss=0.3332, over 3361311.32 frames. ], batch size: 43, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:09:16,392 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=902323.3333333334, ans=0.125
2024-09-26 04:09:22,842 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=902370.0, ans=0.2
2024-09-26 04:09:39,017 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=3.67 vs. limit=15.0
2024-09-26 04:10:02,770 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.4.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 04:10:10,604 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass.scale_min, batch_count=902510.0, ans=0.2
2024-09-26 04:10:27,376 INFO [train.py:1198] (1/4) Epoch 50, batch 2500, loss[loss=0.1652, ctc_loss=0.102, cr_loss=0.3162, over 16329.00 frames. ], tot_loss[loss=0.1844, ctc_loss=0.1176, cr_loss=0.3338, over 3359724.99 frames. ], batch size: 36, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:10:34,033 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=902556.6666666666, ans=0.1
2024-09-26 04:10:37,374 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=256, metric=12.18 vs. limit=15.0
2024-09-26 04:10:38,429 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.330e+02 1.408e+02 1.529e+02 2.285e+02, threshold=2.816e+02, percent-clipped=0.0
2024-09-26 04:10:56,346 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.const_attention_rate, batch_count=902603.3333333334, ans=0.025
2024-09-26 04:11:49,379 INFO [train.py:1198] (1/4) Epoch 50, batch 2550, loss[loss=0.1562, ctc_loss=0.09771, cr_loss=0.2922, over 16946.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1181, cr_loss=0.3348, over 3360775.50 frames. ], batch size: 42, lr: 2.36e-03, grad_scale: 8.0
2024-09-26 04:12:26,108 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=256, metric=4.20 vs. limit=15.0
2024-09-26 04:12:34,941 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.balancer2.prob, batch_count=902883.3333333334, ans=0.125
2024-09-26 04:12:47,360 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.conv_module2.balancer2.prob, batch_count=902930.0, ans=0.125
2024-09-26 04:12:52,115 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=902930.0, ans=0.025
2024-09-26 04:13:02,781 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder_embed.convnext.hidden_balancer.prob, batch_count=902976.6666666666, ans=0.125
2024-09-26 04:13:12,204 INFO [train.py:1198] (1/4) Epoch 50, batch 2600, loss[loss=0.1863, ctc_loss=0.1223, cr_loss=0.3199, over 17147.00 frames. ], tot_loss[loss=0.1859, ctc_loss=0.1188, cr_loss=0.3354, over 3349108.63 frames. ], batch size: 48, lr: 2.36e-03, grad_scale: 8.0
2024-09-26 04:13:14,016 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.const_attention_rate, batch_count=903023.3333333334, ans=0.025
2024-09-26 04:13:25,160 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.312e+02 1.402e+02 1.467e+02 1.641e+02, threshold=2.804e+02, percent-clipped=0.0
2024-09-26 04:13:52,937 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer2.prob, batch_count=903116.6666666666, ans=0.125
2024-09-26 04:13:56,036 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=903116.6666666666, ans=0.0
2024-09-26 04:14:14,531 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer2.min_positive, batch_count=903163.3333333334, ans=0.05
2024-09-26 04:14:24,246 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward2.hidden_balancer.prob, batch_count=903210.0, ans=0.125
2024-09-26 04:14:25,650 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.feed_forward1.out_proj.dropout_p, batch_count=903210.0, ans=0.1
2024-09-26 04:14:27,871 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.self_attn1.whiten.whitening_limit, batch_count=903210.0, ans=22.5
2024-09-26 04:14:34,856 INFO [train.py:1198] (1/4) Epoch 50, batch 2650, loss[loss=0.1698, ctc_loss=0.1082, cr_loss=0.3079, over 17264.00 frames. ], tot_loss[loss=0.1857, ctc_loss=0.1187, cr_loss=0.335, over 3348888.18 frames. ], batch size: 46, lr: 2.36e-03, grad_scale: 8.0
2024-09-26 04:14:38,285 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer1.prob, batch_count=903256.6666666666, ans=0.125
2024-09-26 04:14:59,069 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 04:15:41,620 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.nonlin_attention.whiten2.whitening_limit, batch_count=903443.3333333334, ans=15.0
2024-09-26 04:15:42,968 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer1.prob, batch_count=903443.3333333334, ans=0.125
2024-09-26 04:15:51,380 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module1.balancer1.prob, batch_count=903443.3333333334, ans=0.125
2024-09-26 04:15:56,662 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.whiten, num_groups=1, num_channels=384, metric=3.76 vs. limit=12.0
2024-09-26 04:15:57,314 INFO [train.py:1198] (1/4) Epoch 50, batch 2700, loss[loss=0.1491, ctc_loss=0.09388, cr_loss=0.2759, over 17047.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.118, cr_loss=0.3335, over 3342836.54 frames. ], batch size: 39, lr: 2.36e-03, grad_scale: 8.0
2024-09-26 04:16:12,403 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.330e+02 1.437e+02 1.530e+02 2.387e+02, threshold=2.875e+02, percent-clipped=0.0
2024-09-26 04:16:41,663 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward2.hidden_balancer.prob, batch_count=903583.3333333334, ans=0.125
2024-09-26 04:16:43,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer1.prob, batch_count=903583.3333333334, ans=0.125
2024-09-26 04:17:10,729 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module2.balancer2.prob, batch_count=903676.6666666666, ans=0.125
2024-09-26 04:17:14,185 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=903676.6666666666, ans=0.0
2024-09-26 04:17:15,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=903676.6666666666, ans=0.125
2024-09-26 04:17:20,229 INFO [train.py:1198] (1/4) Epoch 50, batch 2750, loss[loss=0.2173, ctc_loss=0.1422, cr_loss=0.3756, over 16461.00 frames. ], tot_loss[loss=0.1838, ctc_loss=0.1173, cr_loss=0.3327, over 3350755.41 frames. ], batch size: 66, lr: 2.36e-03, grad_scale: 8.0
2024-09-26 04:17:28,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.bypass.scale_min, batch_count=903723.3333333334, ans=0.2
2024-09-26 04:17:28,381 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.attention_skip_rate, batch_count=903723.3333333334, ans=0.0
2024-09-26 04:17:39,065 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.conv_module1.balancer2.prob, batch_count=903770.0, ans=0.125
2024-09-26 04:18:11,141 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.ff2_skip_rate, batch_count=903863.3333333334, ans=0.0
2024-09-26 04:18:12,645 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.nonlin_attention.balancer.prob, batch_count=903863.3333333334, ans=0.125
2024-09-26 04:18:35,257 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.const_attention_rate, batch_count=903910.0, ans=0.025
2024-09-26 04:18:42,701 INFO [train.py:1198] (1/4) Epoch 50, batch 2800, loss[loss=0.1554, ctc_loss=0.09467, cr_loss=0.3039, over 17120.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1177, cr_loss=0.3341, over 3356338.84 frames. ], batch size: 40, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:18:49,255 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=903956.6666666666, ans=0.0
2024-09-26 04:18:50,898 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.pos_emb_skip_rate, batch_count=903956.6666666666, ans=0.0
2024-09-26 04:18:57,718 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.318e+02 1.422e+02 1.529e+02 1.768e+02, threshold=2.845e+02, percent-clipped=0.0
2024-09-26 04:19:02,764 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.out_combiner.scale_min, batch_count=904003.3333333334, ans=0.2
2024-09-26 04:19:20,987 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=12.52 vs. limit=15.0
2024-09-26 04:19:22,156 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.ff3_skip_rate, batch_count=904050.0, ans=0.0
2024-09-26 04:19:25,235 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=904050.0, ans=0.125
2024-09-26 04:19:30,765 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.2.conv_module1.whiten, num_groups=1, num_channels=384, metric=4.01 vs. limit=15.0
2024-09-26 04:19:33,343 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=904096.6666666666, ans=0.1
2024-09-26 04:19:34,709 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=904096.6666666666, ans=0.125
2024-09-26 04:19:34,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.out_combiner.scale_min, batch_count=904096.6666666666, ans=0.2
2024-09-26 04:19:52,928 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=5.28 vs. limit=15.0
2024-09-26 04:19:57,191 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_skip_rate, batch_count=904143.3333333334, ans=0.0
2024-09-26 04:19:58,578 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass_mid.scale_min, batch_count=904143.3333333334, ans=0.2
2024-09-26 04:20:04,785 INFO [train.py:1198] (1/4) Epoch 50, batch 2850, loss[loss=0.1643, ctc_loss=0.1031, cr_loss=0.3059, over 17068.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1175, cr_loss=0.3333, over 3356832.90 frames. ], batch size: 46, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:20:13,058 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.conv_module1.whiten, num_groups=1, num_channels=192, metric=8.44 vs. limit=15.0
2024-09-26 04:20:34,577 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.conv_module2.balancer1.prob, batch_count=904236.6666666666, ans=0.125
2024-09-26 04:20:46,211 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.self_attn_weights.whiten_keys, num_groups=8, num_channels=256, metric=2.37 vs. limit=6.0
2024-09-26 04:20:47,419 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_skip_rate, batch_count=904283.3333333334, ans=0.0
2024-09-26 04:21:29,402 INFO [train.py:1198] (1/4) Epoch 50, batch 2900, loss[loss=0.1978, ctc_loss=0.1264, cr_loss=0.3569, over 17324.00 frames. ], tot_loss[loss=0.1861, ctc_loss=0.1189, cr_loss=0.3358, over 3342159.01 frames. ], batch size: 51, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:21:42,284 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.318e+02 1.386e+02 1.483e+02 2.505e+02, threshold=2.771e+02, percent-clipped=0.0
2024-09-26 04:21:53,035 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=4.90 vs. limit=15.0
2024-09-26 04:22:16,061 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=904563.3333333334, ans=0.125
2024-09-26 04:22:34,778 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=904610.0, ans=0.0
2024-09-26 04:22:39,410 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.feed_forward1.hidden_balancer.prob, batch_count=904610.0, ans=0.125
2024-09-26 04:22:44,286 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.balancer.prob, batch_count=904610.0, ans=0.125
2024-09-26 04:22:50,867 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.bypass.skip_rate, batch_count=904656.6666666666, ans=0.09899494936611666
2024-09-26 04:22:52,047 INFO [train.py:1198] (1/4) Epoch 50, batch 2950, loss[loss=0.1916, ctc_loss=0.1215, cr_loss=0.3507, over 17237.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1182, cr_loss=0.3346, over 3341085.75 frames. ], batch size: 50, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:23:28,306 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.62 vs. limit=15.0
2024-09-26 04:23:35,882 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.const_attention_rate, batch_count=904750.0, ans=0.025
2024-09-26 04:23:40,210 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=6.97 vs. limit=8.0
2024-09-26 04:23:43,966 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer2.prob, batch_count=904796.6666666666, ans=0.125
2024-09-26 04:23:50,043 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=904796.6666666666, ans=0.125
2024-09-26 04:24:14,375 INFO [train.py:1198] (1/4) Epoch 50, batch 3000, loss[loss=0.215, ctc_loss=0.1409, cr_loss=0.3702, over 15520.00 frames. ], tot_loss[loss=0.186, ctc_loss=0.1189, cr_loss=0.3358, over 3351683.22 frames. ], batch size: 89, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:24:14,376 INFO [train.py:1221] (1/4) Computing validation loss
2024-09-26 04:24:26,601 INFO [zipformer.py:1858] (1/4) name=encoder.encoders.3.encoder.layers.0.self_attn_weights, attn_weights_entropy = tensor([1.6705, 2.3057, 2.3891, 2.3679, 2.4766, 2.1036, 2.3048, 1.6893],
       device='cuda:1')
2024-09-26 04:24:30,369 INFO [train.py:1230] (1/4) Epoch 50, validation: loss=0.03495, ctc_loss=0.03495, cr_loss=1.037e-14, over 944034.00 frames. 
2024-09-26 04:24:30,370 INFO [train.py:1231] (1/4) Maximum memory allocated so far is 21613MB
2024-09-26 04:24:41,774 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.ff2_skip_rate, batch_count=904890.0, ans=0.0
2024-09-26 04:24:42,933 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.332e+02 1.410e+02 1.508e+02 3.404e+02, threshold=2.821e+02, percent-clipped=1.0
2024-09-26 04:24:48,029 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.self_attn_weights.pos_emb_skip_rate, batch_count=904936.6666666666, ans=0.0
2024-09-26 04:25:02,444 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=256, metric=13.22 vs. limit=15.0
2024-09-26 04:25:31,886 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_module1.balancer1.prob, batch_count=905076.6666666666, ans=0.125
2024-09-26 04:25:42,694 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.conv_module2.balancer1.prob, batch_count=905076.6666666666, ans=0.125
2024-09-26 04:25:48,679 INFO [train.py:1198] (1/4) Epoch 50, batch 3050, loss[loss=0.2163, ctc_loss=0.141, cr_loss=0.3762, over 15056.00 frames. ], tot_loss[loss=0.1862, ctc_loss=0.1189, cr_loss=0.3366, over 3358608.65 frames. ], batch size: 89, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:27:06,908 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.whiten, num_groups=1, num_channels=512, metric=4.52 vs. limit=12.0
2024-09-26 04:27:09,366 INFO [train.py:1198] (1/4) Epoch 50, batch 3100, loss[loss=0.194, ctc_loss=0.1235, cr_loss=0.3522, over 17295.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1179, cr_loss=0.3342, over 3366548.40 frames. ], batch size: 49, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:27:15,834 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.scale_min, batch_count=905356.6666666666, ans=0.2
2024-09-26 04:27:21,791 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.333e+02 1.400e+02 1.486e+02 1.966e+02, threshold=2.799e+02, percent-clipped=0.0
2024-09-26 04:27:43,853 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=905450.0, ans=0.1
2024-09-26 04:27:48,596 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.const_attention_rate, batch_count=905450.0, ans=0.025
2024-09-26 04:27:55,430 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=905450.0, ans=0.0
2024-09-26 04:28:03,897 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=384, metric=8.78 vs. limit=15.0
2024-09-26 04:28:13,271 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.0.whiten, num_groups=1, num_channels=256, metric=12.65 vs. limit=12.0
2024-09-26 04:28:14,508 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.feed_forward1.out_proj.dropout_p, batch_count=905543.3333333334, ans=0.1
2024-09-26 04:28:19,080 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.attention_skip_rate, batch_count=905543.3333333334, ans=0.0
2024-09-26 04:28:27,180 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.conv_module2.balancer2.prob, batch_count=905543.3333333334, ans=0.125
2024-09-26 04:28:29,961 INFO [train.py:1198] (1/4) Epoch 50, batch 3150, loss[loss=0.186, ctc_loss=0.123, cr_loss=0.3151, over 17011.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1186, cr_loss=0.3353, over 3369050.95 frames. ], batch size: 56, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:28:53,252 INFO [scaling.py:1024] (1/4) Whitening: name=encoder_embed.out_whiten, num_groups=1, num_channels=192, metric=7.25 vs. limit=8.0
2024-09-26 04:28:56,914 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.prob, batch_count=905636.6666666666, ans=0.125
2024-09-26 04:29:26,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=905730.0, ans=0.5
2024-09-26 04:29:48,814 INFO [train.py:1198] (1/4) Epoch 50, batch 3200, loss[loss=0.1876, ctc_loss=0.1208, cr_loss=0.3342, over 17364.00 frames. ], tot_loss[loss=0.1846, ctc_loss=0.1178, cr_loss=0.334, over 3372916.62 frames. ], batch size: 48, lr: 2.36e-03, grad_scale: 32.0
2024-09-26 04:29:53,719 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=905823.3333333334, ans=0.125
2024-09-26 04:30:01,215 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.309e+02 1.397e+02 1.487e+02 2.037e+02, threshold=2.793e+02, percent-clipped=0.0
2024-09-26 04:30:09,481 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.conv_skip_rate, batch_count=905870.0, ans=0.0
2024-09-26 04:30:11,186 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=905870.0, ans=0.1
2024-09-26 04:30:34,833 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.1.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 04:30:36,340 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=905963.3333333334, ans=0.125
2024-09-26 04:30:51,280 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.whiten, num_groups=1, num_channels=256, metric=4.00 vs. limit=12.0
2024-09-26 04:31:07,458 INFO [train.py:1198] (1/4) Epoch 50, batch 3250, loss[loss=0.1949, ctc_loss=0.1271, cr_loss=0.3393, over 16991.00 frames. ], tot_loss[loss=0.185, ctc_loss=0.1181, cr_loss=0.3343, over 3363672.32 frames. ], batch size: 56, lr: 2.36e-03, grad_scale: 32.0
2024-09-26 04:31:20,146 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer2.min_abs, batch_count=906056.6666666666, ans=0.5
2024-09-26 04:31:27,017 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.bypass.scale_min, batch_count=906103.3333333334, ans=0.2
2024-09-26 04:31:28,386 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.0.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 04:32:09,615 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=192, metric=3.84 vs. limit=10.0
2024-09-26 04:32:27,560 INFO [train.py:1198] (1/4) Epoch 50, batch 3300, loss[loss=0.187, ctc_loss=0.1203, cr_loss=0.3337, over 17314.00 frames. ], tot_loss[loss=0.1851, ctc_loss=0.1182, cr_loss=0.3346, over 3351083.33 frames. ], batch size: 49, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:32:41,661 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.281e+02 1.376e+02 1.500e+02 2.072e+02, threshold=2.752e+02, percent-clipped=0.0
2024-09-26 04:33:07,132 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.balancer2.prob, batch_count=906383.3333333334, ans=0.125
2024-09-26 04:33:46,249 INFO [train.py:1198] (1/4) Epoch 50, batch 3350, loss[loss=0.1972, ctc_loss=0.1248, cr_loss=0.362, over 16985.00 frames. ], tot_loss[loss=0.1847, ctc_loss=0.1178, cr_loss=0.3344, over 3354104.24 frames. ], batch size: 58, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:34:38,820 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.bypass.skip_rate, batch_count=906663.3333333334, ans=0.035
2024-09-26 04:34:40,496 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=906663.3333333334, ans=0.025
2024-09-26 04:34:46,890 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.2.self_attn_weights, loss-sum=0.000e+00
2024-09-26 04:35:06,873 INFO [train.py:1198] (1/4) Epoch 50, batch 3400, loss[loss=0.1888, ctc_loss=0.1211, cr_loss=0.3383, over 16955.00 frames. ], tot_loss[loss=0.1845, ctc_loss=0.1177, cr_loss=0.3343, over 3351287.51 frames. ], batch size: 42, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:35:20,635 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.315e+02 1.394e+02 1.507e+02 2.409e+02, threshold=2.789e+02, percent-clipped=0.0
2024-09-26 04:35:27,366 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.out_combiner.scale_min, batch_count=906803.3333333334, ans=0.2
2024-09-26 04:35:42,784 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.bypass_mid.scale_min, batch_count=906850.0, ans=0.2
2024-09-26 04:36:14,675 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.1.feed_forward3.out_whiten, num_groups=1, num_channels=512, metric=8.76 vs. limit=15.0
2024-09-26 04:36:24,724 INFO [train.py:1198] (1/4) Epoch 50, batch 3450, loss[loss=0.1569, ctc_loss=0.1018, cr_loss=0.2755, over 17197.00 frames. ], tot_loss[loss=0.1849, ctc_loss=0.118, cr_loss=0.3342, over 3350926.47 frames. ], batch size: 41, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:36:26,641 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.conv_module1.balancer1.prob, batch_count=906990.0, ans=0.125
2024-09-26 04:36:42,671 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward3.hidden_balancer.prob, batch_count=907036.6666666666, ans=0.125
2024-09-26 04:37:12,453 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.0.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.04 vs. limit=6.0
2024-09-26 04:37:18,270 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.feed_forward1.out_proj.dropout_p, batch_count=907130.0, ans=0.1
2024-09-26 04:37:35,473 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.bypass.scale_min, batch_count=907176.6666666666, ans=0.2
2024-09-26 04:37:44,968 INFO [train.py:1198] (1/4) Epoch 50, batch 3500, loss[loss=0.2167, ctc_loss=0.1462, cr_loss=0.3525, over 11746.00 frames. ], tot_loss[loss=0.1839, ctc_loss=0.1174, cr_loss=0.3328, over 3344667.21 frames. ], batch size: 123, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:37:58,718 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.340e+02 1.405e+02 1.556e+02 2.203e+02, threshold=2.811e+02, percent-clipped=0.0
2024-09-26 04:38:24,439 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.1.bypass.skip_rate, batch_count=907316.6666666666, ans=0.035
2024-09-26 04:38:26,178 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=907316.6666666666, ans=0.125
2024-09-26 04:38:32,355 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.nonlin_attention.balancer.prob, batch_count=907363.3333333334, ans=0.125
2024-09-26 04:38:32,462 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.balancer1.prob, batch_count=907363.3333333334, ans=0.125
2024-09-26 04:38:48,545 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten2, num_groups=1, num_channels=512, metric=4.82 vs. limit=15.0
2024-09-26 04:39:05,112 INFO [train.py:1198] (1/4) Epoch 50, batch 3550, loss[loss=0.1819, ctc_loss=0.1157, cr_loss=0.3312, over 17303.00 frames. ], tot_loss[loss=0.1841, ctc_loss=0.1174, cr_loss=0.3334, over 3349099.51 frames. ], batch size: 49, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:39:22,461 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=907503.3333333334, ans=0.0
2024-09-26 04:39:31,616 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.2.conv_module2.balancer1.min_positive, batch_count=907503.3333333334, ans=0.025
2024-09-26 04:39:34,685 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.conv_module2.balancer2.prob, batch_count=907550.0, ans=0.125
2024-09-26 04:39:42,541 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.attention_skip_rate, batch_count=907550.0, ans=0.0
2024-09-26 04:39:44,469 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.5.encoder.layers.1.conv_module2.whiten, num_groups=1, num_channels=256, metric=9.11 vs. limit=15.0
2024-09-26 04:39:48,912 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.ff3_skip_rate, batch_count=907550.0, ans=0.0
2024-09-26 04:40:22,837 INFO [train.py:1198] (1/4) Epoch 50, batch 3600, loss[loss=0.1683, ctc_loss=0.1075, cr_loss=0.3039, over 17300.00 frames. ], tot_loss[loss=0.1842, ctc_loss=0.1175, cr_loss=0.3333, over 3348599.36 frames. ], batch size: 51, lr: 2.36e-03, grad_scale: 32.0
2024-09-26 04:40:23,009 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.self_attn_weights.pos_emb_skip_rate, batch_count=907690.0, ans=0.0
2024-09-26 04:40:36,919 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.290e+02 1.354e+02 1.454e+02 2.078e+02, threshold=2.707e+02, percent-clipped=0.0
2024-09-26 04:40:47,975 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.3.ff3_skip_rate, batch_count=907736.6666666666, ans=0.0
2024-09-26 04:41:00,435 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.0.feed_forward1.out_whiten, num_groups=1, num_channels=192, metric=9.70 vs. limit=15.0
2024-09-26 04:41:01,041 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=907783.3333333334, ans=0.1
2024-09-26 04:41:05,568 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.balancer1.prob, batch_count=907783.3333333334, ans=0.125
2024-09-26 04:41:09,582 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.0.layers.1.self_attn_weights.whiten_keys, num_groups=4, num_channels=128, metric=2.72 vs. limit=6.0
2024-09-26 04:41:29,088 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.2.conv_module1.balancer1.prob, batch_count=907876.6666666666, ans=0.125
2024-09-26 04:41:40,073 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module1.balancer2.prob, batch_count=907876.6666666666, ans=0.125
2024-09-26 04:41:42,942 INFO [train.py:1198] (1/4) Epoch 50, batch 3650, loss[loss=0.1938, ctc_loss=0.1257, cr_loss=0.3407, over 17297.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.118, cr_loss=0.3342, over 3351604.53 frames. ], batch size: 46, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:41:46,301 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=907923.3333333334, ans=0.125
2024-09-26 04:42:00,792 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.3.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=4.51 vs. limit=10.0
2024-09-26 04:42:11,181 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.balancer1.prob, batch_count=907970.0, ans=0.125
2024-09-26 04:42:34,597 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.0.bypass_mid.scale_min, batch_count=908063.3333333334, ans=0.2
2024-09-26 04:43:01,636 INFO [train.py:1198] (1/4) Epoch 50, batch 3700, loss[loss=0.1862, ctc_loss=0.1205, cr_loss=0.3286, over 17344.00 frames. ], tot_loss[loss=0.184, ctc_loss=0.1173, cr_loss=0.3334, over 3353387.35 frames. ], batch size: 48, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:43:12,989 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.out_combiner.scale_min, batch_count=908156.6666666666, ans=0.2
2024-09-26 04:43:17,256 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.236e+02 1.309e+02 1.373e+02 1.463e+02 1.965e+02, threshold=2.746e+02, percent-clipped=0.0
2024-09-26 04:43:17,868 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.4.encoder.layers.2.feed_forward3.out_whiten, num_groups=1, num_channels=384, metric=15.25 vs. limit=15.0
2024-09-26 04:43:20,615 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.1.feed_forward1.hidden_balancer.prob, batch_count=908203.3333333334, ans=0.125
2024-09-26 04:43:44,198 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.feed_forward1.out_proj.dropout_p, batch_count=908250.0, ans=0.1
2024-09-26 04:43:45,900 INFO [scaling.py:1120] (1/4) WithLoss: name=encoder.encoders.3.encoder.layers.0.self_attn_weights, loss-sum=0.000e+00
2024-09-26 04:43:56,281 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.0.feed_forward1.hidden_balancer.prob, batch_count=908296.6666666666, ans=0.125
2024-09-26 04:44:19,038 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.0.self_attn2.whiten, num_groups=1, num_channels=256, metric=11.70 vs. limit=22.5
2024-09-26 04:44:21,191 INFO [train.py:1198] (1/4) Epoch 50, batch 3750, loss[loss=0.1913, ctc_loss=0.1227, cr_loss=0.343, over 17236.00 frames. ], tot_loss[loss=0.1848, ctc_loss=0.1179, cr_loss=0.3346, over 3352445.40 frames. ], batch size: 50, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:44:28,206 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.feed_forward2.out_whiten, num_groups=1, num_channels=512, metric=4.97 vs. limit=15.0
2024-09-26 04:44:47,905 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.4.encoder.layers.1.bypass.scale_min, batch_count=908436.6666666666, ans=0.2
2024-09-26 04:45:00,463 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.2.encoder.layers.0.const_attention_rate, batch_count=908483.3333333334, ans=0.025
2024-09-26 04:45:05,189 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module2.balancer1.min_positive, batch_count=908483.3333333334, ans=0.025
2024-09-26 04:45:39,937 INFO [train.py:1198] (1/4) Epoch 50, batch 3800, loss[loss=0.1831, ctc_loss=0.1145, cr_loss=0.3433, over 16930.00 frames. ], tot_loss[loss=0.1856, ctc_loss=0.1185, cr_loss=0.3355, over 3356577.74 frames. ], batch size: 42, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:45:40,241 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.feed_forward1.out_proj.dropout_p, batch_count=908623.3333333334, ans=0.1
2024-09-26 04:45:41,785 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.balancer2.prob, batch_count=908623.3333333334, ans=0.125
2024-09-26 04:45:48,619 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.0.conv_module2.whiten, num_groups=1, num_channels=512, metric=3.50 vs. limit=15.0
2024-09-26 04:45:55,606 WARNING [optim.py:487] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.341e+02 1.416e+02 1.505e+02 2.139e+02, threshold=2.833e+02, percent-clipped=0.0
2024-09-26 04:45:59,114 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.5.encoder.layers.1.feed_forward3.hidden_balancer.prob, batch_count=908670.0, ans=0.125
2024-09-26 04:46:05,777 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.1.nonlin_attention.whiten1, num_groups=1, num_channels=288, metric=4.36 vs. limit=10.0
2024-09-26 04:46:11,179 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.1.encoder.layers.1.conv_module1.balancer2.prob, batch_count=908716.6666666666, ans=0.125
2024-09-26 04:46:26,206 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.conv_module2.whiten, num_groups=1, num_channels=512, metric=8.91 vs. limit=15.0
2024-09-26 04:46:36,513 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.2.conv_module2.balancer1.prob, batch_count=908763.3333333334, ans=0.125
2024-09-26 04:46:50,426 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_skip_rate, batch_count=908810.0, ans=0.0
2024-09-26 04:46:55,882 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.3.encoder.layers.2.nonlin_attention.whiten1, num_groups=1, num_channels=384, metric=5.95 vs. limit=10.0
2024-09-26 04:46:58,242 INFO [train.py:1198] (1/4) Epoch 50, batch 3850, loss[loss=0.2044, ctc_loss=0.135, cr_loss=0.3473, over 11872.00 frames. ], tot_loss[loss=0.1867, ctc_loss=0.1194, cr_loss=0.3362, over 3312831.29 frames. ], batch size: 124, lr: 2.36e-03, grad_scale: 16.0
2024-09-26 04:47:00,093 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.nonlin_attention.balancer.prob, batch_count=908856.6666666666, ans=0.125
2024-09-26 04:47:17,710 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.1.balancer2.prob, batch_count=908903.3333333334, ans=0.125
2024-09-26 04:47:28,999 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.3.encoder.layers.0.conv_module1.balancer1.prob, batch_count=908950.0, ans=0.125
2024-09-26 04:47:38,893 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.1.encoder.layers.1.nonlin_attention.whiten2, num_groups=1, num_channels=256, metric=8.08 vs. limit=15.0
2024-09-26 04:47:56,756 INFO [scaling.py:214] (1/4) ScheduledFloat: name=encoder.encoders.0.layers.0.conv_module2.balancer2.prob, batch_count=908996.6666666666, ans=0.125
2024-09-26 04:48:00,464 INFO [scaling.py:1024] (1/4) Whitening: name=encoder.encoders.2.encoder.layers.0.self_attn1.whiten, num_groups=1, num_channels=384, metric=17.52 vs. limit=22.5
2024-09-26 04:48:10,118 INFO [train.py:1496] (1/4) Done!